EPA-460/3-76-011
April 1976
         FEDERAL TEST PROCEDURE
                       AND SHORT TEST
             CORRELATION ANALYSES
      U.S. ENVIRONMENTAL PROTECTION AGENCY
           Office of Air and Waste Management
        Office of Mobile Source Air Pollution Control
          Emission Control Technology Division
              Ann Arbor, Michigan 48105

-------
                                  EPA-460/3-76-011
FEDERAL TEST PROCEDURE
         AND SHORT TEST
  CORRELATION  ANALYSES
                     by

                Mobile Systems Group
                Aerospace Corporation
                  P.O. Box 92957
              Los Angeles, California 90009
                Contract No. 68-01-0417
           EPA Project Officer: F. Peter Hutchins
                  Prepared for

         ENVIRONMENTAL PROTECTION AGENCY
            Office of Air and Waste Management
         Office of Mobile Source Air Pollution Control
           Emission Control Technology Division
              Ann Arbor, Michigan 48105

                   April 1976

-------
This report is issued by the Environmental Protection Agency to report
technical data of interest to a limited number of readers.  Copies are
available free of charge to Federal employees, current contractors and
grantees, and nonprofit organizations - as supplies permit - from the
Air Pollution  Technical Information Center, Environmental Protection
Agency,  Research Triangle Park, North Carolina 27711; or,  for a fee,
from the National Technical Information Service, 5285 Port Royal Road,
Springfield, Virginia 22161.
This report was furnished to the Environmental Protection Agency by
Aerospace Corporation, Los Angeles California 90009. in fulfillment
of Contract No. 68-01-0417.  The contents of this report are reproduced
herein as received from Aerospace Corpcration.  The opinions, findings,
and conclusions expressed are those of the author and not necessarily
those of the Environmental Protection Agency. Mention of company or
product names is not to be considered as an endorsement by the Environmen-
tal Protection Agency.
               Publication No. EPA-460/3-76-011
                                    ii

-------
                               FOREWORD
              This report, prepared by The Aerospace Corporation for
the U.S. Environmental Protection Agency, Emission Control Technology
Division,  presents the results of a statistical analysis of the degree of
correlation between five short tests and the 1975 Federal Test Procedure.
The correlation analyses were based on experimental test data from 147
1974-model-year vehicles, composed of three inertia test weight groups,
and on 40 catalyst-equipped experimental vehicles.
              The results of the study are presented in six sections.
Section 1  contains a summary of the  study  results.  The background, scope,
objectives, and  method of approach are given in Section 2.  The short tests,
test conditions,  and test fleet composition  are described and discussed in
Section 3. Section 4 describes the data-screening procedures, the primary
statistical tools used in the correlation analyses, and results of the
statistical analysis in detail for the catalyst-equipped experimental vehicle
fleet.  Similar results for the 1974 model year in-use fleet and a five-
vehicle defect test fleet are presented in Sections 5 and 6, respectively.
              The conduct of this analysis effort resulted in over 1000
pages of correlation table printouts,  regression plots, scattergrams,  etc.
This information is  summarized in the tables and figures presented in the
report; the voluminous  printout material is not included in order to enhance
the readability of the report.  However, the printout material is  on file at
the Emission Control Technology Division  of EPA, Ann Arbor, Michigan,
and may be borrowed for limited periods for reproduction for purposes of
detailed examination.
                                    111

-------
                          ACKNOWLEDGMENT
               During the course of this study,  Mr. F.  P.  Hutchins of the
Environmental Protection Agency's Emission Control Technology Division,
who served as EPA Project Officer for the study,  provided valuable guidance
and assistance.  His efforts  are gratefully acknowledged.
               Dr.  John Thacker was principally responsible for the statis-
tical analysis effort reported herein.  The following technical personnel of
The Aerospace Corporation also made valuable contributions to the analyses
performed under this contract:
                              W.  B.  Lee
                              R. F. Janz
                              A. M.  Timmer
                                       M. G. Hinton, Group Director
                                       Mobile Systems
Approved by:
      Itzer, General Manager
    fronment and Energy
   Snservation Division

-------
                               CONTENTS
1.    SUMMARY	      1_1
      1. 1    Catalyst-Equipped Experimental Vehicle Fleet  ....      1-2
             1. 1. 1    Direct Relatability Results	      1-2
             1. 1. 2    Contingency Table Analysis Results	      1-9
      1.2    In-Use 1974 Model Year Vehicle Fleet  	      1-26
             1. 2. 1    Direct Relatability Results	      1-26
             1.2.2    Contingency Table Analysis Results	      1-32
      1. 3    Defect Data From Catalyst-Equipped Experimental
             Vehicle Fleet	      1-42
             1. 3. 1    Nature of Defects and Statistical Impact. . .      1-42
             1. 3. 2    Contingency Table Analysis Results	      1-42
      1.4    General Overview Remarks	      1-47
             1. 4. 1    Mode  vs Bag ST	      1-47
             1.4. 2    Single Mode vs Weighted Mode Tests	      1-47
             1. 4. 3    Garage Instrument vs Laboratory
                      Analyzer	      1-47
             1. 4. 4    Correlation Coefficient vs Contingency
                      Table Analysis	      1-48
             1. 4. 5    Relative Impact on Air Quality	      1-49
2.    INTRODUCTION	      2-1
      2. 1    Background and Objectives	      2-1
      2.2    Study Scope	      2-2
      2.3    Method of Approach	      2-4
      2.4    Organization of Report	      2-5
3.    TEST CHARACTERISTICS AND PROCEDURES	      3-1
      3. 1    Short Tests	      3-1
             3. 1. 1    General	      3-1
             3. 1. 2    ST Definition	      3-2
             3. 1. 3    Short  Test  Sequence	      3-6
                                    vn

-------
                          CONTENTS (Continued)
      3. 2    Test Fleets	      3-7
             3. 2. i    Catalyst-Equipped Experimental
                      Vehicle Fleet (CEV)	      3-7
             3. 2. 2    Defect Test Fleet	      3-8
             3.2.3    In-Use 1974 Model Year Vehicle Fleet  ...      3-9
4.    CATALYST-EQUIPPED EXPERIMENTAL VEHICLE
      FLEET	      4-1
      4. 1    Preliminary Statistical Analyses	      4-1
             4. 1. 1    Data Screening	      4-1
             4. i. 2    Multivariate Analysis of Variance	      4-3
             4. 1. 3    Canonical Correlation Analysis	      4-5
             4. 1. 4    Summary  of Preliminary Analysis
                      Results	      4-7
      4. 2    Principal Statistical Analysis Techniques and
             Results	      4-8
             4. 2. 1    Correlation Analysis	      4-8
             4. 2. 2    Contingency Table Analysis	      4-17
      4. 3    References for Section 4	      4-97
5.    IN-USE 1974 MODEL YEAR VEHICLE FLEET	      5-1
      5. 1    Correlation Analysis Results	      5-2
      5. 2    Contingency Table  Analysis Results	      5-9
             5. 2. 1    Maximum Correlation Method	      5-9
             5. 2. 2    Bounded Errors of Commission Method .  . .      5-9
      5. 3    Relative Impact on Air Quality	      5-53
             5. 3. 1    By Individual Pollutant	      5-53
             5. 3. 2    Multiple Constituent Tests	      5-57
6.    DEFECT DATA FROM CATALYST-EQUIPPED
      EXPERIMENTAL VEHICLE FLEET	      6-1
      6. 1    Statistical Analysis of Defect Tests	      6-1
                                    vi u

-------
                          CONTENTS (Continued)
             6. 1. 1   Data Selection Procedures	     6-5
      6. 2    Contingency Table Analysis of Defect Data	     6-6
      6.3    Conclusions	     6-12
APPENDIX          DEFECT TEST DESCRIPTIONS	     A-l
                                     IX

-------
                                TABLES



1-1.    ST/FTP Correlation Summary (CEV Fleet)	     1-3

1-2.    Correlations for Weighted Mode Tests (CEV Fleet)  ....     1-6

1-3.    Correlations for Selected Car Deletions:  Federal
        Short Cycle vs FTP (CEV  Fleet)	     1-7

1-4.    ST Correlation Ratings (CEV Fleet)	     1-8

1-5.    Contingency Table	     1-9

1-6.    Assumed FTP Levels for CEV Fleet	     1-11

1-7.    Comparison of Selected ST Hydrocarbon Results:
        CEV Fleet, Bounded Errors of Commission Analysis,
        HC FTP Level = 0.41 gm/mile (E  = constant = 5%). ...     1-15

1-8.    Comparison of Selected ST Hydrocarbon Results: CEV
        Fleet, Bounded Errors of  Commission Analysis, HC
        FTP Level = 0.90 gm/mile (E   = constant = 5%)	     1-16

1-9.    Comparison of Selected ST Carbon Monoxide Results:
        CEV Fleet, Bounded Errors of Commission Analysis,
        CO FTP Level = 9. 0 gm/mi  (E = constant =5%)  	     1-18

1-10.   Comparison of Selected ST Carbon Monoxide Results:
        CEV Fleet, Bounded Errors of Commission Analysis,
        CO FTP Level = 3. 4 gm/mi  (E = constant =5%)	     1-19

1-11.   Comparison of Selected ST NOX Results:  CEV
        Fleet, Bounded Errors of  Commission Analysis,
        NO  FTP Level = 3. 1 gm/mi (E  = constant =5%)	     1-22
           jf.                           C
1-12.   Comparison of Three-Constituent Test Results:  CEV
        Fleet, Bounded Errors of  Commission Analysis,
        (Predicted EC = constant =5%)	     1-24

1-13.   Correlation Coefficient Summary:  1974 Model Year
        Fleet	     1-27

1-14.   ST Ratings:  1974 Model Year Fleet	     1-33
                                    XI

-------
                           TABLES (Continued)
 1-15.    Comparison of ST Hydrocarbon Results:  1974 Model
         Year Fleet,  Bounded Errors of Commission Analysis
         (E  = constant =  5%)	    I-37
           c
 1-16.    Comparison of ST Carbon Monoxide Results:  1974
         Model Year Fleet,  Bounded Errors of Commission
         Analysis (E  = constant =5%)	    1-39

 1-17.    Comparison of ST NOX Results:  1974 Model Year
         Fleet, Bounded Errors of Commission Analysis
         (E  = constant =5%)	    1-40

 1-18.    ST Comparison:  1974 Model Year Fleet, Multiple
         Constituent Tests (Actual E  < 2%)	    1-41
                                  C
 1-19.    Defect Analysis Comparison Summary:  Predicted
         Population [% EC = 5,  FTP Level I]  	    1-44

 1-20.    Key Mode Composite Test (Laboratory Data)	    1-45

 1-21.    Short Test Effectiveness; EC = 5%; 1974 Model Year
         Fleet	    1-50

 1-22.    Short Test Effectiveness Values for Multiple Con-
         stituent Tests; 1974 Model Year Fleet	    1-54

4-1.     Number of Cases Available for  Statistical Analysis
         (CEV Fleet)	    4-2

4-2.     Summary of Variance  Components (CEV Fleet)	    4-4

4-3.     Canonical Correlation Coefficients Between the FTP and
         ST for the CEV Fleet (first good data set)	    4-6

4-4.     FTP Composite vs Bag Correlation Summary
         (CEV Fleet)	    4-12

4-5.     ST/FTP Correlation Summary (CEV Fleet)	    4-13

4-6.     ST/FTP Correlations  for Weighted Mode Tests
         (CEV Fleet) (first good data only)	    4-14
                                   xn

-------
                           TABLES (Continued)
4-7.     Correlation Coefficients for Selected Car Deletions;
         Federal Short Cycle vs FTP (CEV Fleet)	    4-15

4-8.     ST Correlation Ratings	    4-18

4-9.     Contingency Table	    4-19

4-10.    Assumed FTP Levels (CEV Fleet)	    4-27

4-11.    Maximum Correlation Summary,  FTP Level I
         (CEV Fleet)	    4-32

4-12.    Maximum Correlation Summary,  FTP Level II
         (CEV Fleet)	    4-33

4-13.    Maximum Correlation Summary,  FTP Level III
         (CEV Fleet)	    4-34

4-14.    Maximum Correlation Summary,  FTP Level IV
         (CEV Fleet)	    4-35

4-15.    Comparison of Selected ST Hydrocarbon Results:
         CEV Fleet, Bounded Errors of Commission Analysis,
         HC FTP Level = 0.90 gm/mile (E  = constant = 5%)	    4-59

4-16.    Comparison of Selected ST Hydrocarbon Results:
         CEV Fleet, Bounded Errors of Commission Analysis,
         HC FTP Level = 0.41 gm/mile (E  = constant =5%)	    4-60

4-17.    Comparison of Selected ST Carbon Monoxide Results:
         CEV Fleet, Bounded Errors of Commission Analysis,
         CO FTP Level = 9. 0 gm/mi (E  = constant = 5%)	    4-71

4-18.    Comparison of Selected ST Carbon Monoxide Results:
         CEV Fleet, Bounded Errors of Commission Analysis,
         CO FTP Level = 3.4 gm/mi (E  = constant = 5%)	    4-72

4-19.    Comparison of Selected ST NOX Results: CEV Fleet,
         Bounded Errors  of Commission Analysis, NOX FTP
         Level = 3. 1 gm/mi (E  = constant = 5%)	    4-78
                             c

4-20.    Key Mode Weighting Factors	    4-79
                                   Xlll

-------
                           TABLES (Continued)
 4-21.    Approximate Standard Deviation for Three-Constituent
         Tests - CEV Fleet, N = 40	   4-95

 5-1.     Correlation Coefficient Summary:  1974 Model
         Year Fleet	   5~3

 5-2.     FTP Composite Versus Bag 2 + 3 Correlation
         Coefficients:  1974 Model Year Fleet	   5-6

 5-3.     ST Ratings:  1974 Model Year Fleet	   5-7

 5-4.     Maximum Correlation Summary; 1974 Model Year Fleet,
         Predicted Population	   5-10

 5-5.     Comparison of ST Hydrocarbon Results:  1974 Model
         Year Fleet, Bounded Errors of Commission Analysis
         (E  = constant = 5%)	   5-18
           c
 5-6.     Comparison of ST Carbon Monoxide Results:  1974
         Model Year Fleet,  Bounded Errors of Commission
         Analysis (E  = constant = 5%)	   5-28

 5-7.     Comparison of ST NOX Results:  1974 Model Year
         Fleet, Bounded Errors of Commission Analysis
         (E  = constant = 5%)	   5-31

 5-8.     Standard Deviation for Three-Constituent Tests:
         1974 Model Year Fleet,  N = 147	   5-31

 5-9.     ST Comparison:  1974 Model Year Fleet;  Multiple
         Constituent Tests  (E   < 2%)	     5-52
                            c

 5-10.    Short Test Effectiveness; E  = 5%; 1974 Model
         Year Fleet	c.	   5-54

 5-11.    Short Test Effectiveness Values for Multiple'
         Constituent Tests;  1974 Model Year Fleet	   5-58

6-1.     ST/FTP Correlation  Coefficient Comparison:  Defect
         Test Vehicles vs Original CEV Fleet (laboratory
         instruments)	   6-2

6-2.     Elementary FTP Statistics: Defect Test Vehicles vs
         Original CEV  Fleet (gm/mi)	   6-3
                                   xiv

-------
TABLES (Continued)
6-3.
6-4.
6-5.
Groups Distinguishable from Baseline
Defect Test Fleet 	
Defect Analysis Comparison Summary
Population [% E = 5, FTP Level I] .
Kev Mode Composite Test (laboratory
Operation:
: Predicted
data) 	
6-4
6-7
6-12
          XV

-------
                                FIGURES



1-1.     Contingency Table Representation	    1-10

1-2.     Bounded Errors of Commission Method	    1-12

1-3.     Variation of EC, Eo, and FF with HC Cut-Point;
         CEV Fleet; Federal Short Cycle;  Bounded Errors
         of Commission Method	    1-13

1-4.     Variation of Ec, Eo, and FF with CO Cut-Point;
         CEV Fleet; Federal Short Cycle;  Bounded Errors of
         Commission Method	    1-17

1-5.     Variation of Ec, Eo, and FF with NOX Cut-Point; CEV
         Fleet; Federal Short Cycle; Bounded Errors of
         Commission Method	    1-20

1-6.     Variability of Predicted Population Results	    1-23

1-7.     Variation of Ec, Eo, and FF with HC and NOX Cut-
         Points; 1974 Model Year Fleet; Federal Short
         Cycle; Bounded Errors of Commission Method	    1-36

1-8.     Variation of Ec, Eo, and FF with CO Cut-Point;
         1974 Model Year Fleet; Federal Short Cycle; Bounded
         Errors of Commission Method	    1-38

1-9.     Impact of Percent Population Sampled on CO
         Removed (Illustrative Example Only) 	    1-52

3-1.     Federal Short  Cycle and Composite NY/NJ Short
         Cycle Test Driving Schedules  	    3-5

4-1.     Correlation Analysis Scattergram	    4-10

4-2.     Contingency Table Representation	    4-20

4-3.     Maximum. Correlation Method	    4-22

4-4.     Bounded Errors of Commission Method	    4-22

4-5.     Weighted Errors Method	    4-23

4-6.     Percent Rejection Method	    4-23
                                   XVII

-------
                           FIGURES (Continued)



4-7.     Parametric Model	   4~25

4-8.     Probability Equations	   4-25

4-9.     Expected Values	   4-26

4-10.    Equations for Parametric Techniques	   4-26

4-11.    Variation of Ec, EQ, and FF with HC FTP Level;
         Maximum Correlation Method; Data Analytic
         Technique; CEV Fleet	   4-29

4-12.    Variation of Ec, Eo, and FF with HC FTP Level;
         Maximum Correlation Method; Parametric Tech-
         nique; CEV Fleet  	   4-30

4-13.    Variation of EC, Eo, and FF with HC FTP Level;
         Maximum Correlation Method; Predicted Popula-
         tion Technique; CEV Fleet  	   4-31

4-14.    Variation of EC, Eo, and FF with HC FTP Level;
         Federal Three-Mode Test; Maximum Correlation
         Method; Predicted Population of CEV Fleet	   4-36

4- 1 5.    Variation of Ec, Eo, and FF with HC FTP Level;
         Key Mode Test; Maximum Correlation Method;
         Predicted Population of CEV Fleet	   4-37

4-16.    Variation of Ec, EQ, and FF with CO FTP Level;
         Maximum Correlation Method; Predicted Population
         of CEV Fleet	   4-38

4-17.    Variation of Ec, EQ, and FF with CO FTP Level;
         Federal Three-Mode Test; Maximum Correlation
         Method; Predicted Population of CEV Fleet	   4-39

4-18.    Variation of Ec, Eo, and FF with CO FTP Level;
         Key Mode Test; Maximum Correlation Method;
         Predicted Population of CEV Fleet	   4-40

4-19.    Variation of Ec, EQ, and FF with HC FTP Level;
         Unloaded 2500 rpm Test; Garage Instruments; Maxi-
         mum Correlation Method; Predicted Population of
         CEV Fleet	   4-41
                                   XVlll

-------
                           FIGURES (Continued)
4-20.    Variation of Ec, Eo, and FF with HC FTP Level;
         Federal Three-Mode Test; Garage Instruments;
         Maximum Correlation  Method; Predicted Population
         of CEV Fleet	   4-42

4-21.    Variation of Ec, Eo, and FF with HC FTP Level;
         Key Mode Test; Garage Instruments; Maximum
         Correlation Method; Predicted Population of CEV
         Fleet	   4-43

4-22.    Variation of Ec, Eo, and FF with CO FTP Level;
         Unloaded 2500 rpm Test; Garage Instruments;
         Maximum Correlation  Method; Predicted Population
         of CEV Fleet	   4-44

4-23.    Variation of Ec, Eo, and FF with CO FTP Level;
         Federal Three-Mode Test; Garage Instruments;
         Maximum Correlation  Method; Predicted Population
         of CEV Fleet	   4-44

4-24.    Variation of Ec, EQ, and FF with CO FTP Level; Key
         Mode Test; Garage Instruments; Maximum Correlation
         Method; Predicted Population of CEV Fleet	   4-45

4-25.    Variation of E0 and Ec at NOX Level of 3. 1 gm/mi;
         CEV Fleet; Maximum  Correlation Method; Pre-
         dicted Population of CEV Fleet	   4-46

4-26.    Variation of EC, EQ, and FF with HC Cut-Point;
         CEV Fleet; Federal Short Cycle; Bounded Errors
         of Commission Method	   4-49

4-27.    Variation of EC, Eo, and FF with HC Cut-Point;
         CEV Fleet; NY/NJ Composite  Test; Bounded
         Errors of Commission Method	   4-50

4-28.    Variation of Ec, Eo, and FF with HC Cut-Point;
         CEV Fleet,  Key Mode  Test; Bounded Errors of
         Commission Method	   4-51

4-29.    Variation of Ec, EQ, and FF with HC Cut-Point;
         CEV Fleet; Key Mode  Test; Garage Instruments;
         Bounded Errors of Commission Method	   4-52
                                    xix

-------
                           FIGURES (Continued)
4-30.    Variation of Ec,  Eo,  and FF with HC Cut-Point;
         CEV Fleet; 1975 FTP HC Standard = 0.41 gm/mi;
         Federal Three-Mode Test; Bounded Errors of
         Commission Method	    4-53

4-31.    Variation of EC,  Eo,  and FF with HC Cut-Point;
         CEV Fleet; 1975 FTP HC Standard = 0. 9 gm/mi;
         Federal Three-Mode Test; Bounded Errors of
         Commission Method	    4-54

4-32.    Variation of Ec,  EQ,  and FF with HC Cut-Point;
         CEV Fleet; 1975 FTP Standard = 0.41 gm/mi;
         Federal Three-Mode Test; Garage Instruments;
         Bounded Errors of Commission Method	    4-55

4-33.    Variation of Ec,  EQ,  and FF with HC Cut-Point;
         CEV Fleet; 1975 FTP Standard = 0. 9  gm/mi; Federal
         Three-Mode Test; Garage Instruments; Bounded
         Errors of Commission Method	    4-56

4-34.    Variation of Ec,  Eo,  and FF with HC Cut-Point;
         CEV Fleet; Unloaded 2500 rpm Test;  Bounded
         Errors of Commission Method	    4-57

4-35.    Variation of Ec,  Eo,  and FF with HC Cut-Point;
         CEV Fleet; Unloaded 2500 rpm Test;  Garage
         Instruments; Bounded Errors of Commission
         Method  	    4-58

4-36.    Variation of EC,  E0,  and FF with CO Cut-Point;
         CEV Fleet; Federal Short Cycle; Bounded Errors
         of Commission Method	    4-62

4-37.    Variation of Ec,  Eo,  and FF with CO Cut-Point;
         CEV Fleet; 1975 FTP CO Level =3.4 gm/mi; NY/NJ
         Composite Test; Bounded Errors of Commission
         Method	    4-63

4-38.    Variation of EC,  E0,  and FF with CO Cut-Point;
         CEV Fleet; Key Mode Test; Bounded Errors of
         Commission Method	    4-64
                                    xx

-------
                           FIGURES (Continued)
4-39.    Variation of EC, EQ, and FF with CO Cut-Point;
         CEV Fleet; Key Mode Test; Garage Instruments;
         Bounded Errors of Commission Method	    4-65

4-40.    Variation of Ec, E , and FF with CO Cut-Point;
         CEV Fleet; 1975 FTP CO Level =3.4 gm/mi;
         Federal Three-Mode Test; Bounded Errors of
         Commission Method	    4-66

4-41.    Variation of EC, EQ, and FF with CO Cut-Point;
         CEV Fleet; 1975 FTP Level =3.4 gm/mi; Federal
         Three-Mode Test; Garage Instruments; Bounded
         Errors of Commission Method	    4-67

4-42.    Variation of E  , EQ, and FF with CO Cut-Point;
         CEV Fleet; Unloaded 2500 rpm Test; Bounded
         Errors of Commission Method	    4-68

4-43.    Variation of E_, EQ, and FF with CO Cut-Point;
         CEV Fleet; Unloaded 2500 rpm Test; Garage
         Instruments; Bounded Errors of Commission
         Method	    4-69

4-44.    Variation of Ec, EQ, and FF with NOX Cut-Point;
         CEV Fleet; Federal Short Cycle Test; Bounded
         Errors of Commission Method	    4-73

4-45.    Variation of EC, EQ, and FF with NOX Cut-Point;
         CEV Fleet; NY/NJ Composite Test; Bounded Errors
         of Commission Method	    4-74

4-46.    Variation of EC, EQ, and FF with NOX Cut-Point;
         CEV Fleet; Key Mode Test; Bounded Errors of
         Commission Method	    4-75

4-47.    Variation of Ec, Eo, and FF with NOX Cut-Point;
         CEV Fleet; Federal Three-Mode Test; Bounded
         Errors of Commission Method	    4-76

4-48.    Variation of Ec, Eo, and FF with NOX Cut-Point;
         CEV Fleet; Unloaded 2500 rpm Test; Bounded
         Errors of Commission Method	    4-77
                                    xxi

-------
                           FIGURES (Continued)
4-49.    Variation of E  and EQ for Key Mode and Weighted
         Key Mode Tests; CEV Fleet; 1975 FTP HC Level =
         0.41 gm/mi; Bounded Errors of Commission Method ....    4-80

4-50.    Variation of Eo and EC for Key Mode and Weighted
         Key Mode Tests; CEV Fleet; 1975 FTP CO Level =
         3.4 gm/mi; Bounded  Errors of Commission Method	    4-81

4-51.    Variation of EQ and Ec for Key Mode and Weighted
         Key Mode Tests; CEV Fleet; 1975 FTP NOX Level =
         3. 1 gm/mi; Bounded  Errors of Commission Method  ....    4-82

4-52.    Variability of Predicted Population Results	    4-85

4-53.    Computation Flow Chart	    4-86

4-54.    Variation of Actual Ec, Eo, and FF with Predicted EC;
         Federal Short Cycle;  Three-Constituent Test; Bounded
         Errors of Commission Method; CEV Fleet; 1975
         FTP Level 1	    4-88

4-55.    Variation of Actual E^ Eo, and FF with Predicted EC;
         Federal Short Cycle;  Three-Constituent Test; Bounded
         Errors of Commission Method; CEV Fleet; 1975
         FTP Level 2	    4-88

4-56.    Variation of Actual E^ EQ, and FF with Predicted Ec;
         Federal Short Cycle;  Three-Constituent Test; Bounded
         Errors of Commission Method; CEV Fleet; 1975
         FTP Level 3	    4-89

4-57.    Variation of Actual E^ E0, and FF with Predicted Ec;
         Federal Short Cycle;  Three-Constituent Test; Bounded
         Errors of Commission Method; CEV Fleet; 1975
         FTP Level 4	    4-89

4-58.    Variation of Actual Ec, Eo, and FF with Predicted Ec;
         Federal Three-Mode; Laboratory Instruments; Three-
         Constituent  Test; Bounded Errors of Commission
         Method; CEV Fleet; 1975 FTP Level 1	    4-90

4-59.    Variation of Actual Ec, EQ, and FF with Predicted Ec;
         Federal Three-Mode; Laboratory Instruments; Three-
         Constituent  Test; Bounded Errors of Commission
         Method; CEV Fleet; 1975 FTP Level 2	    4-90
                                   xxn

-------
                           FIGURES (Continued)
4-60.   Variation of Actual Ec, EQ,  and FF with Predicted Ec;
        Federal Three-Mode; Laboratory Instruments; Three-
        Constituent Test; Bounded Errors of Commission
        Method; CEV  Fleet; 1975 FTP Level 3	    4-91

4-61.   Variation of Actual EC, EQ,  and FF with Predicted EC;
        Federal Three-Mode; Laboratory Instruments; Three-
        Constituent Test; Bounded Errors of Commission
        Method; CEV  Fleet; 1975 FTP Level 4	    4-91

4-62.   Variation of Actual EC, EQ,  and FF with Predicted EC;
        Federal Three-Mode; Garage Instruments; Three-
        Constituent Test; Bounded Errors of Commission
        Method; CEV  Fleet; 1975 FTP Level 1	    4-92

4-63.   Variation of Actual Ec, Eo,  and FF with Predicted EC;
        Federal Three-Mode; Garage Instruments; Three-
        Constituent Test; Bounded Errors of Commission
        Method; CEV  Fleet; 1975 FTP Level 2	    4-92

4-64.   Variation of Actual EC, EQ,  and FF with Predicted EC;
        Federal Three-Mode; Garage Instruments; Three-
        Constituent Test; Bounded Errors of Commission
        Method; CEV  Fleet; 1975 FTP Level 3	    4-93

4-65.   Variation of Actual EC, EQ,  and FF with Predicted Ec;
        Federal Three-Mode; Garage Instruments; Three-
        Constituent Test; Bounded Errors of Commission
        Method; CEV  Fleet; 1975 FTP Level 4	    4-93

5-1.    Variation of Ec,  Eo, and FF with HC and NOX Cut-
        Point; 1974 Model Year Fleet; Federal Short Cycle
        Test; Bounded Errors of Commission Method	    5-11

5-2.    Variation of EC,  EQ, and FF with HC and NOX Cut-
        Point; 1974 Model Year Fleet; NY/NJ Composite
        Test; Bounded Errors of Commission Method	    5-12

5-3.    Variation of Ec,  Eo, and FF with HC Cut-Point;
        1974 Model Year Fleet; Key Mode Test; Bounded
        Errors of Commission Method	    5-13
                                   XXlll

-------
                          FIGURES (Continued)
5-4.     Variation of EC, E , and FF with HC Cut-Point;
         1974 Model Year Fleet; Key Mode  Test; Garage
         Instruments; Bounded Errors of Commission
         Method  	    5-14

5-5.     Variation of Ec, Eo, and FF with HC Cut-Point; 1974
         Model Year Fleet; Federal Three-Mode Test; Bounded
         Errors of Commission Method	    5-15

5-6.     Variation of Ec, Eo, and FF with HC Cut-Point; 1974
         Model Year Fleet; Federal Three-Mode Test; Garage
         Instruments; Bounded Errors of Commission Method ....    5-16

5-7.     Variation of Ec, Eo, and FF with HC and NOX Cut-
         Point; 1974 Model Year Fleet; Unloaded 2500 rpm
         Test; Bounded Errors of Commission Method	    5-17

5-8.     Variation of EC, E , and FF with HC Cut-Point;
         1974 Model Year Fleet; Unloaded 2500 rpm Test;
         Garage Instruments; Bounded Errors of Commission
         Method  	    5-17

5-9.     Variation of EC, Eo, and FF with CO Cut-Point;
         1974 Model Year Fleet; Federal Short Cycle Test;
         Bounded Errors of Commission Method	    5-20

5-10.    Variation of Ec, EQ, and FF with CO Cut-Point;
         1974 Model Year Fleet; NY/NJ Composite Test;
         Bounded Errors of Commission Method	    5-21

5-11.    Variation of Ec, EQ, and FF with CO Cut-Point;
         1974 Model Year Fleet; Key Mode  Test; Bounded
         Errors of Commission Method	    5-22

5-12.    Variation of Ec, Eo, and FF with CO Cut-Point;
         1974 Model Year Fleet; Key Mode  Test; Garage
        Instruments; Bounded Errors of Commission Method. . . .    5-23

5-13.    Variation of Ec, Eo, and FF with CO Cut-Point;
         1974 Model Year Fleet; Federal Three-Mode Test;
         Bounded Errors of Commission Method	    5-24
                                   XXIV

-------
                           FIGURES (Continued)
5-14.   Variation of Ec,  E0,  and FF with CO Cut-Point;
        1974 Model Year Fleet; Federal Three-Mode Test;
        Garage Instruments;  Bounded Errors of Commission
        Method	    5-25

5-15.   Variation of EC,  E ,  and FF with CO Cut-Point;
        1974 Model Year Fleet; Unloaded  2500 rpm Test;
        Bounded Errors of Commission Method	    5-26

5-16.   Variation of EC,  E   and FF with CO Cut-Point;
        1974 Model Year Fleet; Unloaded  2500 rpm Test;
        Garage Instruments;  Bounded Errors of Commission
        Method	    5-27

5-17.   Variation of EC,  EQ,  and FF with NOX Cut-Point;
        1974 Model Year Fleet; Key Mode Test; Bounded
        Errors of Commission Method	    5-29

5-18.   Variation of Ec,  Eo,  and FF with NOX Cut-Point;
        1974 Model Year Fleet; Federal Three-Mode Test;
        Bounded Errors of Commission Method	    5-30

5-19.   Variation of EC,  Eo,  and FF with Instrument Type;
        HC; 1974 Model Year Fleet; Key Mode Test; High
        Speed Mode;  Bounded Errors  of Commission Method ....    5-33

5-20.   Variation of Ec,  Eo,  and FF with Instrument Type;
        CO; 1974 Model Year Fleet; Key Mode Test; High
        Speed Mode;  Bounded Errors  of Commission Method ....    5-34

5-21.   Variation of Ec,  Eo,  and FF with Instrument Type;
        HC; 1974 Model Year Fleet; Key Mode Test; Low
        Speed Mode;  Bounded Errors  of Commission Method ....    5-35

5-22.   Variation of Ec,  Eo,  and FF with Instrument Type;
        CO; 1974 Model Year Fleet; Key Mode Test; Low
        Speed Mode;  Bounded Errors  of Commission Method ....    5-36

5-23.   Variation of Ec,  Eo,  and FF with Instrument Type;
        HC; 1974 Model Year Fleet; Key Mode Test; Idle
        Mode;  Bounded Errors of Commission Method	    5-37
                                   xxv

-------
                          FIGURES (Continued)
5-Z4.   Variation of Ec, Eo,  and FF with Instrument Type;
        CO; 1974 Model Year Fleet; Key Mode Test; Idle
        Mode; Bounded Errors of Commission Method	    5-38

5-25.   Variation of Ec, Eo,  and FF with Instrument Type;
        HC; 1974 Model Year Fleet; Federal Three-Mode Test;
        High Speed Mode;  Bounded Errors of Commission
        Method	    5-39

5-26.   Variation of Ec, E0,  and FF with Instrument Type;
        CO; 1974 Model Year Fleet; Federal Three-Mode Test;
        High Speed Mode;  Bounded Errors of Commission
        Method	    5-40

5-27.   Variation of Ec, Eo,  and FF with Instrument Type;
        HC; 1974 Model Year Fleet; Federal Three-Mode Test;
        Low Speed Mode; Bounded Errors of Commission
        Method	    5-41

5-28.   Variation of Ec, Eo,  and FF with Instrument Type;
        CO; 1974 Model Year Fleet; Federal Three-Mode Test;
        Low Speed Mode; Bounded Errors of Commission
        Method	    5-42

5-29.   Variation of Ec, Eo,  and FF with Instrument Type;
        HC; 1974 Model Year Fleet; Federal Three-Mode Test;
        Idle Mode; Bounded Errors of Commission Method	    5-43

5-30.   Variation of Ec, Eo,  and FF with Instrument Type;
        CO; 1974 Model Year Fleet; Federal Three-Mode Test;
        Idle Mode; Bounded Errors of Commission Method	    5-44

5-31.   Variation of Ec, Eo,  and FF with Instrument Type;
        HC; 1974 Model Year Fleet; Unloaded 2500 rpm Test;
        Bounded Errors of Commission Method	    5-45

5-32.   Variation of Ec, EQ,  and FF with Instrument Type;
        CO; 1974 Model Year Fleet; Unloaded 2500 rpm Test;
        Bounded Errors of Commission Method	    5-46

5-33.   Variation of Actual EC,  Eo, and FF with Predicted
        Ec; Federal Short Cycle; Three-Constituent Test;
        Bounded Errors of Commission Method; 1974 Model
        Year Fleet	    5-48
                                  XXVI

-------
                           FIGURES (Continued)
5-34.   Variation of Actual Ec,  E0,  and FF with Predicted
        Ec; Federal Three-Mode; Three-Constituent Test;
        Laboratory Instruments; Bounded Errors of Com-
        mission Method; 1974 Model Year Fleet  	    5-49

5-35.   Variation of Actual Ec,  Eo,  and FF with Predicted
        Ec; Federal Three-Mode; Three-Constituent Test;
        Garage Instruments; Bounded Errors of Commission
        Method; 1974 Model Year Fleet	    5-50

5-36.   Impact of Percent Population Sampled on CO Removed
        (Illustrative Example Only)	    5-56

6-1.    Variation of Ec, Eo, and FF with HC Cut-point;
        Original CEV Fleet;  Key Mode Test; 1975 FTP
        Level = 0.41 gm/mi; Bounded Errors of Commission
        Method	    6-8

6-2.    Variation of Ec, Eo, and FF with HC Cut-point;
        Defect Tests Only; Key Mode Test;  1975 FTP
        Level = 0. 41 gm/mi; Bounded Errors of Commission
        Method	    6-8

6-3.    Variation of Ec, Eo, and FF with CO Cut-point;
        Original CEV Fleet;  Key Mode Test; 1975 FTP
        Level = 3.4 gm/mi; Bounded Errors of  Commission
        Method	    6-9

6-4.    Variation of Ec, Eo, and FF with CO Cut-point;
        Defect Tests Only; Key Mode Test;  1975 FTP
        Level =3.4 gm/mi; Bounded Errors of  Commission
        Method	    6-9

6-5.    Variation of Ec, Eo, and FF with NOX Cut-point;
        Original CEV Fleet;  Key Mode Test; 1975 FTP
        Level = 3.1 gm/mi; Bounded Errors of  Commission
        Method	    6-10

6-6.    Variation of Ec, E0, and FF with NOX Cut-point;
        Defect Tests Only; Key Mode Test;  1975 FTP
        Level = 3.1 gm/mi; Bounded Errors of  Commission
        Method	    6-11
                                   XXVll

-------
1.   SUMMARY

-------
                               1.  SUMMARY


               A series of statistical analyses was performed to determine
the degree of "correlation" that exists between five specific short tests (STs)
and the federal emission certification test procedure (FTP) for new vehicles.
This work was performed to determine if "reasonable correlation with certi-
fication test procedures" exists; this is a condition precedent to the promulga-
tion of  regulations that impose  the in-use warranty provisions  of Sec. 207 (b)
of the Clean Air  Act of 1970 upon the motor vehicle manufacturers.
               The basis for the analyses was ST and FTP test data from
three vehicle fleets:
         •     A catalyst-equipped experimental vehicle fleet (40 vehicles)
         •     An in-use 1974 model year vehicle fleet (147 vehicles)
         •     A catalyst-equipped defect test fleet (5 vehicles)
Each of the vehicles in these fleets was tested by the FTP and the following STs:
         •     Federal Short Cycle
         •     New York/New Jersey (NY/NJ) Composite
         •     Clayton Key Mode
         •     Federal Three-Mode
         •     Unloaded 2500 rpm
The first two of these STs are CVS (constant volume  sampling) or bag-type
tests wherein a test technician  drives the car on the dynamometer  in accord-
ance with a prescribed driving  pattern.  The vehicle  exhaust is diluted by
the CVS procedure, and a single sample  bag of diluted exhaust is collected
for the  ST.   The  latter three STs are categorized as  modal or volumetric.
In these tests, the test technician operates the vehicle on a dynamometer at
a fixed  vehicle speed and dynamometer load, or unloaded at a fixed engine
rpm, or at idle.  The vehicle tailpipe exhaust is sampled directly, and the
concentration of  each pollutant  is measured and recorded.  The Clayton Key
Mode and the Federal Three-Mode STs each have high-speed,  low-speed,
                                     1-1

-------
 and idle modes.  The Unloaded 2500 rpm ST is a high-speed test with the
 transmission in neutral at 2500 engine rpm.
               Hydrocarbon (HC) and carbon monoxide (CO) measurements
 were recorded with both laboratory analyzers and garage-type instruments
 for most of the volumetric tests (Key Mode, Federal Three-Mode, Unloaded
 2500  rpm). All oxides of nitrogen (NO ) measurements were  made with
 laboratory analyzers.
               Two different statistical analysis methods were used to assess
 "correlation" --a conventional correlation analysis, and a contingency table
 analysis.
               The principal results of the study are summarized in the
 following sections.  Because of the many variables involved (three test fleets,
 five STs, three emission constituents, two types of measurement instruments,
 etc.), the  results are  presented first as a function of fleet type; then overview
 statements or findings  are presented which provide more general conclusions,
 where appropriate.
 1. 1            CATALYST-EQUIPPED EXPERIMENTAL
               VEHICLE FLEET
 1.1.1         Direct Relatability Results
               A  conventional correlation analysis was performed for the
catalyst-equipped vehicle (CEV) fleet for each of the five  short tests; a sum-
mary of the ST/FTP correlation coefficients  obtained is given in Table 1-1.
The correlation coefficient (r) is the quantitative measure of relatability
between the results of the  short test and the FTP-  The closer r is to 1, the
better the relation.  No relationship is indicated by r = 0.  Negative r indi-
cates an inverse  relation between the observed test results. For a test
sample size (N) of 40 or 39, a computed correlation of less than 0.35 indicates
that the ST and the FTP pollutants are uncorrelated with 95 percent confidence.
For N = 25 or 26, this  threshold is approximately 0.4.
                                    1-2

-------
           Table  1-1.  ST/FTP Correlation Summary (CEV Fleet)
Short Test
Federal Short Cycle
NY/NJ Composite
Key Mode
(Laboratory)

Key Mode
(Garage)

Federal Three -Mode
(Laboratory)

Federal Three -Mode
(Garage)

2500 rpm
(Laboratory)
2500 rpm
(Garage)
Good
Data
Set(a)
First
Second
Average
First
Second
Average
First
Second
First
Second
First
Second
First
Second
First
Second
First
Second
Test
Mode


High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle


N(b)
39
25
39
39
25
40
40
26
40
26
31
26
40
26
40
26
40
26
"r" -ST/FTP Correlation'0*
Coefficient
HC
0.87
0.91
0.93
0.92
0.92
0.95
0.61
0.53
0.92
0.57
0.53
0.97
0.73
0.73
0.88
0.51
0.39*
0.32*
0.87
0.79
0.80
0.68
0.52
0.94
0.76
0.73
0.78
0.69
0.42
0.62
0.47
0.37*
0.50
0.36*
CO
0.81
0.42
0.83
0.77
0.71
0.68
0.26*
0.39
0.54
0.30*
0.31*
0.40
0.37
0.21*
0.52
0.08*
0.09*
-0.03*
0.08*
0.22*
0.48
0.20*
0.27*
0.34*
0.24*
0.21*
0.52
0. 12*
0.03*
0.39*
0.30*
0.25*
0. 14*
0. 25*
N0x
0.62
0.47
0.53
0.61
0.51
0.61
0.79
0.20*
0.27*
0.86
0 . 04*
0 . 04*


0.89
0.03*
0. 13*
0.92
-0.28*
0.08*


0.23*
0.23*

(a)  First Good Data:  This data set contains the observations of the first FTP and ST,
    both of which are valid.
    Second Good Data: This data set contains the second pair of FTP and ST obser-
    vations, both of which are valid.
    Average Data:  This data set contains the average of the FTP  and ST observations
    on each car (for the Federal Short Cycle and NY/NJ Composite only).
(b)  Number of cars in data set
(c)  The  correlation is statistically significant at the 95% confidence level except when
    indicated by an asterisk.
                                           1-3

-------
 1.1.1.1      Hydrocarbon Emission
              The bag-type STs (Federal Short Cycle and NY/NJ Composite)
 and the idle mode of the volumetric or modal tests (Key Mode, Federal Three-
 Mode) in general show superior HC tracking characteristics.  However,  on
 the Federal Three-Mode, the high-speed mode has a slightly higher correla-
 tion in some instances than the idle mode.  The Unloaded 2500 rpzn ST has
 much poorer HC  correlation.
 1.1.1.2      Carbon Monoxide Emission
              The bag-type STs exhibit the superior CO tracking character-
 istics, but of a lower correlation level than that achieved for HC.  The idle
 mode of the volumetric tests has higher correlation than the high and low
 speed modes, but with a rather  poor correlation coefficient level.   The
 Unloaded 2500 rpzn ST is essentially uncorrelated for CO.
 1.1.1.3      Oxides of Nitrogen Emission
              The high-speed modes of the volumetric tests display the best
 ability to track NO .  The bag-type tests correlate with NO  , but at a much
                  3fc                                     Ji
 lower coefficient level.  The idle and low-speed volumetric modes and the
 Unloaded 2500 rpm ST are uncorrelated with NO .
                                              Ji
 1.1.1.4      Modal vs Bag  Tests
              On the basis of HC and CO correlation, as noted above,  the
bag tests (Federal Short Cycle and NY/NJ Composite) are preferable to the
modal- or volumetric-type ST.  The volumetric STs show acute deficiencies
in tracking CO.   However, the high-speed modes of the  volumetric ST  have
superior NO correlation.
            JL
              An analysis of variance indicated that the percent error due to
testing was higher for bag tests  than many of the modal  tests (using the same
laboratory instruments).  This higher testing error may be due to variations
of vehicle operation while trying to follow the driving profile of the short
driving test procedure, rather than due to the bag collection method, per se.
The lower testing error of the volumetric tests,  on the other hand, may  be
                                   1-4

-------
due to the simplicity of the test procedure itself, in that the measurements
are taken at stabilized engine operating conditions.
1.1.1.5       Weighted Modal Tests
              A multiple regression analysis was performed for the three-
mode volumetric tests on the first good data  set.  The purpose of this analysis
was to empirically determine the linear combinations of like constituents of
the three-mode  readings that have maximum correlation with the FTP.  The
results are shown in Table 1-2,  along with the maximum correlation using
only a single reading on each constituent.  As can be seen from the table,
the weighted combination correlations are not significantly greater than the
correlation of the best  single reading.
1. 1. 1. 6      Laboratory Analyzers vs Garage Instruments
              The largest differences between the correlation results of the
two measurement techniques occur on the second good data sets.  However,
the sample size of the second good data set,  26 cars, is risky for inference
purposes.  In general,  there is a greater variation in the correlation esti-
mates of first good data and second good data for the garage  analyzer than
for the laboratory analyzer, as shown in Table 1-1.
              The most striking difference between laboratory and garage
data is for HC on the Federal Three-Mode.  The laboratory measurements for
first good data indicate the best mode to be high speed, while the correspond-
ing garage instrument readings indicate the idle mode as superior.  This is
inconsistent with the results for HC on the  Clayton Key Mode where both
instrument types indicated the idle mode as superior.   Firm inferences are
tenuous due to differences in sample size.
              CO correlation deficiency is common to both measurement
techniques.  Due to the low concentration of CO being emitted in the CEV
fleet, this may be a measurements problem,  in general,  rather than a defi-
ciency in ST structure.
                                    1-5

-------
               Table 1-2.  Correlations for Weighted Mode Tests (CEV Fleet)
Short Test
Key Mode
Laboratory
Garage
Federal Three -Mode
Laboratory
Garage
N(a)

40
40

31
40
Weighted Correlation(b)
Coefficient
HC

0.93
0.91

0.91
0.81
CO

0.55
0.58

0.48
0.53
N0x

0.83


0.90

Best Single -Mode^c'.,.
Correlation Coefficient* '
HC

0.92 (I)
0.88 (I)

0.87 (H)
0.78 (I)
CO

0.54 (I)
0.52 (I)

0.48 (I)
0.52 (I)
N0x

0.79 (H)


0.89 (H)

(a)  Number of cars in first good data set; the first pair of FTP and ST observations, both
    of which are valid
(b)  Correlations are  statistically significant at the 95% confidence level

(c)  H = high speed mode
     I = idle mode

-------
1.1.1.7
Sensitivity of Correlation Re suit s
              Selected extreme data points were deleted and the correlation
coefficient recalculated for the Federal Short Cycle ST to illustrate variability
due to the data sample.  As  shown in Table 1-3, the correlation coefficient
is extremely sensitive to a small percentage of the data points.
            Table 1-3.  Correlations for Selected Car Deletions:
                        Federal Short Cycle vs FTP (CEV Fleet)

Number of Cars Deleted


0
1
2
3
4
(a)
Correlation Coefficient*


HC

0.872
0.657
0.656
--
--

CO

0.810
0.673
0.639
--
--

NO
X
0.621
0.690
0.633
0.823
0.755
         (a)  Significant at the 95% confidence level
1. 1. 1.8
ST Correlation Ratings
The following qualitative rating scale was used to rate the ST:
              Rating
         (U)   Unacceptable

         (P)   Poor

         (F)   Fair
         (G)   Good
         (E)   Excellent
                                    Description
                   Constituent is uncorrelated at the 95 percent
                   confidence level
                   Constituent is correlated at the 95 percent
                   confidence level, but with correlation less
                   than 0.6
                   Correlation between 0.6 and 0.7
                   Correlation between 0.7 and 0.9
                   Correlation between 0.9 and 1.0
                                   1-7

-------
              For rating the three-mode volumetric ST, the mode with the
 highest rating was used.  Table 1-4 shows the ratings of the ST on each
 pollutant on this basis.

               Table 1-4.  ST Correlation Ratings (CEV Fleet)
Short Test
Federal Short Cycle
NY/NJ Composite
Key Mode
Laboratory
Garage
Federal Three -Mode
Laboratory
Garage
2500 rpm Unloaded
Laboratory
Garage
Rating
HC
G
E

E (I)(a)
G (I)

G(H)
G (I)

P
P
CO
G
G

P (I)
P (I)

P(I)
P (I)

U
U
NOX
F
P

G(H)


G(H)


U

         (a) I = idle mode, H = high speed mode

              In general, the STs have less difficulty tracking HC than CO
and NO .  Excluding the Unloaded 2500 rpm ST  (which has either "P" or "U"
       j£
ratings for all three pollutants), the bag-type  and modal STs all have "G" to
"E" ratings for HC.  In the case of CO, the bag-type STs have "G" ratings,
whereas the modal STs are rated in the "P" category.  This situation is re-
versed in the case of NO , where the modal STs have "G" ratings and the
bag-type STs are rated "F" to "P".  Hence, the choices among the STs for
CO and NO  implementation may be more limited than for HC.
                                   1-8

-------
1. 1.2
Contingency Table Analysis Results
               The contingency table analysis technique was used to establish
the ST pass-fail levels for each pollutant.  The contingency table is defined in
Table  1-5, along with its associated parameters.   A pictorial demonstration
of its application to a given data set is shown in Figure 1-1.  It can be seen
that, for a given data set, part of the analysis is concerned with the criteria
used to select the ST cut-points. In this regard, the bounded errors of com-
mission method was used extensively to establish trends for the variations in
E  , E , FF, and PP.  In this method,  the ST cut-points are selected to mini-
  c   o                                           c
mize E  while  holding the E  below a specified level.  It thus  permits a direct
                       Table 1-5.  Contingency Table

II .u
T3 M
9J O
OH w
Pass
Fail
Total
True = FTP
Pass
a
c
a + c
Fail
b
d
b + d
Total
a + b
c + d
n = a + b
+ c + d
                a = number of correctly passed vehicles (PP)
                b = number of errors of omission (E  )
                c = number of errors of commission (E  )
                d = number of correctly failed vehicles (FF)
          Sensitivity = a/(a + c)
          Specificity = b/(b + d)
          False positive error = b/(a + b)
          False negative error = c/(c + d)
          Correlation index =
                                         ad - be
                              [(a + b)(a + c)(b + d)(c + d)]
                                                        1/2
                                    1-9

-------
                                           FF, CORRECTLY
                                           FAILED VEHICLES
                    Er, ERROR OF
                     c
                    COMMISSION
    ST CUT-POINT
t
oc
to
a
  	Ji uui  rv/im	1	
         PP, CORRECTLY
         PASSED VEHICLES
                                     O_
                                     t
                                             EQ| ERROR OF

                                             OMISSION
                   FTP MEASUREMENT
        Figure 1-1.  Contingency Table Representation
                           1-10

-------
answer to the question,  "For a given permissible level of E ,  what level of
                                                          c
EQ is associated with the ST, and with what impact on air quality (inferred
from number of FF and EQ vehicles)?" This method is illustrated in Fig-
ure 1-2.  The policy  decision is the maximum allowable E  .
                                                        c
              With regard to procedural technique, a bivariate normal dis-
tribution model was fitted to a particular data set by incorporating the  corre -
lation coefficient, mean values, and standard deviations of  the data set.  The
ST cut-points were then determined by using the model for  the predicted
population of the CEV fleet.
              As the appropriate FTP standards to which the CEV fleet was
designed were uncertain, four sets of FTP cut-points were  used in the analysis,
as specified in Table 1-6.  The bound of the errors of commission was varied
from 5 percent to 1 percent in 1-percent increments, with the  values 0.5
percent and 0. 1 percent also included.
1. 1.2.1
Hydrocarbon Emission

The variation of E „  E ,  and FF as a function of HC cut-point
                  C   O
was graphically determined for each ST examined.  The results for the Federal
Short Cycle are shown in Figure 1-3 to indicate the general nature of the
              Table 1-6.  Assumed FTP Levels for CEV Fleet

Level

I
II
in
IV
Emission Levels, gm/mi
HC

0.41
0.60
0.75
0.90
CO

3.4
5.0
7.0
9-0
NO
X
3. 1
3. 1
3. 1
3.1
                                    1-11

-------
       MINIMIZE EA SUBJECT TO Er < /%
                 o             \f
     I
     CO
            NOT TO

          EXCEED Y%
           ST

        CUT-POINT , '

                             MINIMIZE
                    Q.


                    t
           FTP MEASUREMENT
Figure 1-2.  Bounded Errors of Commission Method
                   1-12

-------
0.3      0.4
                0.5
0.6     0.7     0.8
 HC CUT-POINT, gm/mi
1.0     1.1
Figure 1-3.   Variation of Ec, Eo, and FF with HC Cut-Point;
  CEV Fleet; Federal Short Cycle; Bounded Errors of Com-
  mission Method
                             1-13

-------
tradeoffs available for policy formulation.  Reducing the EC increases EQ
and decreases FF.  All STs had similar trends.  To illustrate specific values
and trends among the STs, Tables  1-7 and 1-8 summarize data from the
graphical displays at HC FTP levels of 0.41 and 0.90.   On the average, at
both FTP levels, the bag tests (Federal Short Cycle and NY/NJ Composite)
have lower E and higher FF at the fixed E  =  5 percent than do the volumetric
tests.  However, the idle mode of the Clayton Key Mode (with either laboratory
or garage instruments) test produces similar results.  The Unloaded 2500 rpm
test is very poor on a comparative basis.
1.1.2.2       Carbon Monoxide Emission
              The variation of E  , E  ,  and FF as  a function of CO cut-point
                               CO
was also graphically determined for each ST examined,  and for the range of
CO FTP values  selected in Table 1-6 (CO = 3.4 to 9). Figure 1-4 indicates
results for the Federal Short Cycle.  As in the preceding case of hydrocarbon
emissions, these displays  indicated the tradeoffs possible between E  , E  ,
and FF.   However, for CO FTP levels above 3.4, the general or average CO
levels  of the  CEV fleet were sufficiently low; i.e.,  a very high percentage of
the vehicles exceeded the 5-,  7-. and 9-gm/mi requirements, so that both
E  and FF percentage values were very small for all of the short test pro-
cedures.  This characteristic is summarized in Table 1-9 for the CO FTP
level of 9 gm/mi;  the E  and FF values are less  than 1 percent for all the STs.
              At the 3.4 level,  however, as shown in Table 1-10, the bag
tests were sufficiently discriminatory to identify FF values above 20 percent,
with E values in the 14- to 16-percent range. The volumetric tests, on the
other hand, all had high E  values (30- to 40-percent range), with very low
FF values (<16).
1.1.2.3       Oxides of Nitrogen Emission
              The variations of EC, E  t  and FF as a function of NO  cut-
point were also graphically determined for each ST examined, for the single
NO FTP value of 3. 1 gm/mi examined in the study. Figure 1-5 illustrates
results for the Federal Short Cycle.
                                   1-14

-------
Table 1-7.  Comparison of Selected ST Hydrocarbon Results:
            CEV Fleet, Bounded Errors of Commission Analysis,
            HC FTP Level =0.41 gm/mile (E  = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
Idle
Low Speed
High Speed
Clayton Key Mode (Garage)
Idle
High Speed
Federal Three -Mode (Laboratory)
Idle
Low Speed
High Speed
Federal Three -Mode (Garage)
Idle
Low Speed
High Speed
2500 rpm Unloaded (Laboratory)
2500 rpm Unloaded (Garage)
Parameter, %
E
o
11
7

7
35
30

10
21

17
17
11

18
22
20
38
37
FF
56
60

61
32
37

57
45

52
38
51

46
44
47
28
30
                              1-15

-------
Table 1-8.  Comparison of Selected ST Hydrocarbon Results:
           CEV Fleet,  Bounded Errors of Commission Analysis,
           HC FTP Level = 0.90 gm/mile (E  = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
Idle
Low Speed
High Speed
Clayton Key Mode (Garage)
Idle
High Speed
Federal Three -Mode (Laboratory)
Idle
Low Speed
High Speed
Federal Three -Mode (Garage)
Idle
Low Speed
High Speed
2500 rpm Unloaded (Laboratory)
2500 rpm Unloaded (Garage)
Parameter, %
E
o
9
6.5

6
21
19.5

9.5
16

15
15
10

14
16
15
23
23
FF
22
25

22
9
10

22
16

21
21
25

17
16
17
8
9
                              1-16

-------
    30
    25
    20
o
    10
                                 	EQ
                                 	FF
V       /
A     /

 A
                             *
                                *
                                  »
                      1975 FTP
                      LEVEL
                     —7.0
                      -9.0
                                                          AT 5.0. 7.0. AND 9.0
                                                               FF < 5%
                               CO CUT-POINT, gm/mi
      Figure 1-4.   Variation of Ec,  Eo, and FF with CO Cut-Point;
        CEV Fleet; Federal Short Cycle; Bounded Errors of Com-
        mission Method
                                    1  17

-------
Table 1-9.  Comparison of Selected ST Carbon Monoxide Results:
           CEV Fleet, Bounded Errors of Commission Analysis,
           CO FTP Level =9.0 gm/mi (E  = constant = 5%)
           Short Test
                                            Parameter,  %
                                                        FF
 Federal Short Cycle
 NY/NJ Composite
 Clayton Key Mode (Laboratory)
     Idle
     Low Speed
     High Speed
 Clayton Key Mode (Garage)
     Idle
     Low Speed
     High Speed
 Federal Three-Mode (Laboratory)
     Idle
     Low Speed
     High Speed
 Federal Three-Mode (Garage)
     Idle
     Low Speed
     High Speed
 2500 rpm Unloaded  (Laboratory)
 2500 rpm Unloaded  (Garage)
                             1 18

-------
Table 1-10.  Comparison of Selected ST Carbon Monoxide Results:
            CEV Fleet,  Bounded Errors of Commission Analysis,
            CO FTP Level = 3.4 gm/mi (E  = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
Idle
Low Speed
High Speed
Clayton Key Mode (Garage)
Idle
Low Speed
High Speed
Federal Three -Mode (Laboratory)
Idle
Low Speed
High Speed
Federal Three -Mode (Garage)
Idle
Low Speed
High Speed
2500 rpm Unloaded (Laboratory)
2500 rpm Unloaded (Garage)
Parameter, %
E
0
14
16

29
33
36

29
36
33

33
40
43

29
36
36
35
38
FF
22
20

15
11
8

14
7
10

16
8
7

14
7
7.5
8
6
                              1-19

-------
    25
    20
     15
 j°

 "o
     10
o

LU
Q_
                                                    FF<5.
     0
                          5                   6

                          NOX CUT-POINT, gm/mi
            Figure  1-5.  Variation of Ec, Eo, and FF with
             NOX Cut-Point; CEV Fleet; Federal Short
             Cycle; Bounded Errors of Commission Method
                                1-20

-------
               The significant results at the E  level of 5 percent are
summarized in Table 1-11 for comparative purposes. As can be noted, the
high-speed mode of the volumetric tests (Clayton Key Mode and Federal
Three-Mode) produced the highest FF values and the lowest E  values, and
are thus indicated to be superior for NO  discrimination purposes.
                                       Ji
1.1.2.4       Weighted Three-Mode  Tests
               Contingency table analyses were also made for two different -
weighted Key Mode  tests. The results indicated that the weighted volumetric
tests are not significantly better than the best single  mode, as was also con-
cluded from conventional correlation  analyses (see Sec.  1. 1. 1.5).
1.1.2.5       Variance Effects
               Since the variations in E  , E , and FF with ST cut-point noted
                                      co                     r
previously are predictions from the data, the variability of the predictions
was analyzed.  The uncertainty in the predicted results increases when
decreasing the bounds of the errors of commission, as illustrated in Figure 1-6
for HC  on the Federal Short Cycle.
1.1.2.6       Laboratory vs Garage  Instruments
               Tables 1-7 through 1-10 indicate that generally similar  levels
of E  and FF were obtained with both laboratory and  garage analyzers  for the
HC and CO ranges examined for the CEV fleet.
1.1.2.7       Modal vs Bag Tests
               In terms of HC and CO discrimination, as  noted above,  the bag
tests are superior to the modal ST.  The modal STs all have high EQ and low
FF values.   In terms of NO   discrimination, the high-speed mode of the
                          Ji
volumetric ST was superior.   These results agree  with those predicted from
conventional correlation analysis in Sec.  1. 1. 1.4.
1.1.2.8       Multiple-Constituent Tests
               In addition to analyzing each pollutant  individually, an analysis
was made for three-constituent tests.  In a three-constituent  test,  a car fails
                                    1-21

-------
     Table 1-11.  Comparison of Selected ST NO  Results:
                  CEV Fleet, Bounded Errors or Commission Analysis,
                  NO  FTP Level = 3.1 gm/mi (E   = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)* '
Idle
Low Speed
High Speed
fa)
Federal Three -Mode (Laboratory)* '
Idle
Low Speed
High Speed
2500 rpm Unloaded (Laboratory)^'
Parameter, %
E
o
9.5
10

13
14
6.5

11
11
3
13
FF
5.5
5

2
<2
8.5

1
1
8.5
2
      (a)  Garage-type analyzers for NO  were not available for ST
          evaluation.
                                      x
the ST if any of its HC, CO,  and NO  measurements exceed the previously
                                  Ji
determined cut-points. These tests are applicable to the bag tests, the
unloaded test, and the individual modes of the three-mode volumetric tests.
              The three-constituent test results for the Federal Short Cycle
and the Federal Three-Mode (high speed and idle modes only) were computed
and graphically summarized as a function of predicted E  .   Table  1-12 sum-
marizes these results for the predicted E value of 5 percent.  Both the
laboratory and garage instrument results are displayed for the Federal Three-
Mode short test.
                                    1-22

-------
 01
                           I
              ERROR BAR INDICATES PLUS AND MINUS ONE
              STANDARD DEVIATION OF THE ESTIMATE
234567
    HC FEDERAL SHORT CYCLE, gm/mi
Figure 1-6.  Variability of Predicted Population Results
                            1-23

-------
Table 1-12.  Comparison of Three-Constituent Test Results:
            CEV Fleet,  Bounded Errors of Commission Analysis,
            (Predicted E  = constant = 5%)
Short Test
Federal Short Cycle



Federal Three -Mode
(Laboratory)
Idle



High Speed



Federal Three -Mode
(Garage)
Idle



High Speed



FTP
Level
1
2
3
4

1
2
3
4
1
2
3
4

1
2
3
4
1
2
3
4
Parameter, %
Actual EC
5
8
7.5
7.5

0
3
3
13
3.5
9.5
9.5
13

10
2.5
0
5
10
15
10
5
Actual E
o
16
16
8
6

36
28
16
16
16
13
10
7

36
25
14
13.5
30
18
6
6.5
Actual FF
48
25
20
17

36
17
16
7
54
35
23
16

30
18
16
11.5
36
25
24.5
18
                             1-24

-------
              With laboratory instrument measurements,  as the  FTP
cut-points increase from level Set I to level Set IV. the resulting actual errors
of commission tend to increase for the given predicted level of errors of com-
mission.  This trend is not present for the garage instrument results shown.
              A comparison of the modes on the Federal Three-Mode test
shows that,  for the fixed  predicted percent E , the high  speed mode has a
higher percent FF and lower percent E than does the idle mode.  This is true
regardless of instrumentation or FTP level.  However, the actual percent E
is generally lower on the  idle mode than on the high-speed mode, but this
difference is not always significant.
              A comparison of different modes or ST should be made on a
fixed actual percent E  basis.  This is, of course,  difficult to do because of
the computational procedure followed.  It  can be approximately performed,
however.  Consider comparing the Federal Short Cycle to the Federal Three-
Mode.  At FTP level I, the actual  percent E  is approximately the  same for
the high-speed mode and  the Federal Short Cycle (statistically, they are the
same). Now, comparing  the percent FF and percent E   values, percent FF
and percent E  are both higher on  the high-speed mode than the Federal Short
Cycle.  This difference is not statistically significant at  the 95 percent  confi-
dence level,  and the two tests would have  to be judged as equal.  Also,  at the
95 percent confidence level, the high-speed mode is superior  to the idle mode.
              The  differences between laboratory and garage instruments
are quite predictable,  based upon the  previous  results from individual pollu-
tants.  For the fixed predicted percent E  , on their respective modes,
        a.    Actual percent E  is higher for garage instruments than for
              laboratory instruments
        b.    Actual percent FF is lower for garage  instruments than for
              laboratory instruments
        c.    Actual percent E  is higher for garage instruments than for
              laboratory instruments.
                                    1-25

-------
 1.2           IN-USE 1974 MODEL. YEAR VEHICLE FLEET
 1.2.1         Direct Relatability Results
              A conventional correlation analysis was made for the 1974
 model year fleet to assess direct relatability between the five  short tests
 and the FTP. The method was the same as described for the CEV fleet in
 Sec. 1.1.1.  The resulting ST/FTP correlation coefficients for HC, CO,
 and NO  are summarized in Table 1-13 for the three individual inertia test
       ji
 weight groups (A = 4000 Ib, B = 2750 Ib, and C = 5500 Ib) and for the pooled
 vehicle population (combined groups A, B,  and C).
 1.2. 1. 1      Hydrocarbon Emission
              For the pooled fleet, the bag-type STs (Federal Short Cycle
 and NY/NJ Composite),  the idle mode of the modal STs with laboratory
 analyzers, and the Unloaded 2500 rpm ST with laboratory analyzers in gen-
 eral exhibit the better HC tracking characteristics.
              For Group A, similar results apply.
              For Group B, the  results are similar to the pooled fleet
except that in some instances the low-speed mode of the Key Mode and the
low and high-speed modes of the  Federal Three-Mode test have a. slightly
higher correlation coefficient than the idle mode.
              For Group C,  none of the STs are able to track HC with any
reasonably high  degree of correlation.
 1.2.1.2       Carbon Monoxide Emission
              For the pooled fleet, the bag-type STs, the idle and low-speed
modes  of the modal tests with laboratory analyzers, and the Unloaded 2500 rpm
ST with laboratory analyzers in general exhibit the better CO tracking
characteristics.
              For Groups A and B, similar results apply except that the
low-speed  mode is superior to the idle mode in the modal tests.
                                   1-26

-------
Table 1-13.  Correlation Coefficient Summary:
             1974 Model Year Fleet

Short Test


Federal
Short
Cycle


NY/NJ
Composite


Key Mode
(Laboratory)










Key Mode
(Garage)





Vehicle
Group 'a'


Pooled

A
B
C
Pooled
A
B
C
Pooled


A


B


C


Pooled


A



Test
Mode











High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle

N(b)


147

50
48
49
147
50
48
49
147


50


48


49


145


50


ST/FTP Correlation
Coefficient*0)

HC

0.932

0.933
0.897
0.383
0.906
0.9H
0.920
0.513
0.757
0.776
0.793
0.590
0.595
0.723
0.812
0.868
0.825
0.238*
0.228*
0.460
0.528
0.545
0.455
0.228*
0.151*
0.245*

CO

0.905

0.972
0.897
0.476
0.890
0.950
0.857
0.498
0.518
0.769
0.739
0.514
0.827
0.704
0.262*
0.738
0.650
-0.195*
0.435
0.757
0.507
0.472
0.470
0.563
0.652
0.372

NO
X
0.355

0.780
0. 104*
0.674
0.060*
0.733
0.005*
0.611
0.521
0.419
0.463
0.562
0.495
0.381
0.731
0.635
0.548
0.555
0.580
0.571






                      1-27

-------
Table 1-13.  Correlation Coefficient Summary:
             1974 Model Year Fleet (Continued)
Short Test





Federal
Three -Mode
(Laboratory)









Federal
Three -Mode
(Garage)



Vehicle
Group'3-'
B


C

Pooled
A


B


C


Pooled
A


Test
Mode
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
N(b)
46


49

147
50


48


49


145
50


ST/FTP Correlation
Coefficient^)
HC
0.478
0.765
0.692
0. 191*
0.198*
0. 100*
0.766
0.771
0.803
0. 507
0.523
0.709
0.890
0.859
0.851
0.522
0.533
0.252*
0.474
0.531
0.632
0. 138*
0. 107*
0.660
CO
0.362
0.540
0.560
-0.221*
-0.091*
0.229*
0.604
0.729
0.734
0.717
0.801
0.724
0.278*
0.737
0.622
0.159*
0.592
0.733
0.387
0.409
0.476
0.533
0.597
0.397
N0x





0.467
0.453
0.411
0.492
0.664
0.369
0.722
0.611
0.665
0.552
0.707
0.639



	 . 	
                     1-28

-------
              Table 1-13.  Correlation Coefficient Summary:
                           1974 Model Year Fleet (Continued)
Short Test






2500 rpm
Unloaded
(Laboratory)

2500 rpm
Unloaded
(Garage)

Vehicle
Group**)
B


C


Pooled
A
B
C
Pooled
A
B
C
Test
Mode
High
Low
Idle
High
Low
Idle




N
46


49


147
50
48
49
147
50
46
49
ST/FTP Correlation
Coefficient^)
HC
0.536
0.763
0.717
0.095*
-0.008*
-0.060*
0. 809
0.832
0.865
0. 107*
0.574
0.487
0.781
-0.064*
CO
0.268*
0.539
0.550
-0.083*
0.239*
0.392
0.740
0.812
0.724
0.350
0.447
0.676
0.684
-0.051*
NO
X






0.447
0.524
0. 577
0.679


(a)A = Chrysler (4000 Ib)
  B = Ford (2750 Ib)
  C = Chevrolet (5500 Ib)
  Pooled = Groups A + B + C

  'Number of cars in the data set
(c)
  'The correlations are statistically significant at the 95 percent confidence
  level  except where indicated by an asterisk.  ST and FTP uncorrelated
  for correlations below 0.28.
                                   1-29

-------
              For Group C, the idle mode of the Key Mode and Federal
Three-Mode (with laboratory analyzers) are superior, although the low-
speed modes of these STs and the bag-type tests are correlated with CO at
lower correlation coefficient levels.  The  other STs are essentially uncor-
related with CO for Group C.
1.2.1.3      Oxides of Nitrogen Emission
              For the pooled fleet,  all modes of the Key Mode and Federal
Three-Mode and  the Unloaded 2500 rpm ST have similar correlations in the
0. 41 and 0. 52 range; the Federal Short Cycle correlates at a lower value
(0. 36),  while the NY/NJ Composite bag test  is uncorrelated.
              For Group A, however, the bag-type STs have the highest
correlation coefficients observed for NO   (0.73 to  0.78), while the modal
                                      2£
and Unloaded 2500 rpm ST results are similar to those of the pooled fleet.
              For Group B, the results are  similar to those  for the pooled
fleet except that the bag-type STs are not correlated at all for NO  .  Here the
                                                               J*.
range of correlation coefficients for the modal and Unloaded 2500 rpm STs is
from 0.55  to 0.73, with the highest values obtained in the high-speed mode.
              The Group C results are similar to  those for Group A.
              There is no single ST with good NO  correlation across the
                                               JL
1974 model year  fleet population.
1.2.1.4      Modal vs Bag Tests
              In terms of HC and CO emissions correlation,  the bag-type
STs are superior for Groups A, B, and the pooled  population.  For Group C
the idle mode of the Key Mode and Federal Three-Mode STs has the higher
correlation for CO; the idle mode of the Key Mode  and the low- and high-
speed modes of the Federal Three-Mode are essentially the same  as the bag-
type STs in terms of HC discrimination capability.
              In terms of NO  correlation,  the bag-type and modal STs are
essentially equivalent for Group C, whereas the bag-type STs are  clearly
superior for Group A.   The modal tests are  superior for Group B  and the
pooled fleet.
                                   1-30

-------
1.2.1.5       Laboratory Analyzers vs Garage Instruments
               From an HC and CO correlation viewpoint, the garage
analyzers are inferior  to the laboratory analyzers in that they have lower
correlation coefficients than the laboratory analyzers for HC and CO in each
corresponding test mode.  They do, however, tend to identify the  same
superior test modes as the laboratory analyzers, and can have reasonably
high correlation coefficients, although there is a wide variation for the three
groups examined in the 1974 model year fleet.
               To illustrate, consider the idle mode of the Federal Three-
Mode ST, which for the pooled fleet resulted in representatively high corre-
lation  coefficients for HC and CO with laboratory analyzers: 0.80 and 0.73,
respectively.  With garage instruments,  these correlation coefficients
dropped to 0.63 and 0.48, respectively.
               In the  case of Group A,  the laboratory analyzer HC and CO
values were 0.71 and 0.72, whereas the garage instrument values were
reduced to 0.66 and 0.40.
               Group B HC and CO values for laboratory analyzers were 0. 85
and 0.62.  With garage instruments, they were lowered to 0.72 and 0.55.
               In the  case of Group C,  the HC correlation coefficient of 0. 25
with laboratory instruments was not statistically significant at the 95 percent
confidence level, while the CO correlation coefficient was 0.73.  With garage
instruments, these values dropped to -0. 06 and 0. 39, respectively.   In addi-
tion, all other ST test modes with garage instruments were uncorrelated for
HC and CO for Group C.  This was the only group exhibiting these charac-
teristics, although it also had generally poorer HC and CO correlation
coefficients than the other groups when laboratory analyzers were used.
               This group-peculiar characteristic raises the issue as to
whether it is related  to inertia test weight factors or to vehicle manufacturer,
since each inertia test  weight group was made by a different automotive
company.  There are insufficient data to evaluate this issue at this time;
however, a  comparison can be made between the 2750-lb Pintos of Group B
                                    1-31

-------
above and the 5000-lb Galaxies of the CEV fleet in Section 1. 1, since both
were manufactured by the Ford Motor Company.
              Again, using the idle mode of the Federal Three-Mode ST for
comparison purposes, the use of garage instruments instead of laboratory
instruments for the CEV Galaxies reduced the HC correlation  coefficient
from 0.80 to 0.78, and increased the CO correlation coefficient from 0.48
to 0.52 (see Table 1-1). These ranges are similar to those reported above
for Group B (Pintos), even though the  Galaxies were  catalyst-equipped and
the Pintos were not.
              Thus,  it appears that additional examinations may be required
of possible manufacturer-related effects (e.g.,  idle fuel-air ratio tolerance
bands and quality control measures) in order to fully understand their impact
upon measurement instrument type for short test purposes.
1. 2. 1.6      ST  Ratings
              ST  ratings,  using the scale  established for the CEV fleet in
Sec. 1.1.1.8, are given in Table 1-14.  As can be seen, no single ST per-
forms  consistently well on all three individual groups,  or on a pooled basis.
Generally, the STs are unable to track HC and CO emission levels on
Group  C.
              As with the CEV fleet,  the bag-type STs have higher ratings
than the volumetric tests.  The Unloaded 2500 rpm ST  shows substantially
higher correlation for the 1974 model  year fleet than for the CEV fleet (as
shown  in Table 1-4).  The extreme CO tracking deficiency for the CEV fleet
data is  not evident for the 1974 model  year fleet.
1.2.2         Contingency Table Analysis Results
              A contingency table analysis,  using the bounded errors of
commission method described in Sec.  1. 1.2 for the CEV fleet, was  also
performed for the  1974 model year fleet, with the results as discussed
below.
                                   1-32

-------
Table 1-14.  ST Ratings:  1974 Model Year Fleet
Short Test
Federal Short Cycle



NY/NJ Composite



Key Mode
(Laboratory)


Key Mode
(Garage)


Federal Three -
Mode (Laboratory)


Federal Three -
Mode (Garage)


Vehicle
Group(a)
Pooled
A
B
C
Pooled
A
B
C
Pooled
A
B
C
Pooled
A
B
C
Pooled
A
B
C
Pooled
A
B
C
Ratings (b)
HC
E
E
G
P
E
E
E
P
G (I)(c)
G (I)
G (L)
P (D
P (D
U
G (L)
U
G (I)
G (I)
G (H)
P (L)
F (I)
F (I)
G (L)
U
CO
E
E
E
P
G
E
G
P
G (L)
G (L)
G (L)
G (I)
P (H)
F (L)
P (D
U
G (I)
G (L)
G (I)
G (I)
P (D
P (L)
P (D
P (D
NO
X
P
G
U
F
U
G
U
F
P (H)
P (H)
G (H)
P (L)



P (H)
F (L)
G (H)
G (L)




                      1-33

-------
            Table 1-14.  ST Ratings:  1974 Model Year Fleet
                         (Continued)
Short Test

2500 rpm Unloaded
(Laboratory)


2500 rpm Unloaded
(Garage)


Vehicle
Group la>

Pooled
A
B
C
Pooled
A
B
C
Ratings *b)
HC

G
G
G
U
P
P
G
U
CO

G
G
G
P
P
F
F
U
NO
X
P
P
P
F




(a)A = Chrysler (4000 Ib)
(b)
B = Ford (2750 Ib)
C = Chevrolet (5500 Ib)
Pooled = Groups A + B + C

Rating scale as in Sec. 1.1.1.8
 I . idle
  L = low speed mode
  H = high speed mode
                                  1-34

-------
1.2.2.1       Hydrocarbon Emission
              The variation of EC>  EQ, and FF as a function of HC cut-point
was graphically determined for each ST.  The results for the Federal Short
Cycle are shown in Figure 1-7 to indicate the general trends.   All STs had
similar trends.   To illustrate  specific values and trends among the STs,
Table 1-15 summarizes data for the E  value of 5 percent.
              The bag tests (Federal Short Cycle and NY/NJ Composite)
have lower EQ and higher FF at the fixed E  = 5 percent condition than do
the volumetric tests.  There is little difference shown between the various
volumetric STs.
1.2.2.2       Carbon Monoxide  Emission
              The variations of  E , E , and FF as a function of CO cut-point
                                 co                                r
were also graphically determined.  Figure 1-8 indicates results for the
Federal Short Cycle.
              To illustrate specific values and trends among the STs,
Table 1-16 summarizes data from the figures for the E   value  of 5 percent.
              The bag-type STs (Federal Short Cycle and NY/NJ Composite)
exhibit excellent CO tracking characteristics; the EQ values are considerably
better (lower) than the volumetric tests, and the FF values are the highest.
When garage-type instruments are used, the E values  are essentially doubled
(over laboratory instrument values) and FF values are significantly  reduced.
1.2.2.3       Oxides of Nitrogen Emission
              Figure 1-7 also indicates the variation of EC, EQ, and FF as
a function of NO  cut-point for the Federal Short Cycle.
               x    c
              The significant  results for each  ST at the EC level of  5 per-
cent are summarized in Table  1-17 for comparative purposes.   As can be
noted, all STs identified very low percentages of correctly failed vehicles
(FF), less than  5 percent, while having significant errors of omission,
approximately 15 percent.
                                   1-35

-------
   30
   25
   20
o
at.
uu
   10
    0
                    \
                     *
                      *
 '.HC


 \


  \


   \
    X

   /
   /
  /
  /
 /
 /
/HC
                          /   \
                                                     	FF
                                          NOV
      0
                    HC AND  NOX CUT-POINT, gm/mi
    Figure  1-7.  Variation of Ec, Eo, and FF with HC and NOX

     Cut-Points; 1974 Model Year Fleet; Federal Short Cycle;

     Bounded Errors of Commission Method
                              1-36

-------
Table 1-15.  Comparison of ST Hydrocarbon Results:  1974 Model Year
            Fleet, Bounded Errors of Commission Analysis
            (E  = constant = 5%)
              Short Test
                                                Parameter, %
                                                            FF
    Federal Short Cycle
    NY/NJ Composite
    Clayton Key Mode (Laboratory)
        Idle
        Low Speed
        High Speed
    Clayton Key Mode (Garage)
        Idle
        Low Speed
        High Speed
    Federal Three-Mode (Laboratory)
        Idle
        Low Speed
        High Speed
    Federal Three-Mode (Garage)
        Idle
        Low Speed
        High Speed
    2500 rpm Unloaded (Laboratory)
    2500 rpm Unloaded (Garage)
 6.5
 8.5

16
17
18

11.5
14
13

15.5
17.5
18

17
14
12
16
16
34.5
32

24. 5
23.6
22.5

29
27
28

25
23
23

24
27
29
26
26
                               1-37

-------
  651-
  60 -
  55 -
£ 50
  45 -
  40 -
»



- 25


UJ
-o 20
z
UJ°
I—

~~ O A5
• 1 1
o.
10



5
Q
N '
X
N /
% X
\ /
\ /
\ /
\ /
A
/ \
— f \
\
\
\
X »
X \
X >
— x^
i i i r* 	 1 —
15 20 25 30 35 40

                                                                     0
                                                                    FF
                                                          •
                                                              \
                                                             45
50
                                 CO CUT-POINT, gm/mi
     Figure 1-8.   Variation of Ec,  Eo, and FF with CO Cut-Point;
       1974 Model Year Fleet;  Federal Short Cycle; Bounded Errors
       of Commission Method
                                  1-38

-------
Table 1-16.  Comparison of ST Carbon Monoxide Results:  1974 Model
            Year Fleet, Bounded Errors of Commission Analysis
            (E  = constant = 5%)
              c
             Short Test
                                               Parameter^ %
                                             E
             FF
   Federal Short Cycle
   NY/NJ Composite
   Clayton Key Mode (Laboratory)
       Idle
       Low Speed
       High Speed
   Clayton Key Mode (Garage)
       Idle
       Low Speed
       High Speed
   Federal Three-Mode (Laboratory)
       Idle
       Low Speed
       High Speed
   Federal Three-Mode (Garage)
       Idle
       Low Speed
       High Speed
   2500 rpm Unloaded (Laboratory)
   2500 rpm Unloaded (Garage)
 7
 8

19
18
35


35
35
37

20
20
29

35
31
30
19
33
65
64

53
54
38


38
38
35

53
52
43

37
41
42
53
40
                               1-39

-------
    Table 1-17.  Comparison of ST NOX Results: 1974 Model Year Fleet,
                Bounded Errors of Commission Analysis
                (E  = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
Idle
Low Speed
High Speed
Federal Three -Mode (Laboratory)
Idle
Low Speed
High Speed
2500 rpm Unloaded (Laboratory)
Parameter, %
E
o
14.5
16.5
13.5
14
13.5
14
14
14
14
FF
3
1.5
<5
<5
<5
<5
<5
<5
4
1.2.2.4       Single-Constituent Test Results
              On the average,  the bag-type tests have lower E  and higher
FF for a fixed rate of E  than do the volumetric tests.  However^  FF rates
in the 30 percent range can be achieved with any of the tests.  For a fixed
percent FF, the percent E  is determined since the sum of FF and E  is the
FTP rejection rate. Thus,  the "best" test for fixed percent FF is the one
with the lowest percent E  .  In general, the bag-type STs are better in this
respect.  However, the actual level of percent E  on the volumetric tests is
still quite low.   For example, at 30 percent FF on the CO Federal Short
Cycle, the percent  E  is  essentially zero.  For CO on the Key Mode low-
speed mode, percent E  is 0.65 percent for laboratory instruments  and
3. 85 percent for garage instruments.
                                   1-40

-------
1.2.2.5
Multiple-Constituent Tests
              In addition to analyzing each pollutant individually, an analysis
was made for multiple-constituent tests.  The method of analysis and compu-
tational procedures were the same as for the CEV fleet,  as discussed in
Sec. 1. 1.2.8.
              Three-constituent test results for the Federal Short Cycle
and the Federal Three-Mode (high-speed and idle modes only) were com-
puted and graphically summarized as a function of predicted E  .  Table 1-18
summarizes these results using laboratory instruments for predicted E
values <2 percent.
              For the actual percent E  less than 2 percent, the laboratory
results of the Federal Three-Mode and the Federal Short Cycle are com-
parable.  Table 1-18 indicates the minimum and maximum for percent FF
and percent E , while percent E  is less than 2 percent.  There is little
difference between the idle mode and the Federal Short Cycle.  Over this
range of percent E  ,  the idle mode would appear favorable to the Federal
Short Cycle due to the lower value of percent E  on the idle mode.
            Table  1-18.  ST Comparison:  1974 Model Year Fleet,
                        Multiple Constituent Tests
                        (Actual E S 2%)
Short Test
Federal Short Cycle
Federal Three -Mode
Idle
High
% FF
Min
25

22
5
Max
36

38
42
% E
o
Min
44

42
38
Max
55

58
75
               A comparison of instrument types showed that the laboratory
instruments are  generally preferable.
                                   1-41

-------
 1. 3           DEFECT DATA FROM CATALYST-EQUIPPED
               EXPERIMENTAL VEHICLE FLEET
 1.3.1         Nature of Defects and Statistical Impact
               Upon completion of the FTP and ST tests performed on the
 CEV fleet as described in Sec. 1. 1, 95 defect tests were performed on 5 of
 the vehicles from the 40-vehicle CEV fleet.
               The 95 defect tests simulated a wide variety of malfunctions
 which could occur in typical passenger  cars.  The general categories of
 defects are defective ignition components; changes in ignition timing, dwell,
 and spark advance; faulty carburetion; defective valves; clogged air filters;
 and faulty emission control components.  The defects were introduced indi-
 vidually and mixed.
               Correlation analyses were performed  to determine the statis-
 tical character of the  defect test data.  Many of the defect tests were either
 replications or produced similar data.  The HC correlations  are consistently
 higher, over 0. 9, among the defect data than the previous 40-car CEV fleet.
 Addition of all defect data to the original CEV fleet data would significantly
 distort the population  characteristics with regard to HC.  CO and NO  dis-
                                                                   x
 tortion would also occur, although not as pronounced as with  HC.
               As the assumption of independence of the observations is
 crucial to contingency table analysis, the 95  defect tests were statistically
 pruned to 24 tests representing 24 independent defective vehicles.   These
 data are considered to  represent a population distinct from the original
 40-car population.  Of these 24, 6 had no Federal Three-Mode (laboratory)
 data, and 5 had no Key Mode (laboratory) data.
 1.3.2          Contingency Table Analysis Results
               The analysis proceeded in two stages.  The original  CEV fleet
 population was first analyzed,  using first good  data.  The  analysis method
was the bounded errors of commission procedure which established the ST
                                    1-42

-------
 cut-points (see Sec.  1. 1.2).  Percent E was varied from 10 percent to
 1 percent in 1-percent increments, with the addition of points at 0. 5 percent
 and 0. 1 percent.  Immediately following analysis of the original CEV fleet,
 the defect population was analyzed.  The contingency table results were cal-
 culated for  this population using the cut-points previously determined from
 the original CEV fleet population.  The computations were performed at
 each  of the  EC settings.  Thus, the analysis is merely an assessment of how
 well a test  constructed using data with an unknown mix of normal and defect
 operation will perform on a population of defective vehicles known to represent
 extreme departures from normal operation.  A summary of the analysis on
 each  constituent is given in Table 1-19.  The ST cut-points were established
 for E  less than or equal to 5 percent, and the FTP level was Level I.
 (HC = 0.41  gm/mi,  CO = 3.4 gm/mi, NO  =  3. 1 gm/mi).
                                        Ji
 1.3.2. 1       Hydrocarbon Emission
               In all cases, each ST produced significantly higher FF values
 and lower EQ values for the defect fleet than  for the  original CEV fleet.  The
 percent E   for the defect fleet was  generally lower and varied from 0. 97 to
 8. 68 percent,  as compared with the 5 percent level used in the CEV fleet to
 select the HC cut-point values.
 1.3.2.2       Carbon Monoxide Emission
               Each ST produced significantly higher FF values and lower E
 values for the  defect fleet than for the original CEV  fleet, except for the
 Unloaded 2500 rpm ST with garage  analyzers, where the E   values were
 essentially  the same.  The percent E  for the defect fleet was generally
 somewhat higher than the 5 percent level used in the CEV fleet, varying
from  4.48 to 16. 5 percent.
 1.3.2.3        Oxides of Nitrogen Emission
               Both  FF and E  values were significantly higher  for the defect
fleet than for the original CEV fleet,  for each ST.  The percent E  for the
                                   1-43

-------
Table  1-19.  Defect Analysis Comparison Summar
              Population [% E  = 5, (a) FTP Level
                                                        Predicted

Short Test
Federal Short Cycle


NY/NJ Composite



Key Mode (Laboratory)











Key Mode (Garage)





Federal Three-Mode
(Laboratory)








Federal Three -Mode
(Garage)





2500 rpm Unloaded
(Laboratory)

2500 rpm Unloaded
(Garac'-l

Test
Mode







High



Low



Idle



High

Low

Idle

High



Low


Idle


High


Low

Idle






	 '
No. of
Defect
Cars
24


24



i9











24





18









24






24


24


Pollu-
tant
HC
CO
N0x
HC
CO
NO
X
HC
CO
NO
X
HC
CO
NO
X
HC
CO
NO
X
HC
CO
HC
CO
HC
CO
HC

CO
NO
HC
CO
NO
HC
CO
NO
X
HC

CO
HC
CO
HC
CO
HC
CO
NO
X
HC
CO
Original
CEV Fleet
%EQ
11.0
14. 1
9.60
7.24
16. 1
9.77

30.4
36.0
6.87

35.3
33.0
.13.8

6.79
28.6
13.4

21.8
33. 3
22. 3
36.5
10. 38
29.2
10.8

43. 5
2.93
16.9
40.9
10.9
16.6
33.4
10.6
19.5

36. 1
22.0
36.5
18.0
29.2
38.7
34.9
12.9
37.0
37.7
% FF
55.9
22. 1
5.36
59.6
20. 1
5.19

36.8
7.75
8.69

37 0
10.8
1.76

60.5
15.2
2.20

45.4
10. 1
44.9
6.79
56.8
14. 1
58. 1

6. 10
8.75
52.0
8.68
6.73
52.3
16.1
1.05
47.7

7.21
45.2
6.81
49. 1
14. 1
28.5
8.46
1.83
30.2
5.62
Defect
Fleet
%E0
5.40
6.28
36.6
5.31
7.85
18 4

6.47
22.2
8.55

6.36
17.2
45.0

6.01
10.8
45.4

8.02
23.9
8. 16
32.0
8.03
11.7
9.85

16. 14
5.65
10.6
20. 1
50.5
10.5
17.0
54. 1
8.47

23.6
8. 16
30.6
6.81
12.9
13.7
21.0
47.7
15.0
39.9
% FF
69.0
65.2
16.9
69.1
63.4
35. 19

67.6
48.3
52.2

67.7
53.2
15.8

68. 1
59.7
15.4

66.4
47.4
66.3
39.3
66.4
59.6
71.6

52.4
53.6
70.8
48.5
8.78
70.9
51.6
5. 17
66.0

47.7
65.8
40.7
67.6
58.4
60.7
50.4
5.93
59. 5
31. 37
%*p
Ec
4.21
6. 11
1.22
6.02
9.3
10. 5

2.84
11.4
9.31

2.42
13.8
11. 3

5.56
6.26
8. 34

3.63
12.03
5. 37
16.5
8.68
7.29
4.74

7.48
6.05
4. 54
10. 1
2.30
6.55
10.6
0.88
3.75

11.6
5. 16
13.7
6. 13
4.48
0. 97
10.3
2.26
1. 55
8.74
1.11,
     = ^r,
        > constant for original CEV fleet
Ib)
  HC  = 0.41 cm
  CO   5 4 cm/fi
  NO   3 I Km/r,
                                1-44

-------
defect fleet was generally higher than the 5 percent level used in the CEV
fleet,  varying from 0.88 percent to  11.3 percent.
1. 3.2.4
Multiple-Constituent Tests
               The results of a three-constituent test for the Key Mode
(laboratory) and a nine-constituent test for the Key Mode (laboratory) are
shown in Table 1-20.  These results are typical for all the multi-constituent
test analyses made.  As can be  seen,  the multiple-constituent ST had notice-
able improvements in FF discrimination over values  obtained for the original
CEV fleet,  with essentially no E .
        Table 1-20.   Key Mode Composite TesVa' (Laboratory Data)
Test Type
Three -Constituent
High Speed
Low Speed
Idle
Nine - Cons tituent
Original
CEV Fleet
% FF

27. 5
22. 5
60.0
62. 5
%EC

5.00
5.00
5.00
12. 50
% E0

37. 5
42.5
5.00
2. 50
Defect
Fleet
% FF

89. 5
73. 7
89. 5
94.7
% Ec

0
0
0
0
% E0

10. 5
26. 3
10. 5
5.26
(a)
   % Ec < 5; FTP Level I (HC = 0.41 gm/mi,  CO = 3.4 gm/mi, NOx =
   3. 1  gm/mi)
1. 3.2  5
General Comments
              A review of the above typical results illustrates that the short
tests perform well at isolating a population of defective cars.  This is noted
                                    1-45

-------
by the general tendency for percent FF to increase and percent E  to decrease
in the defect population.  Although percent E  decreased for HC, this was not
generally true for CO and NO .
              The sources of the errors of commission and omission are
twofold.  The first and usual source is that of the test procedures; i.e. ,
measurement errors.  The second  source is due to mixing of defects.  An
observation was classified as a defective car if any component of the vehicle
was defective.  Hence, all the NO  data  analyzed are not representative of
NO defects, for example. The multiple-constituent tests (which tend to
eliminate mixing errors) show a  very high probability, greater than 70 per-
cent,  of detecting defect vehicles (note that all the defective cars failed the
FTP at Level I).
              In conclusion, the  ST/FTP tracking of defective vehicles is
very good.
                                   1-46

-------
1.4            GENERAL OVERVIEW REMARKS
1.4.1          Mode vs Bag ST
1.4.1.1        Individual Pollutants
               In all the analyses conducted, the bag tests were shown to be
technically superior in analyzing HC and CO.  Mode-type STs are preferable
to bag STs when considering their relative performances on NO  .  However,
all STs are deficient in analyzing NO  .  As the dominant variables in both
                                   ji.
fleets  are HC and CO, the bag tests are preferable under these conditions.
               The complexity of implementation of bag-type STs could be a
major deterrent to their universal acceptance.  The mode STs are more
desirable in  this respect, especially if garage-type instruments  are deemed
suitable.  A  clear choice is not possible without a full analysis in which the
objectives and constraints of an implementation procedure are specified.
1.4.1.2        Multiple-Constituent Tests
               The clear superiority possessed  by  the bag-type ST is not
present when comparing tests  on a multiple-constituent basis.   In both the
CEV fleet and the 1974 model year fleet, the Federal Short Cycle is
approximately equivalent to the high-speed mode of the Federal Three-Mode
with laboratory instruments.
1.4.2          Single Mode vs  Weighted Mode Tests
               Analysis of weighted mode  tests  shows  only very minor im-
provements in correlation over a single-mode ST.  As a weighted-mode ST
would be of increased complexity; this option demands little attention.
1.4.3          Garage Instrument vs Laboratory Analyzer
               The garage instruments offer additional tradeoffs within  the
volumetric test area.  Garage instruments reduce  the technical  sophistication
of the ST while, at the same time, reducing the complexity of implementation.
Technically,  the garage instrument tests are  inferior  to the laboratory
                                    1-47

-------
instrument tests in that the garage instruments have higher percent testing
errors for a given modal test.  However;  they provide additional options
under a full-scale tradeoff study.
1.4.4         Correlation Coefficient vs Contingency Table Analysis
              The usefulness of the correlation coefficient is confined to
measuring direct relatability.  It is useful in identifying critical areas for
further research and in providing a relative overview,  such as  ranking of
the ST.
              For analyzing the tradeoff between impact on air quality and
cost to the public, a contingency table analysis which admits a policy decision
is most favorable.  The public costs are defined as those incurred by  the
manufacturer and/or those incurred by the environment.  Constraints  are
easily incorporated and, thus, an appropriate policy or set of policies can
be identified. The method of bounded errors of commission is recommended
as the procedure for contingency table analysis.   The policy decision  is the
bound on percent of allowable  errors of  commission.  The  effect of the
policy is measured in percent FF and percent errors of omission.  Other
measures such as relative impact, discussed below, are also available.  In
short,  it allows the policy-maker  to control quantifiable economic costs and
to assess the impact on air quality.
                                   1-48

-------
1.4.5          Relative Impact on Air Quality
1.4.5.1        By Individual Pollutant
               The FTP standards, or cut-points,  can be interpreted as
establishing the desired impact on air quality in that the FTP cut-points fix
the percent of the population classified as high-polluting vehicles.  If the
FTP were used as the test procedure in  an inspection/maintenance program
which tested all vehicles (i. e. , as the ST),  the  relative impact on air quality
would ideally be 100 percent; that is, all the vehicles that are failures are
in fact identified as such.
               Similarly,  the effectiveness of the various STs can also be
used as a measure of impact on air quality,  where "ST effectiveness" is
defined as:

            or_  ,r  ..               % FF for the  short test           ,,  ..
            ST effectiveness = fl FTP failures  in same population      (U1)

                                   % FF
                               % FF + % E
                                          o

Thus,  on this basis,  the ST is always less effective than the FTP, in
proportion to the percent of errors of omission (E ) associated with a given
ST.  Table  1-21 shows the ST effectiveness values for the 1974 model year
fleet for an E  rate of 5 percent.  These values indicate the relative impact
on air quality of the ST  as compared with the impact of the FTP on air quality,
for the E conditions  shown.
         c
               Actual benefit or impact is dependent upon the user's needs
and constraints.  One measure of benefit would be the tons of pollutant re-
moved from the atmosphere on an annual basis in a given region by the use
of an ST in an inspection/maintenance program.  This can be approximated
by the  relationship:

        Tons removed = ST effectiveness X A pollutant to be removed
                        in population X % population sampled          (1-2)
                                    1-49

-------
          Table 1-21.  Short Test Effectiveness; E  = 5%
                      1974 Model Year Fleet
Short Test
Federal Short Cycle
NY/NJ Composite
Key Mode
Laboratory
Garage
Federal Three -Mode
Laboratory
Garage
2500 rpm Unloaded
Laboratory
Garage
ST Effectiveness^
HC
0.83
0.78

0.58
0.34

0.61
0.41

0.61
0.39
CO
0.90
0.88

0.76
0. 51

0.72
0.48

0.73
0.47
NOX
0. 17
0.06

0.28


0.22


0.22

%FF
HC
34
32

24 (I)(b)
14 (L)

25(1)
17 (I)

25
16
CO
65
64

55 (L)
37 (H)

52 (I)
35 (I)

53
34
NOX
3
1

5 (I)


4(H)


4

(a)
   ST Effectiveness =
   where
                       FF
                     FTP Fails
(b)
   FTPHC Fails = 41.09%
   FTP CO Fails = 72. 35%
   FTP NOX Fails = 17. 8%

 I = idle mode
L = low speed mode
H = high speed mode
                              1-50

-------
       where
          effectiveness =
       and
       A pollutant to be removed in population = average value for the
                                                population of HC,  CO,
                                                or NOX, in tons/year,
                                                in excess of that per-
                                                mitted by the FTP
                                                standard; it is based
                                                on the  FTP failures
                                                and corresponding
                                                emission values ob-
                                                served in the popula-
                                                tion, and vehicle-miles-
                                                traveled characteristics
This relationship ignores those additional benefits  likely to occur if the failed
vehicles were repaired and achieved emission levels below the FTP standards
after repair.
              Equation (1-2) indicates areas  of tradeoff that should be ex-
amined prior to the implementation of a specific inspection/ maintenance
program.  Figure  1-9 depicts one aspect of such tradeoffs.  This figure is
an illustrative plot of Eq.  (1-2) for  two different ST (Federal Short Cycle,
and Unloaded 2500 rpm -with garage instruments) as used for CO emissions.
As indicated in Table 1-21, their effectiveness values are 0.90 and 0.47,
respectively; i. e. , as compared with the CO discrimination capability of the
FTP procedure,  they are 90 and 47 percent as effective as the FTP in iden-
tifying vehicles which fail the FTP test on CO.  Thus, to achieve the same
benefit in total CO pollutant removal, the percentage of the population that
must be  sampled by the Unloaded 2500 rpm ST is approximately double  that
which must  be sampled with the Federal  Short Cycle ST. Alternatively stated,
for any given percent sampling of the population, the use of the Federal Short
Cycle ST would result in approximately double the  amount of CO removed.
              The complexity of program implementation can be measured
in annual cost.   The  cost components would include such items as annual
                                    1-51

-------
IX)
i
o
o
                             FEDERAL
                             SHORT
                             CYCLE
                                             2500 rpm
                                             UNLOADED
            PERCENT VEHICLE POPULATION SAMPLED

     Figure 1-9-   Impact of Percent Population Sampled on CO
       Removed (Illustrative Example Only)
                             1-52

-------
operating expenses, maintenance expenses, and amortized initial development
and installation expenses.  The ST requiring laboratory instrumentation would
have substantial initial procurement costs, and higher annual maintenance and
operating expenses than those using garage instruments.  The bag-type ST
requires more skilled personnel and a CVS station.  The bag ST and multi-
mode tests also require a dynamometer.  Thus, the ST can be ranked
according to cost as follows:
         •     Federal Short Cycle,  NJ/NY Composite
         •     Three-Mode volumetric with laboratory  instruments
         •     Three-Mode volumetric with garage instruments
         •     2500 rpm Unloaded with laboratory instruments
         •     2500 rpm Unloaded with garage instruments
For those inspection/maintenance programs targeted to 100  percent inspection
                                        *
of all vehicles, the above  ranking of ST by cost would appear valid.  However,
if less than 100 percent inspection is envisioned for some reason, then addi-
tional factors should be considered.  For example, the unit cost of a program
(per vehicle) would be expected to decrease as the percent of the population
sampled increases.  Thus, in the example  of Figure 1-9, if  the program were
targeted to a defined level of CO removal,  a cost-benefit analysis might be an
appropriate method to select the ST and the percentage sampled for minimum
cost purposes.  The type of constraint normally imposed on  a tradeoff study
would typically be total annual cost; however,  additional constraints on per-
cent E  or percent rejected (E  plus FF) are also admissible under this
approach.  Other areas of consideration are effective sampling and  site
selection, importance of the pollution source as a function of geographic
location, social impact, etc.
1.4.5.2        Multiple Constituent  Tests
               Short test  effectiveness is also a useful measure of test quality
for the multiple-constituent test,  although  the pollutant removal implications
of Eq.  (1-2) must apply on an individual pollutant basis.  Shown in Table  1-22
                                    1-53

-------
          Table 1-22.  Short Test Effectiveness Values for Multiple
                      Constituent Tests; 1974 Model Year Fleet *a>
Short Test
Federal Short Cycle


Federal Three-Mode
(Laboratory Instruments)
Idle
High
Federal Three-Mode
(Garage Instruments)
Idle
High
ST Effectiveness
0.77
0.373
0.314

0.483
0.568

0.330
0.374
Percent E
Predicted^b)
5
0.05
0.01

5
5

5
5
Actual
8.84
2.04
0.68

0.00
2.72

0.00
0.69
 (a)FTp failures =

   Using bounded errors of commission method of analysis

are the effectiveness values for the Federal Short Cycle and the Federal
Three-Mode.  Comparison  of the test-to-test effectiveness values should,
of course,  be made at points where the actual percent E  is equal; however,
this can be only approximated with the existing data.
              The technical favorability of the Federal Short Cycle is
diminished when comparing on the basis of equivalent percent E .  Although
the Federal Short Cycle effectiveness is 0.77 at actual percent  E  equal to
8. 84,  it is  reduced to 0. 373 and 0. 314 for actual percent E  values of 2. 04
and 0. 68,  respectively.  However,  as shown in Table 1-22,  the effectiveness
values of the high-speed mode of the Federal Three-Mode ST with laboratory
                                   1-54

-------
and garage instruments are 0. 568 (actual percent E  = 2. 72) and 0. 374 (actual
percent E  = 0. 69), respectively.  Comparable effectiveness values for the
idle mode with laboratory and garage instruments are 0. 483 and 0. 330,  re-
spectively, both with actual percent E  equal to 0.  Thus,  in the actual per-
cent E  range below approximately 3, the Federal Three-Mode ST with
garage instruments (idle  or high-speed  mode) is  essentially equivalent to the
Federal Short Cycle in effectiveness while  the Federal Three-Mode ST with
laboratory instruments has a higher  effectiveness than the Federal Short
Cycle.
               Although the favorability of  the laboratory instruments over
the garage instruments persists under this method of comparison, considera-
tion of program complexity could bias test desirability in favor of the Federal
Three-Mode with garage  instruments.
                                    1-55

-------
2.   INTRODUCTION

-------
                            2.  INTRODUCTION
2. 1            BACKGROUND AND OBJECTIVES

               With regard to compliance by vehicles and engines in actual

use with the certification emission  standards established for a vehicle at
the time of its manufacture,  the Clean Air Act of 1970 stipulates in
Sec. 207 (b):

               "If the Administrator determines that

         (i)    there are available testing methods and procedures to
               ascertain whether, when in actual use throughout its
               useful life .... , each vehicle and engine to which
               regulations .... apply complies with the emission
               standards of such regulations,

         (ii)   such methods  and procedures are in accordance with
               good engineering practices, and

         (iii)  such methods  and procedures are reasonably capable
               of being correlated with tests conducted under
               section 206 (a) (1), then --

                     "(1)   he shall establish such methods and pro-
                           cedures by regulation,  and
                     "(2)   at such time as he determines that inspec-
                           tion facilities or equipment are available
                           for purposes of carrying out testing
                           methods and procedures established under
                           paragraph (1), he shall prescribe regula-
                           tions which shall require manufacturers
                           to warrant the emission control device or
                           system of each new motor vehicle or new
                           motor vehicle engine .... for its useful
                           life."

Thus,  there are the essential requirements of "availability," "conformance

with good engineering practices, " and "reasonable correlation with certifi-

cation  test procedures" which must be met prior to  the promulgation of

regulations which impose the in-use warranty provisions of Sec.  207 (b)

upon the motor vehicle manufacturers.
                                    2-1

-------
               The states of New York and New Jersey have developed
 short emission tests for potential use in inspection/maintenance (I/M)
 programs in their areas.  The Clayton Manufacturing  Company also devel-
 oped a short test procedure for use in I/M programs.  More recently, the
 EPA has  developed  short tests similar to  those of New York,  New Jersey,
 and Clayton.  Thus,  there are a number of tests "available" to determine
 the exhaust emissions of in-use vehicles;  these test methods and procedures
 "conform with good engineering practices11 in that they utilize well-
 recognized emission-measurement equipment and techniques.
               These tests are "short" in  duration (approximately 3 to 5
 minutes)  in order to (a) minimize the inconvenience of the motoring public
 (and thereby maximize cooperation), and  (b) minimize capital costs of
 inspection stations by maximizing  the number of vehicles a  given facility
 could test.  They have been  structured for "simplicity" in order to (a)
 reduce the potential for procedural errors, and (b) to  reduce test costs.
 As a result, all such tests require that the vehicle  be  tested in a "hot"
 condition; i.e., at its normal operating temperature.
               There remains the requirement to demonstrate "reasonable
 correlation with certification test procedures, " i. e. , with the Federal Test
 Procedure (FTP) used in the certification of new motor vehicles.  Therefore,
 the present study was performed with the  principal objective of  analyzing
 emission  data from  both short tests (STs) and FTP tests of the same vehicles
 in order to determine the degree of "correlation" which exists between vehicle
 exhaust emissions as determined by an ST and the FTP.   A second objective
 was to analyze continuous trace data from these tests to form the basis for
 the development of a new and "better correlating"  short test procedure,
 should the need occur.
 2.2            STUDY SCOPE
               The basis for the analyses  was ST  and FTP data  from three
vehicle fleets:
                                   2-2

-------
        a.    Catalyst-Equipped Experimental Vehicle (CEV) Fleet

              This fleet comprised 40 catalyst equipped "1975-
              prototype" models that had been operated in California
              in Ford vehicle test programs.  These vehicles were
              tested by Olson Laboratories in Anaheim,  California.

        b.    In-Use  1974 Model Year Vehicle Fleet

              This fleet comprised 147  in-use 1974 model year  cars
              in three groups of approximately 50 cars each, repre-
              senting different inertia weight  classes (subcompact,
              intermediate, and full size) and three  different auto
              manufacturers.  These vehicles were  procured by
              Olson Laboratories,  Livonia, Michigan, from the
              greater Detroit area and  tested by EPA in the Ann
              Arbor test facility.

        c.    Defect Test Fleet
              This fleet comprised five of the catalyst-equipped Ford
              vehicles from the CEV fleet noted above.  Approximately
              95 "defect" tests were  conducted on these vehicles. The
              defect tests included such items as spark plug misfiring,
              carburetor misadjustment,  defective valves,  and
              degraded catalysts.   These tests were performed by
              Olson Laboratories,  Anaheim,  California.

              Each of the above vehicles was  tested by the FTP and the

following STs:

        •     Federal Short Cycle

        •     NY/NJ  Composite

        •     Clayton Key Mode
        •     Federal Three-Mode

        •     Unloaded 2500 rpm

For the volume trie-type tests (Clayton Key Mode,  Federal Three-Mode,

and Unloaded 2500 rpm),  both laboratory and garage-type instruments were

used to record HC and CO measurements.  Garage-type instruments were

included in the event that higher-accuracy laboratory analyzers  would not

be compatible with the working environment of a typical automotive  garage

or a large-scale vehicle testing station.  All the NOx readings were made
                                    2-3

-------
with laboratory analyzers due to the unavailability of an appropriate
garage-type NO  instrument.
2.3            METHOD OF APPROACH
               The primary thrust of the work performed under this contract
was statistical in nature.  Two complementing methods were employed to
assess Sec. 207 (b) correlation --a conventional correlation analysis and a
contingency table analysis.   The conventional correlation analysis addresses
the question of direct relatability between the ST and the FTP by examining
the relationships present in the data.  The results are of  great usefulness
in indicating the extent to which each ST tends to track the FTP.  The
contingency table analysis addresses the relatability of ST and FTP on a
pass-or-fail level.  Each data point is examined, and a determination is
made as to whether the auto passed or failed the FTP and passed or failed
the ST.  Thus,  errors of commission (E ),  errors of omission (E  ),
correct passes  by each test (PP), and correct fails by each test (FF) are
identified. Hence the technique allows for the study of the tradeoffs between
errors and correct identifications.
               The conventional correlation  analysis, being purely an
analysis of the data, does not permit policy  decision as a variable or
parameter.  Contingency table analysis, on  the other hand, permits the
integration of policy decision in that it provides  for the determination of
the ST pass/fail cut-points. Thus, policy decision entered the analysis as
a quantifiable variable, and a study indicating the impact  of various policies
was performed in the contingency table  analysis.  One important method
reflecting impact to policy is that of the method  of bounded errors of com-
mission.  In  this scheme, limits are set on  the maximum permissible per-
centage of errors of commission, and the ST cut-points are selected to
yield minimum errors of omission within this constraint.  This analysis
permits a direct answer to  the question, "For a given permissible level of
errors of commission,  what level of errors of omission is associated with
                                    2-4

-------
a given test, and with what impact on air quality (inferred from the
percentage of FF and  EQ vehicles)?11

               These two methods of analysis,  each representing different
interpretations of Sec. 207 (b) correlation, were applied to both the CEV

fleet and the 1974 model year in-use fleet.  They were also applied to the

defect test fleet to (a) determine the statistical character of the specific

defect tests, and (b) to examine the ability of the STs to detect defective
vehicles  of this nature.

2.4            ORGANIZATION OF REPORT

               The  results  of the  study are reported in the following order
and context:

         Section 3 - Test Characteristics and Procedures
                    Defines the five short tests used, describes
                    the test conditions and procedures, and dis-
                    cusses the composition of the three test
                    fleets.

         Section 4 - Catalyst-Equipped Experimental Vehicle Fleet

                    Defines and discusses, for the CEV fleet, the
                    statistical analysis techniques and results for
                    the correlation and contingency table analyses
                    conducted.

         Section 5 - In-Use 1974  Model Year Vehicle Fleet

                    Defines and discusses, for the 1974 model year
                    fleet,  the statistical analysis techniques  and
                    results for the correlation and contingency table
                    analyses conducted.

         Section 6 - Defect Data from Catalyst-Equipped  Experimental
                    Vehicle Fleet
                    Defines and discusses the  analysis techniques and
                    results from the analyses  made to determine the
                    statistical character of the defect tests and to
                    examine  the ability of the various short tests to
                    detect defective vehicles.
                                    2-5

-------
3.   TEST CHARACTERISTICS AND PROCEDURES

-------
             3.  TEST CHARACTERISTICS AND PROCEDURES
              In this program, five short tests (STs) and the 1975 Federal
Test Procedure (FTP) were performed on three test fleets. This section
defines the various STs, describes the test conditions, and discusses the
composition of the test fleets.
3.1           SHORT TESTS
3. 1. 1         General
              Two classes of STs were involved, and these may be catego-
rized as (a) modal or volumetric and (b) as driving trace or CVS.  Both sets
of nomenclature are used in this report, depending upon the aspect of the
test structure that is pertinent to the discussion.  In the modal tests, the
test technician operates  the vehicle on a dynamometer at a fixed vehicle
speed and dynamometer  load, or unloaded at a fixed  engine rpm, or  at idle.
The vehicle tailpipe exhaust is sampled directly,  and the concentration of
each pollutant is measured and recorded in percent,  of in parts per  million,
of the undiluted exhaust.  Three modal STs were used:
         •     Clayton Key Mode
         •     Federal Three-Mode
         •     Unloaded  2500 rpm
The Clayton Key Mode and Federal Three-Mode STs each had high speed,
low speed, and idle modes.
              For  the second class of ST,  the test technician drives the
car on the dynamometer in accordance with a prescribed driving pattern on
a driving trace.  The vehicle exhaust is diluted by the CVS procedure, and
a single  sample bag of diluted exhaust is collected for the whole ST.  The
dilute sample is analyzed and the  results usually expressed in grams per
mile.  This procedure requires the same equipment, sampling procedure,
                                   3-1

-------
and analytical equipment as the Federal Test Procedure (FTP) used in the
certification of new vehicles.  The difference is that the driving trace for
the ST is much shorter and simpler.  Two CVS-type STs were used:
        •     Federal Short Cycle
        •     Composite of New Jersey Acid and New York Short
              Test (NY/NJ)
              Both classes of ST involved approximately 2 or 3 minutes
of driving time on the dynamometer.  All STs were  performed with the
engine at its normal operating temperature;  i.e., "hot" tests.
              The HC and CO content of the exhaust gas in the volumetric
tests was measured from samples taken at the same time by two different
classes of instruments.  One  set, called "laboratory analyzers," was
identical (except for range) with the high-accuracy analyzers used in CVS
certification testing.   The second set, called "garage instruments," used
a lower-cost,  lower-accuracy and precision instrument of the type currently
in use  by many automotive service stations for routine  diagnostic work.
The structure of each test is given below.
3.1.2         ST  Definition
3.1.2.1
Clayton Key Mode
              The Clayton Key Mode is a well-known test which has been
in use for several years for diagnostic  emissions testing.
Vehicle
Weight
Class,
Ib
2000 to
2750
2800 to
3750
3800
and up
Transmission
Range /Gear
In lower
gear (3rd)
Drive or
high gear
Drive or
high gear
Dynamomete r
Load, hp
@ mph
15 @ 38
24@46
30 @ 50
Modes
High
Speed
Cruise,
mph
36 to 38
44 to 46
48 to 50
Low
Speed
Cruise,
mph
22 to 25
29 to 32
32 to 35
Idle
Automatic
transmission
drive

                                   3-2

-------
3.1.2.2
Federal Three-Mode
              The Federal Three-Mode differs from the Clayton Key Mode
in that it uses dynamometer loadings simulating the average power that
occurs at the appropriate speeds in the FTP where the vehicle is accelerat-
ing  (decelerations are not included).  This results in  a higher dynamometer
loading for the Three-Mode as compared with the Key Mode at the low speed
condition, and, for vehicles with an  inertia weight greater than 4500 Ib,  at
the  high speed setting also.
Vehicle
Weight
Class, Ib
Up to
2500
2501 to
3500
3501 to
4500
Above
4500
Transmi s si on
In lower gear
for 30-mph
test (3rd gear)
Drive or
high gear
Drive or
high gear
Drive or
high gear
High Speed
Mode
Speed,
mph
50
50

50

50

Load,
hp
21
26

31

36

Low Speed
Mode
Speed,
mph
30
30

30

30

Load,
hp
9
12

15

18

Idle Mode
Automatic
transmission
in neutral






3.1.2.3

neutral.
3.1.2.4
Unloaded 2500 rpm
This is a high-speed test:  2500 rpm, transmission in

Federal Short Cycle
              The Federal Short Cycle was derived from the FTP.  Accel-
erations and decelerations are representative of those encountered in the
FTP, and average speed is nearly the same as the three-bag FTP driving
cycle (21. 70 mph and 21. 27 mph,  respectively).
                                   3-3

-------
                                          Time in Mode, sec
              This is a nine-mode,  125-sec CVS test that follows the
driving schedule shown below and plotted in the top half of Figure 3-1.

                     Mode
              0-16 mph acceleration
             16 - 29 mph acceleration
                 29 mph cruise
             29-37 mph acceleration
             37 - 42 mph acceleration
             42 - 37 mph deceleration
             37 - 20 mph deceleration
             20-0  mph deceleration
             Idle
The test does not include engine startup or shutdown.  The dynamometer
loadings follow the procedure as required for the FTP.
3.1.2.5       Composite NY/NJ
              This is a six-mode,  75-sec CVS test that follows the driving
cycle shown below and plotted in the lower half of Figure 3-1.

                     Mode                Time in Mode,  sec
            Idle                                  22
             0-30 mph acceleration              15
                 30 mph cruise                    15
            30 - 10 mph deceleration              12
                 10 mph cruise                     7
            10-0  mph deceleration               4
                                                  75
                                   3-4

-------
                        FEDERAL SHORT CYCLE
                               60
                             TIME, sec
                  120
                  COMPOSITE NY/NJ SHORT CYCLE
CO
  30
  20
   10
             20        40        60
                   TIME, sec
80
  Figure 3-1.  Federal Short Cycle and Composite NY/NJ
               Short Cycle Test Driving Schedules
                           3-5

-------
The test does not include engine startup or shutdown.  All vehicles are
tested at an inertia weight of 3000 Ib and a dynamometer loading of 3. 5 hp
at 30  mph.
3.1.3        Short Test Sequence
              A short test sequence consists of the following tests and
soak periods in  the order shown.
         a.    Completion of 1975 FTP
         b.    Soak - 6 minutes
         c.    Clayton Key Mode
         d.    Soak - 6 minutes
         e.    Federal Three-Mode
         f.    High-speed Unloaded 2500 rpm test
         g.    Soak - 6 minutes
         h.    Federal Short Cycle
         i.    Soak - 6 minutes
         j.    Composite NY/NJ
The 6-minute soak procedure is performed as follows:  after completion of
the preceding test, the  vehicle engine  is stopped, the vehicle hood is closed
if it was open, and the auxiliary air cooling fan is turned on if it was not
previously in use.  The fan remains in operation for 3  minutes. At the end
of 3 minutes, the auxiliary air fan is turned off and the vehicle's engine is
started.  The engine is allowed to idle in neutral for 3  minutes. Upon the
completion of this 3-minute idle period, the next test in the sequence is
initiated.
              During the entire ST, the vehicle hood is closed and the
auxiliary cooling fan is not in operation.
              In the modal tests, the  car is to  be operated in each mode
until the emissions stabilize.   In the CVS tests, driving trace procedures
and tolerance (and transmission shift  points, if applicable) are the same
as for the FTP.
                                   3-6

-------
3.2           TEST FLEETS
3.2.1         Catalyst-Equipped Experimental Vehicle
              Fleet (CEV)
3.2.1.1       Type of Car
              These 40 vehicles were all 1973 Ford Galaxies,  owned by the
Ford Motor Company.  They were equipped by Ford with an oxidizing
catalyst for control of HC and CO.   The emission control system also
included air injection and exhaust gas recirculation (EGR).  All had
8-cylinder, 400 CID engines with two-barrel carburetors  and automatic
transmissions.  Manufacturer's specifications for ignition timing and dwell
were 12° BTDC and 24° to 30°, respectively.  Axle ratio was 3.0, and tire
size was HR 78-15. The  FTP inertia weight at which the vehicles were
tested was 5000 Ib.
3.2.1.2       Prior Use
              At the time of receipt of these vehicles by the testing labora-
tory (Olson Laboratories, Inc., Anaheim, California), the odometer  readings
ranged from 7000 to 36,000 miles, with an average of 21,000.  Prior to these
tests, the cars were primarily used by Federal and California  state
employees in a number of locations  throughout California.  The driving
pattern was highly variable,  ranging from primarily stop-and-go city traffic
to primarily high-speed highway driving.  Vehicle and emission system
maintenance was performed essentially in accordance with Ford Motor
Company recommended procedure.  Emission system maintenance on some
vehicles was performed by Ford Motor Company, while for others it  was
performed by local Ford dealers or motor pool personnel,  following pro-
cedures established and monitored by Ford Motor Company.
3.2.1.3       Test Conditions
              Upon receipt of a vehicle,  the as-received fuel was drained
and test fuel (indolene  clear) was added.  The car was operated for
                                    3-7

-------
approximately 10 minutes, at which time a vehicle inspection was
performed.  This consisted primarily of measuring engine tune,  idle CO
and HC,  inspecting fluid levels, and verifying the existence and operation
of emission control devices.
              For  the first 20 vehicles of this fleet, a short test sequence
(as defined in Sec.  3. 1.3) was performed right after this inspection,  after
which the vehicle was placed in cold soak for the first FTP.  Immediately
following the FTP,  another short test sequence was performed, and the
vehicle placed in cold soak for the second FTP.  A third short  test sequence
was performed immediately upon completion of the second FTP.  For the
latter 20 cars of this fleet,  the first short test sequence (after  the vehicle
inspection) was  deleted.
              All test conditions, instrumentation, and procedures for the
1975 FTP were  as prescribed in the Federal Register, with one addition.  The
concentrations of HC, CO, NO  , and CO- in the undiluted exhaust  were also
                            X        b
continuously measured and recorded during each FTP.  The  sampling train
and analyzers used for this were the same as those used for the volumetric
short test cycles.   These continuous trace data were used to gain insight
into the emission generation characteristics of various portions of  the FTP.
              The  first group of 20 cars was tested during the  period
8 September to 25 October 1974. A group of 10 cars was  tested between
11 and 16 December 1974.   A final group of  10 cars was tested between
22 January and 19 February  1975.
3.2.2         Defect Test Fleet
              Upon completion of the CEV fleet vehicle tests described
above, 95 defect tests were performed on 5 of the 40 vehicles of the CEV
fleet.  These simulated a wide variety of malfunctions that could occur in
a typical passenger car:  defective ignition components; changes to ignition
timing, dwell, and spark advance; faulty carburetion; defective valves;
clogged air filters; and faulty emission control components.  A detailed
listing of all defects is given in the Appendix.
                                   3-8

-------
              For each defect, one FTP was performed, followed by a
short test sequence.  For each FTP, additional continuous trace recordings
were made of the concentration of HC, CO, NO  , and CO_ in the undiluted
                                            X        Ct
vehicle exhaust, as was described for the normal vehicle tests.  In 20 of the
tests,  catalyst bed temperature and exhaust flow rate were measured and
recorded for the duration of the FTP and each ST.
3.2.3         In-Use 1974 Model Year Vehicle  Fleet
3.2.3.1       Types of Cars
              This fleet comprised in-use 1974 model year vehicles. There
were 49 Ford Pintos,  49 Chevrolets (Caprice and Impala), and 49 Dodge/
Plymouths (Coronet,  Charger, Satellite).  The  Pintos were 140 CID, tested
at 2750-lb inertia weight, the Dodge/Plymouths were 318 CID, tested at
4000-lb inertia weight, and the Chevrolets were 400 CID, tested at 5500-lb
inertia weight.
              All cars had automatic transmission.  The emission control
systems were EGR plus air injection for the Chevrolets and Pintos, and
EGR for the Plymouth/Dodges.
              The rear axle ratio was 2. 73 for the Chevrolets and 3.40 for
the Pintos.  The Plymouth/Dodges had a ratio of 2.94.
3.2.3.2       Prior  Use
              These vehicles were all privately owned, and were from the
greater Detroit metropolitan area.  The as-received odometer readings
ranged from 3000  to 20,000 miles, with  an average of 11,000.  There was
no significant difference in the odometer readings between any of the manu-
facturers' subgroups of 49 cars.  No information is available concerning
the detailed driving pattern or maintenance history for any of the cars.
3.2.3.3       Test Conditions
              Testing was performed by the EPA Emissions  Laboratory
at Ann Arbor.  Each  car was tested once by the 1975 FTP, immediately after
                                   3-9

-------
which a short test sequence (as defined in Sec.  3.1.3,  with one exception)
was performed.  The one exception pertains to the Key Mode test.  For the
1974 model year fleet, the Key Mode tests were run at a fixed set of speeds.
These  speeds were 48 to 50 mph for the high speed mode, and 32 to 34 mph
for the low speed mode,  regardless of the test vehicle  inertia weight.
Thus,  the Pintos  were the only vehicles affected, as all other vehicles fall
in the same weight class for the Key Mode test. The Key Mode tests for
both the 1974 model year fleet and the CEV fleet were thus all run at the
same sets of speeds.   The dynamometer inertia and horsepower settings
were made in accordance with the test vehicle inertia weight,  per the
standard Key Mode format.
              Twenty-five of the cars that failed the FTP were tuned by
EPA and retested by the same procedure described above.  This tuneup was
parametric in that adjustments were made as required in an effort to bring
ignition timing, dwell, etc., within manufacturer's specifications, but no
new components were installed, regardless of the condition  of the  existing
ones.
                                   3-10

-------
4.  CATALYST-EQUIPPED EXPERIMENTAL
            VEHICLE FLEET

-------
      4.  CATALYST-EQUIPPED EXPERIMENTAL VEHICLE FLEET
              This section summarizes the results of statistical analyses
conducted to determine the degree of correlation existing between the various
short tests (STs) and FTP tests conducted on the catalyst-equipped experi-
mental vehicle (CEV) fleet.  Preliminary analyses  are discussed in Sec.  4. 1;
the principal statistical analysis techniques and results are summarized in
Sec.  4. 2.
4. 1           PRELIMINARY STATISTICAL ANALYSES
              Preliminary analyses were made to  assess  data quality and
statistical structure.  Of specific concern were the following goals:
         a.    Determine the data acceptable for further processing.
         b.    Determine the variation within each  test procedure.
         c.    Determine the vehicle-to-vehicle variation.
         d.    Determine the intrinsic variables and statistical structure
              of each of the tests.
         e.    Determine the distribution properties of the test data.
              Goals a and e were met by simple data screening techniques.
A multivariable  analysis of variance was used to meet goals b  and c, while d
was addressed by a canonical correlation analysis.  These techniques/analyses
are briefly discussed below.
4. 1. 1         Data Screening
              All of the basic test data for the CEV fleet were  received by
The Aerospace Corporation for processing.  A screening procedure was
developed to evaluate these data,  and to provide an annotated data base for
subsequent statistical analyses.  All inputs to the statistical data base were
derived directly from the test data traces.   All  apparent anomalies  and/or
discrepancies in the data were examined and an effort made to reconcile
them.  Discussions were held with the testing laboratory and with EPA/ECTD,
                                   4-1

-------
as appropriate, to resolve these situations.  In some cases, certain tests,
or portions thereof,  were deleted from the data base.  Of the 40 cars of the
CEV fleet, the final  data base contained 26 cars with two valid FTP tests,
and 14 cars with one valid FTP  test.  A few  short test (ST) results were
deleted,  as were various  isolated values for a given pollutant for a specific
mode.
              After the test data were  put on tape, various descriptive
characteristics of the data were used to detect gross errors in the observa-
tions, in coding and  keypunching, and in including inappropriate  cases.
Generally, this was  accomplished by checking for improper symbols or
characters, such as characters were numbers should be,  for outliers  or
blunders, and for  missing observations.  Erroneous data were reconstructed
where possible; otherwise, the  case was flagged as inappropriate for pro-
cessing.   Table 4-1  summarizes the number of cases available for statistical
analysis.

           Table  4-1.  Number of Cases Available for Statistical
                       Analysis (CEV  Fleet)
                    Test
           Federal Test Procedure
           Federal Short Cycle
           NY/NJ  Composite
           Clayton Key Mode
             Laboratory instruments
             Garage instruments
           Federal Three-Mode
             Laboratory instruments
             Garage instruments
           2500 rpm Unloaded
             Laboratory instruments
             Garage instruments
No. of Cases
     40
     39
     39

     40
     40

     31
     40

     40
     40
                                   4-2

-------
              The mathematical model employed in the contingency table
analysis requires that the data follow a bivariate normal or log-normal
distribution.  This assumption was checked using a combination of histograms,
normal probability plots, and scatter plots (Ref. 4-1).   Generally,  the log of
the data appears normally distributed.
4.1.2         Multivariate Analysis of Variance
              A multivariate analysis of variance (Ref. 4-2) with estimation
of the variance components (Ref. 4-3) was made for the CEV fleet.   The pur-
pose of the analysis of variance  is the comparison of means  when the data are
grouped or classified in one or more  ways.  The CEV fleet data were grouped
according to replication.  The groups were termed first good data and second
good data,  reflecting the original testing sequence. No difference in the data
groups was  discernible.
              The purpose in estimating the variance components is the
quantification of multiple sources of variation.  The sources  of variation
identified in the CEV fleet were  fluctuations between  cars  and measurement
errors within each test.  The results of the variance components analysis
are shown in Table 4-2.  Since there  were 14 cars in the CEV fleet which had
only one valid FTP, the number of cars in the analysis of  Table 4-2 is less
than the number  previously indicated  in Table 4-1 because replicates are re-
quired to analyze variance components.
              Normalized dispersion is defined as

                   _  _  standard deviation of population (S)
                              population mean  (M)

which is a dimensionless quantity.  "D" provides  an effective measure of the
variability of the population as observed by a test. As the fluctuations be-
tween cars are legitimate, a  good indicator of test quality is the percent of
the variation due to testing (ex).  This indicator is defined  as
                               01 =i±Vx 100
                                   s^
                                    4-3

-------
  Table 4-2.  Summary of Variance Components (CEV Fleet)
Test
FTP


Federal Short Cycle


NY/NJ Composite


Key Mode














Federal Three -Mode














2500 rpm Unloaded




No. of Cars(a)
26


25


25


25L
26G













17L
Z6G













26L
Z6G



Test Mode









High speed




Low speed




Idle




High speed




Low speed




Idle









Pollutant'*'
HC
CO
N0x
HC
CO
NO
X
HC
CO
N0x
HCL
HCG
coL
COQ
N0x
HCL
HCG
coL
COG
N0x
HCL
HCG
COL
COQ
N0x
HCL
HCG
COL
coG
NO
X
HCL
HCG
coL
coG
N0x
HCL
HCG
COL
coG
N0x
HCL
HCG
coL
coc
N0x
M
0. 69
2.68
2.57
0. 59
1.06
3.70
31.7
25.4
27.0
199
38.3
0.043
0.046
831
181
38.6
0.012
0.03
1418
289
39.8
0.0075
0.024
193
195
36.6
0.045
0.044
1008
271
42.9
0.018
0.033
2406
202
38.5
0.009
0.028
90. 5
375
59.5
0.021
0.038
464
Units
gm/mi
gm/mi
gm/mi
gm/mi
gm/mi
gm/mi
ppm
ppm
ppm
ppm
ppm
%
%
ppm
ppm
ppm
%
%
ppm
ppm
ppm
%
%
ppm
ppm
ppm
%
%
ppm
ppm
ppm
%
%
ppm
ppm
ppm
%
%
ppm
ppm
ppm
%
%
ppm
D
0.88
0.5
0. 26
1. 15
0.88
0.23
1.49
1.27
0.26
0.47
0.35
0.70
0.42
0.58
0.43
0.54
0.83
0.23
0. 19
1.0 .
0.54
0.64
0.25
0.27
0.65
0.29
0. 67
0.32
0.47
1.03
0.54
0.67
0.33
0. 13
1.61
0.67
0.67
0.31
0.19
2.39
1.66
0.81
0.45
0.30
o.%
3
7
16
22
44
30
19
76
53
10
30
3
32
4
3
65
1
84
27
1
70
12
--
20
6
28
3
50
2
67
77
3
69
14
1
57
4
--
52
4
1
1
50
48
(a)
  Subscripts L and G denote laboratory and garage analyzers
                                  4-4

-------
where
         ST = standard deviation of testing errors
          S = standard deviation of population.
              Missing values for a in Table 4-2 indicate the computationally
degenerate case where ST  is computed to be larger than S .   The ot values
shown in Table 4-2 indicate that the garage analyzers  are of lower quality
(higher a) than the corresponding laboratory instruments.   The ST bag tests
have higher a than many of the volumetric test procedures using laboratory
instruments.  The high ot in the bag tests  may be due to variations within the
driving procedure rather than to instrumentation, while  the low ot associated
with volumetric tests with laboratory analyzers may be due to simplicity of
the procedure plus instrument accuracy.
4.1.3         Canonical  Correlation  Analysis
              Canonical  correlation analysis (Refs. 4-2, 4-4) examines the
relationship between two  sets  of variables.  The problem is to find a linear
combination of a set, X,  of variables  that has maximum correlation with a
linear combination of another  set,  Y,  of variables.  The resulting correlation
is called the canonical correlation coefficient,  and the linear  combinations
are termed the canonical variables.   A second pair of linear combinations
is then looked for that has a maximum correlation and is uncorrelated with
the first pair of linear combinations.   The number of pairs of linear combina-
tions  of the X and Y sets  is equal to the number of variables in the smaller
set (X or Y, whichever is smaller).   The  technique is useful in testing for
independence of two sets  of variables  and in predicting information about a
hard-to-measure set of variables from a  set that is easier  to measure.
              The  canonical correlation coefficients for each ST versus the
FTP are shown in Table 4-3 together  with original correlation coefficients.
The observations used were the first  good data set.  For the EPA Short Cycle
and the NY/NJ Composite, the canonical correlation coefficients do not differ
significantly from the correlation coefficients of the original data.  Slight
improvements can be seen in the three-mode volumetric tests.   However, the
                                    4-5

-------
Table 4-3.  Canonical Correlation Coefficients Between the FTP and ST
             for the CEV Fleet (first good data set)
Test
Federal Short Cycle



NY/NJ Composite


Key Mode
( Laboratory)









Key Mode
(Garage)




Federal Three -Mode
(Laboratory)









Federal Three -Mode
(Garage)




2500 rpm Unloaded
(Laboratory)


2500 rpm Unloaded
(Garage)

No. of Cars
39



39


40










40





31










40





40



40


Test Mode







High speed


Low speed



Idle



High speed

Low speed

Idle

High speed


Low speed



Idle



High speed

Low speed

Idle








Pollutant
HC
CO
NO
X
HC
CO
NO
HC
CO
N0x
HC
CO
NO
X
HC
CO
NO
X
HC
CO
HC
CO
HC
CO
HC
CO
N0x
HC
CO
NO
X
HC
CO
NO
X
HC
CO
HC
CO
HC
CO
HC
CO
NO
X
HC
CO
N0x
Conventional
Correlation
Coefficient
0.87
0.81
0.62

0.92
0.77
0.61
0.74
0.23
0.79
0.70
0.38
0.1 6

0.94
0.04
0.24(»)

0.73
0.37
0.73
0.21
0.88
0.52
0.77
0.16
0.83
0.74
0. 25
0.02

0.78
0.52
0.08
0.73
0.21
0.78
0.52
0.47
0.30
0.23(a>

0.50
0. I4
0. 20
Canonical
Variable
1
2
3

1
2
3
1
2
3






.

1
2
3

-

1
2
3








1
2
3



1
2
3

1
2
3
Canonical
Correlation
Coefficient
0.89
0.80
0. 61

0.92
0.82
0.61
0.96
0.86
0.65
--
.-
..

--

..

0.94
0.85
0.65
--
--
--
0.98
0.93
0.55
--
...


--



0.90
0.84
0.58
..
--
--
0.63
0.49
0.21

0.58
0.38
0. 26<»»
  'a)Not significantly different from 0 at the 95% confidence level.
                                     4-6

-------
physical interpretation of the canonical variables is illusive for these tests.
The canonical correlation coefficients for the unloaded test indicate the
inferior correlation properties of this type procedure.  With the exception of
the unloaded test, the canonical correlation coefficients are significantly
different from zero;  i.e., the tests  are correlated to some degree.
4.1.4         Summary of Preliminary Analysis Results
               The HC and CO observations are  generally more variable than
the NO  readings,  as indicated by the dispersion results for the FTP (see
       X.
Table 4-2).  The test-to-test variation (or) can be quite high and, hence,
repeatability of the test procedures can be poor.
               Canonical variables may offer some advantages in further
analysis.  However,  their interpretation is difficult, and in a first analysis
the original variables seem appropriate.
               A model of the distribution properties of the test data appears
most likely to be log-normal.  This type of model appears appropriate  for
predicting a contingency table for the total vehicle population.
                                     4-7

-------
4.2           PRINCIPAL STATISTICAL ANALYSIS TECHNIQUES
              AND RESULTS
              The statistical procedures utilized consist of two comple-
menting classes:  correlation analysis and contingency table analysis.
Correlation analysis addresses the direct relatability of the ST with the
FTP.  Correlation analysis is an important aid in identifying STs that
have acute deficiencies.  Contingency table analysis approaches the question
of relatability from the viewpoint of the possible  tradeoffs between impact
on air quality and cost to the  public (both direct and indirect).  It is an
important tool to aid in policy formulation and cost-benefit analysis.  The
following sections briefly define each such analysis technique and summa-
rize its associated results.
4.2.1         Correlation Analysis
4.2.1.1      Conventional Method
              A conventional correlation analysis includes the calculation
of the  sample correlation coefficient r, and an a-percent confidence interval
for the population correlation coefficient p, on paired observations.   Letting
(x., y.) i = 1,	, N denote the observations, r is defined by

                              N
                              S (*j - Mx)(y. -  My)
                                     S S
                                      x y

where M , S , and M   S  are the mean and standard deviation of the obser-
        xx        y   y
vations  x. and y^,  respectively.  An a-percent confidence interval is given by
(r-,  r+), where the probability that the interval covers p is a/100.  For the
95 percent interval used in this study
                                    4-8

-------
where z  = 1/2 ln(-j-^-^J, (Fisher's Z statistic, Ref. 4-4).
               The sample correlation coefficient is used as the prime quanti-
tative measure of relatability.  The closer r is to  1, the better the relation.
A lack of relationship is indicated by r = 0.  Negative r indicates an inverse
relation  between the observations,  i.e., if one observation is high, the
other is  low, and vice versa.  The confidence interval is viewed as reflecting
the sensitivity  of the calculations to the data.  The wider the  interval,  the
less predictable  is the correlation coefficient and,  hence,  the relatability.
              A scattergram is also an important device for assessing
direct relatability.  A scattergram is merely a two-dimension plot of the
data pairs (x.,  y.).  This provides for visual examination of the data, which
is crucial in any relatability study. A sample scattergram and the associ-
ated statistics  are shown in Figure 4-1.  Here HC on the Federal Short Cycle
is plotted versus the HC on the FTP for the CEV fleet.  The number of cases
(N) is 39.  The sample correlation coefficient (COR) is 0.872, while the
95 percent interval is (0.768,  0.931).  The regression line of "y" on "x"
(ST on the FTP)  is produced by drawing a straight line between the two points
marked Y on the right and left borders of the plot.  This line represents
a least squares fit of the data (as measured in the y direction).  Similar
scatter grams for each ST and each emission constituent (HC,  CO,  NO )
                                                                    X.
were examined in the course of the study.
              Since the data included replications on some cars, the data
were organized into the following structure for conventional correlation
analysis:
         a.    First Good Data
               This data set contains the observations of the first
              FTP and ST, both of which are valid.
         b.    Second Good Data
               This data set contains the second pair of FTP and
               ST observations, both of which are valid.
                                    4-9

-------
                      • . . T • •
                                                 +• ••
                                                                     • • • •
                                                                                             ••+••• «^« • JE
I
H*
o
u
C

s
c
         (gm/mi)
                3.0
                2.5  *
                2.0  *
                1.5  *
                1.0  »
                .50
    C.O
         X* • • +* • • •^••« • *• • • • ^# • • •*••••*• • • «^ • • ••^••*
            .20       .60        1.0        1.4        1.8        2.2       2.6
                 .40        .80        1.2        1.6        2.0        2.4        2.8
M*   39
COB= .872
95* IHIEBVAL:  ( 0. 76B, 0.931)
                                                                                               3.0
                                                                                                    3.2
                                                          UC FTP  (gm/mi)
                HEAD    ST.DBV.    BEGBBSSIOM LIIE     B2S.HS.
            X   .64338   .53578  X* .99967*1* .13485   .07045
            T   .50870   .46756  1= .76129*1* .01891   .05365
                           Figure 4-1.  Correlation Analysis Scatter gram; CEV Fleet;
                                         Federal Short Cycle HC vs FTP HC

-------
         c.    Average Data
              This data set contains the average of the FTP and ST
              observations on each car (for the Federal Short Cycle
              and NY/NJ Composite only).
              For each ST a correlation analysis was performed on first
good data, second good data,  and average data where appropriate.  The
following sections briefly  summarize the significant results.
4.2.1.1.1    FTP Composite Emissions vs Individual FTP Bags
              To gather insight on correlatable STs, the correlations
between the FTP composite emissions  and the individual FTP bag data were
examined.  The composite values were computed in the standard manner.
The bag data were computed in grams  of each pollutant per bag.  Addition-
ally,  the sum of the bag 2 and bag 3 constituents were computed and the
correlation coefficient with the composite data calculated.  The analysis
was conducted on  both first and second good data.
              Table  4-4 shows the FTP intra-correlations for like  con-
stituents.  Additionally, cross correlation coefficients between dissimilar
pollutants were  computed  (not shown).  NO  appeared to be uncorrelated
with HC and CO.  The cross correlation of HC and CO was typically 0.4 to
0.6.  The results shown in Table 4-4 indicate that both cold (bag 1) and hot
(bag 2,  bag 3, bag 2 + 3) test procedures have a high correlation with the
FTP composite.  Thus,  it may be possible to  develop prototype STs using
sections of the FTP-
4.2.1.1.2    ST vs  FTP Composite Emissions
              A summary of ST/FTP  correlation coefficients  is given in
Table 4-5.  For N = 40 or 39, a computed correlation coefficient greater
than 0.35 indicates that the ST and FTP pollutants are statistically corre-
lated at the 95 percent confidence level.  For  N = 25 or 26, this threshold
is approximately 0.4.
                                   4-11

-------
         Table 4-4.  FTP Composite vs Bag Correlation Summary
                     (CEV Fleet)
FTP
Bag No.
1
2
3
2+3
Good
Data
Set'*'
First
Second
First
Second
First
Second
First
Second
Composite vs Bag ...
Correlation Coefficient' '
HC
0.90
0.91
0.94
0.99
0.84
0.97
0.98
0.99
CO
0.96
0.93
0.90
0.81
0.86
0.87
0.95
0.90
N0x
0.95
0.91
0.87
0.79
0.95
0.97
0.99
0.98
          (a)
          (b)
First good data contained 40 cars
Second good data contained 26 cars

The correlations are statistically significant at the
95% confidence level.
4.2. 1.2
  Multiple Regression Analysis
              A regression analysis evaluates the relationship between a

dependent variable and one  or more independent variables.  This technique

was used to predict the FTP results from three-mode volumetric obser-

vations.  For example, the constants h_, a,, a_, and a_ in
                 HC
                    FTp
are determined so that the correlation between the predicted HC-,-,-, as
                                                              r T IT
given above and the observed HCpTp  is maximum.  The procedure is step-

wise in that an independent variable is added one at a time in order of their

largest contribution to the correlation (Ref.  4-1).  Hence the order of
                                   4-12

-------
          Table 4-5.   ST/FTP Correlation Summary (CEV Fleet)

Short Test



Federal Short Cycle


NY/NJ Composite


Key Mode
(Laboratory)




Key Mode
(Garage)




Federal Three -Mode
(Laboratory)




Federal Three -Mode
(Garage)




2500 rpm
(Laboratory)
2500 rpm
(Garage)
Good
Data
/ \
Set



39
25
39
39
25
40
40


26


40


26


31


26


40


26


40
26
40
26
"r" -ST/FTP Correlation'0'
Coefficient

HC

0.87
0.91
0.93
0.92
0.92
0.95
0.61
0.53
0.92
0.57
0.53
0.97
0.73
0.73
0.88
0.51
0.39*
0.32*
0.87
0.79
0.80
0.68
0.52
0.94
0.76
0.73
0.78
0.69
0.42
0.62
0.47
0.37*
0.50
0.36*

CO

0.81
0.42
0.83
0.77
0.71
0.68
0.26*
0.39
0. 54
0.30*
0.31*
0.40
0.37
0.21*
0.52
0.08*
0.09*
-0.03*
0.08*
0.22*
0.48
0.20*
0.27*
0.34*
0.24*
0.21*
0.52
0. 12*
0.03*
0.39*
0.30*
0.25*
0. 14*
0. 25*

NO
X
0.62
0.47
0.53
0.61
0.51
0.61
0.79
0.20*
0.27*
0.86
0 . 04*
0 . 04*






0.89
0.03*
0.13*
0.92
-0.28*
0.08*






0.23*
0.23*


(a)  First Good Data:  This data set contains the observations of the first FTP and ST,
    both of which are valid.
    Second Good Data: This data set contains the second pair of FTP and ST obser-
    vations, both of which are valid.
    Average Data: This data set contains the average of the FTP and ST observations
    on each car (for  the Federal Short Cycle and NY/NJ Composite only).
(b)  Number of cars in data set
(c)  The correlation is statistically significant at the 95% confidence level except when
    indicated by an asterisk.
                                         4-13

-------
inclusion indicates the mode's relative importance.   The ordering of the
modes varies depending on the ST and pollutant under study.
              A multiple regression  analysis was performed for the three-
mode volumetric tests on first good data.  The purpose of this analysis was
to empirically determine the linear combinations of the three-mode  readings
that have maximum correlation with the FTP.  The linear combinations are
composed of like constituents.  Thus, each linear combination can be con-
sidered as a weighted observation on  HC, CO,  and NOx-  The results are
shown in Table 4-6, along with the maximum correlation coefficient using
only a single reading on each constitutent. As can be seen from Table 4-6,
the weighted combination correlation  coefficients are not significantly
higher than the correlation coefficient of the best single reading.

         Table 4-6.  ST/FTP Correlations for Weighted Mode Tests
                     (CEV  Fleet) (first good data only)
Short Test
Key Mode
Laboratory
Garage
Federal Three-
Mode
Laboratory
Garage
N(a>

40
40


31
40
Weighted Corre-^b)
lation Coefficient
HC

0.93
0.91


0.91
0.81
CO

0.55
0.58


0.48
0.53
NO
X

0.83



0.90

tc\
Best Single -Mode* ' .
Correlation Coefficient* '
HC

0.92 (I)
0.88 (I)


0.87 (H)
0.78 (I)
CO

0.54 (I)
0.52 (I)


0.48 (I)
0.52 (I)
NO
X

0.79 (H)



0.89 (H)

 (a)
 (b)
 (c)
Number of cars in data set
Correlations are statistically significant at the 95% confidence level
H - high speed mode
I  = idle mode
                                    4-14

-------
4.2. 1.3
Correlation Sensitivity Analysis
              As previously mentioned, the sensitivity of the analysis to the
data used can be assessed by using the confidence interval.   To dramatize
this  sensitivity, a worst case approach was examined by deleting selected
extreme data points from the existing data.  Recalculation of the correlation
coefficient was performed to illustrate the variability due to the sample.
This was done in a sequential manner for the Federal Short Cycle.  A sum-
mary of the analysis results for the Federal Short Cycle-FTP  correlations
is shown in Table 4-7.  A review of Table 4-7 values indicates that the re-
sults are extremely sensitive to a small percentage of the data points.
4.2.1.4      Discussion of Selected Correlation Analysis Results
4.2.1.4.1
Shortcomings of the Correlation Coefficient
               The main usage of the correlation coefficient is as an indicator
of direct relatability between ST and FTP.  In this respect it has a number of
deficiencies.   The computed correlation coefficient is sensitive to the location

      Table 4-7.  Correlation Coefficients for Selected Car Deletions;
                  Federal Short Cycle vs FTP (CEV Fleet)
Number of Cars Deleted
0
1
2
3
4
(a)
Correlation Coefficient
HC
0.872
0.657
0.656
--
--
CO
0.810
0.673
0.639
--
--
NO
X
0.621
0.690
0.633
0.823
0.755
          ^Significant at the 95% confidence level
                                    4-15

-------
of a small percentage of the data, as shown in Table 4-7.  It is a summary
statistic in that all the information contained in the data is compressed into
a single number (this is alleviated to some degree by examination of the
scattergrams).
              It is difficult to infer air quality impact from correlation
statistics except in the broadest sense,  and a tradeoff analysis is virtually
impossible based solely on correlation coefficients.
4.2.1.4.2    Mode Tests vs Bag Tests
              On the basis of HC and CO correlation, the bag tests (Federal
Short Cycle and NY/NJ Composite) are  preferable to the mode- or volumetric-
type ST.  The volumetric STs,  in general, show deficiencies in tracking  CO.
The high-speed modes, however, have superior NO  correlation.
                                                  JL
4.2.1.4.3    Laboratory Analyzers vs Garage Instruments
              The largest difference between the correlation results  of the
two measurement techniques occurs on  the second good data sets. There
is a greater variation in the correlation estimates of first good data and
second good data for the garage analyzer than for the laboratory analyzer,
as shown in Table 4-5.  This is most likely due to the combination of  low
CO values for the  CEV fleet,  small sample  size,  and less accurate
instrumentation.
              The most striking difference between laboratory and garage
data is for HC on the  Federal Three-Mode.   The laboratory measurements
indicate the best mode to be high speed, while the  garage readings indicate
the idle mode as superior.  This is inconsistent with the results for HC on
the Clayton Key Mode,  and may be attributed to the difference in the  sample
sizes of the Federal Three-Mode and the Clayton Key Mode tests.
              CO correlation deficiency is  common to both measurement
techniques.   Due to the low concentration of CO being emitted, this may  be
a measurements problem, in general, rather than a deficiency in ST
structure.
                                   4-16

-------
4.2.1.4.4     ST  Correlation Ratings
              The following qualitative rating scale was used to rate the ST:
               Rating                         Description
          (U)  Unacceptable        Constituent is uncorrelated at 95%
                                  confidence level
          (P)  Poor               Constituent is correlated at the 95%
                                  confidence level, but with correla-
                                  tion less than 0.6
          (F)  Fair               Correlation between 0. 6 and 0. 7
          (G)  Good               Correlation between 0.7 and 0. 9
          (E)  Excellent           Correlation between 0. 9 and 1. 0

              For rating the three-mode volumetric ST,  the mode with the
highest rating was used.  Table 4-8 shows the ratings of the  ST  on  each
pollutant on this basis.
              In general,  the STs have less difficulty tracking HC  than CO
and NO .  Excluding the Unloaded 2500 rpm ST (which has either "P11 or "U"
       JL
ratings for all three pollutants),  the bag-type and modal STs all have "G" to
"E" ratings for HC.  In the case of CO, the bag-type STs have "G" ratings,
whereas the modal STs are rated in the "P" category.  This  situation is re-
versed in the case of NO , where the modal STs have "G" ratings and the
                       x
bag-type STs are  rated "F" to "P". Hence, the choices among the STs for
CO and NO  implementation may be more limited than for HC.
4.2.2         Contingency Table Analysis
              The contingency table analysis is used to establish the ST
pass-fail levels for each pollutant.  The contingency table is defined in
Table 4-9,  along with its associated parameters.  A pictorial demonstration
of its  application to a given data  set is  shown in Figure 4-2.   This  figure
                                   4-17

-------
shows that,  for a given data set, ST cut-points must be established in order
to compute the elements of the contingency table.  Four basic approaches
for cut-point determination were considered, which are described as follows.
                    Table 4-8. ST Correlation Ratings
Short Test
Federal Short Cycle
NY/NJ Composite
Key Mode
Laboratory
Garage
Federal Three -Mode
Laboratory
Garage
2500 rpm Unloaded
Laboratory
Garage
Rating
HC
G
E

E (I)(a)
G(I)

G(H)
G(I)

P
P
CO
G
G

P(D
P(D

P(D
P(D

U
U
N0x
F
P

G(H)


G(H)


U

             I = idle mode, H = high speed mode
                                   4-18

-------
       Table 4-9.  Contingency Table

Predicted =
Short Test
Pass
Fail
Total
True = FTP
Pass
a
c
a + c
Fail
b
d
b + d
Total
a + b
c + d
n = a + b
+ c + d
 a = number of correctly passed vehicles (PP)

 b = number of error of omission (E )

 c = number of error of commission (E  )

 d = number of correctly failed vehicles (FF)

Sensitivity = a /(a + c)

Specificity = b/(b + d)

False positive error = b/(a + b)

False negative error = c/(c + d)
   4
Correlation index =
ad - be
                    [(a + b)(a + c)(b + d)(c + d)]
                                              1/2
                     4-19

-------
                                          FF,  CORRECTLY
                                          FAILED VEHICLES
                    Er, ERROR OF
                     c
                    COMMISSION

    ST CUT-POINT
•  •
           •
         i
      •
                        •             .
                          •           i
                               •      i   •
                                            EQ, ERROR OF
^                                    ,      OMISSION
13

<
                     •
         PP, CORRECTLY
         PASSED VEHICLES
                                    ai
                                    ti
                   FTP MEASUREMENT

        Figure 4-2. Contingency Table Representation
                          4-20

-------
4.2.2.1       Analysis Methods Examined
4.2.2.1.1     Maximum Correlation Method
              In this method,  the ST cut-point is selected so that the corre-
lation index (as defined in  Table 4-9) is maximized.  This is an impartial
procedure for finding the STs that give the best correlation .with the FTP
under the terms of the contingency table.  Figure 4-3 graphically illustrates
the procedure.  This method provides for no policy decision.
4.2.2.1.2     Bounded Errors of Commission Method
              The ST cut-points are selected to  minimize the errors of
omission while holding the errors of commission below a specified level.
This method permits a direct answer to the question, "For a given permis-
sible level of errors of commission, what level of errors of omission must
be accepted, and with what impact on air quality  (inferred from the number
of FF vehicles) ?"  This method is pictorially demonstrated in Figure 4-4.
The policy decision is the  maximum allowable errors of commission.
4.2.2.1.3     Weighted Errors Method
              The strategy used in this method is, as indicated by
Figure 4-5,  to minimize a linear combination of  the errors of commission
and the  errors of omission. The linear combination  represents cost to the
public,  where the weights  indicate the relative importance of the two types
of costs: those incurred by manufacturers, versus those due to deterioration
of air quality.  Air quality impact is inferred from the level of FF vehicles.
The policy decision is the  cost structure; that is, the specification of the
weights.
4.2.2.1.4     Percent Rejection Method
              The ST cut-points are determined so that a specified per-
centage of the population is failed by the ST.  This is shown graphically in
Figure 4-6.  The policy decision is the percentage to be rejected by the ST.
                                    4-21

-------
      PARAMETRIC
       TECHNIQUE

USE LINEAR REGRESSION
AS A MODEL
X2 -
          +  B
                                    DATA ANALYTIC
                                      TECHNIQUE

                               SELECT SHORT TEST LEVEL
                               SUCH THAT THE TABLE
                               CORRELATION IS  MAXIMUM
SOLVE FOR C2 WHEN

I.E..
                \s>
                     ST      .  [ :S:
                  CUT-POINT -
                     FTP MEASUREMENT
   Figure 4-3.   Maximum Correlation Method
               MINIMIZE EQ SUBJECT TO EC < Y%
         PARAMETRIC TECHNIQUE ONLY.  USE BIVARIATE
         NORMAL PROBABILITY MODEL
                   NOT TO   |
                  EXCEED Y%  l
                   ST
              CULJLOJJiT*r  •J. "	
                        T    1
                        •  • T"1
                                   MINIMIZE
                           Q.
                           t
                   FTP MEASUREMENT
        Figure 4-4.   Bounded Errors of
                       Commission Method
                         4-22

-------
                  MINIMIZE aEr + 0E
                             C    Q
    PARAMETRIC
     TECHNIQUE
USE BIVARIATE NORMAL
PROBABILITY MODEL
                 WEIGHT a
                  ST
       DATA ANALYTIC
         TECHNIQUE
SELECT THE SHORT TEST LEVEL
SO THAT aEc + #EO IS  MINIMUM
                              •  •
                                WEIGHT/?
                           i a.
                           IE
                  FTP MEASUREMENT'
        Figure 4-5.  Weighted Errors Method
                  REJECT a* OF THE SAMPlf
      PARAMETRIC
       TECHNIQUE
  USE BIVARIATE NORMAL
  PROBABILITY MODEL
         DATA ANALYTIC
           TECHNIQUE
 SELECT THE SHORT TEST LEVEL UNTIL
 SPECIFIED PERCENT IS REJECTED
                  ST
                                   SHORT TEST
                                   REJECTION
                                   REGION
                        l!
                 FTP MEASUREMENT——

        Figure 4-6.  Percent Rejection Method
                           4-23

-------
4.2.2.2       Procedural Techniques Utilized
               The techniques used to compute the ST cut-points and the
contingency table entries are classified as data analytic and parametric.
The data analytic technique uses the data directly without resort to a model.
The parametric procedure uses a model of the data.
4.2.2.2.1     Data Analytic Technique
               The cut-point for each pollutant is determined individually.
For each ST cut-point, the table entries are calculated by counting the
number of data points in each of the  appropriate regions indicated in
Figure 4-2.  Each ST cut-point is  then iteratively varied until the objective
of the particular strategy is  achieved.  This set of ST cut-points is then
taken as the  solution to the strategy  under study.
               This procedure was not applied to the method of bounded
errors of commission.  The bound typically ranged from 5 percent to 0.1
percent.  In  terms of actual  counts,  this range is 2. 0 to 0.04 cars for the
CEV fleet.  The  results would thus be sensitive to a very small portion of
the data.
4.2.2.2.2     Parametric Technique
               The data are first modeled by using a bivariate normal
distribution as shown in Figure 4-7.   Thus the correlation coefficient, mean
values, and standard deviations are  computed from the data and substituted
into the model.   The ST cut-points are then determined by using the model.
Figure 4-8 indicates the pertinent probability calculations for predicting the
table entries.  The predicted table entries are shown in Figure 4-9.
Figure 4-10  shows the equations to be solved to determine the ST cut-points.
              After the ST cut-points have been determined, the  contingency
table results are calculated using both the actual data points and the model
of the data.
                                   4-24

-------
BIVARIATE NORMAL DISTRIBUTION
         e  U(l -/><
WHERE   nvfji2   *   POPULATION MEANS
        0-^0-2   •   POPULATION STANDARD DEVIATIONS
        P      •   CORRELATION COEFFICIENT
        Xj      -   FTP MEASUREMENT
        X2      -   ST MEASUREMENT

                 Figure 4-7. Parametric Model

              PROBABILITY OF ERROR OF COMMISSION
X2)
              Pr{ XjC2} • j j  D (Xr
                                 -C2}-// D(X1.X2)«1d!

              WHERE   Cj • CRITICAL FTP LEVEL
                      C2 - CRITICAL ST LEVEL

                Figure  4-8.  Probability Equations

                               4-25

-------
EXPECTED ERRORS OF COMMISSION AND OMISSION
    EC - N x [ PROBABILITY OF ERROR OF COMMISION ]
    E  • N x [ PROBABILITY OF ERROR OF OMISSION ]
     o
EXPECTED CORRECT FAILURES
    FF • N x [ PROBABILITY OF CORRECT FAILURE ]
    PP  = N - FF - E  - Ert
                 c    o
WHERE   N - NUMBER OF CARS USED

          Figure 4-9.  Expected Values
                   SOLVE FOR C2
  PERCENT REJECTION
                       +00 +00
      0=Pr^X>Cl=l /
      u  rr \*2>^2>   J J
                     -C0c2
  WEIGHTED ERRORS
      MIN (QE. + J8EJ; a + £ • 1
       /•»     co
  BOUNDED ERRORS OF COMMISSION:
      MIN(E); E 
-------
4.2.2.3
Selected Analysis Methods
              Only two of the above four cut-point-level selection strategies
were investigated in any detail:  the maximum correlation method and the
bounded errors of commission method. The maximum correlation method
was chosen for comparison with the previous correlation analysis.  The infor
mation contained in an analysis under the other two strategies is identical
for varying policy decisions. That is,  as the policy is varied under each
strategy, the resulting loci of EC,  EQ> and FF are identical.  Hence, the
bounded errors of commission method was chosen for its particular rele-
vance to the cost to manufacturers and air quality impact.
              As the emissions standard to which the CEV fleet was de-
signed is uncertain,  four sets of FTP cut-points were used in the  analyses.
These are  specified  in Table 4-10.
              Table 4-10.  Assumed FTP Levels (CEV Fleet)
Level
I
II
HI
IV
Emission Level, gm/mi
HC
0.41
0.60
0.75
0.90
CO
3.4
5.0
7.0
9.0
NO
X
3.1
3.1
3.1
3.1
                                   4-27

-------
4.2.2.4       Maximum Correlation Analysis Results
               The problem of presenting the results can best be seen while
observing relationships of EC, EQ,  and FF to changing FTP level.  For
example, Figure 4-11 illustrates a typical plot for HC, using the data
analytic calculation technique.  Similarly, Figure 4-12 shows the same
results using the parametric calculational technique on the actual data points
only, while Figure 4-13 shows the results as predicted from a model of the
data.  Trends are clearly more visible in the predicted population results.
Although these trends are an intrinsic  component of the model,  the actual
magnitudes and rates of change of the trends are due to the data.
               A summary of the  results of the maximum correlation
analysis for the predicted population of the CEV fleet is shown in Tables 4-11
through 4-14.  For  N = 40 or 39,  a. computed table correlation coefficient
greater in magnitude than 0.31 indicates that the ST and FTP pollutants are
statistically correlated at the 95 percent confidence level.  For N = 31,
this threshold is 0.35.   Figures 4-13 to 4-24 depict the relationship of
EC,  EQ> and FF to changing FTP level for HC and  CO on the predicted
population basis.  Figure 4-25 shows the  variation of E and E  for NO .
r                                                     ocx
               The correlation index of the contingency table, as defined in
Table 4-9,  is substantially different than  the computed correlation  coeffi-
cients of Sec. 4.2.1.  Although the relative ranking of the ST may be similar
to that of Table 4-8, experience has shown that contingency table correlation
index is an  unreliable indicator of relatability.  For example,  consider the
extreme case where EQ = 0.0, EC = 0.0,  and FF = 0.01%.  In this case the
correlation index will be 1.0; however, 99. 99% of the data are in the
correctly passed group,  and the correlation index tells nothing about 99.99%
of the data.  This example  also indicates  that the correlation index is a
function of the ST and FTP cut-points.  Although this is desirable for policy
analysis, tradeoffs are best inferred by directly observing the pertinent
quantities.
                                   4-28

-------
                                         O  FSC
                                         A  NY/NJ
                                         D  2500  rpm
                                             UNLOADED
                                         	FF
   0.4      0.5       0.6      0.7      0.8     0.9
                    1975 FTP HC LEVEL, gm/mi
1.0
Figure 4-11.  Variation of Ec, Eo,  and FF with HC
 FTP Level; Maximum Correlation Method; Data
 Analytic Technique;  CEV Fleet
                         4-29

-------
o
   50
   45
   40
   35
    30
   25
   20
   15-
   10
O  FSC
A  NY/NJ
O  2500 rpm UNLOADED


       Ec
	EQ
	FF
         >a
                     I
  I
.4      0.5      0.6      0.7      0.8
               1975 FTP HC LEVEL, gm/mi
                                            0.9
                 1.0
  Figure 4-12.  Variation of Ec, Eo, and FF with HC
   FTP Level; Maximum Correlation Method;  Para-
   metric Technique; CEV Fleet
                          4-30

-------
                                  O  FSC
                                  A  NY/NJ
                                  D  2500 rpm UNLOADED
                                  	FF
    0.4
0.5      0.6     0.7      0.8
       1975 FTP HC LEVEL, gm/mi
Figure 4-13.  Variation of Ec, EQ>  and FF with HC
  FTP Level; Maximum Correlation Method; Pre-
  dicted Population Technique; CEV Fleet
                       4-31

-------
  Table 4-11.   Maximum Correlation Summary,
                   FTP Level I (CEV Fleet)

Short Test
Federal
Short Cycle

NY/NJ
Composite


Key Mode
(Laboratory)








Key Mode
(Garage)





Federal
Three-Mode
(Laboratory)







Federal
Three-Mode
(Garage)




2500 rpm Unloaded


2500 rpm Unloaded
(Garage)


N
39


39



40









40






31









40






50


40



Test
Mode







High



Low


Idle


High


Low

Idle

High



Low


Idle


High


Low

Idle








Pollutant
HC
CO
NOX
HC
CO
NO
X
HC

CO
N0x
HC
CO
NOX
HC
CO
NOX
HC

CO
HC
CO
HC
CO
HC

CO
N0x
HC
CO
NOX
HC
CO
NOX
HC

CO
HC
CO
HC
CO
HC
CO
NOX
HC

CO

*Ec
6.49
10.8
16.3
5.38
12.0
16.7

10.6

23.1
10.0
11.5
20.4
33.6
5.21
17.2
30.4
8.95

20.9
9.05
24.2
6.30
17.8
6.29

23.9
5.49
7.70
21.7
42.9
7.64
17. 1
37.8
8.47

23.7
8.99
24.1
8. 17
17.9
12.2
22.4
32.4
11.8

25.6

*Eo
8.53
8.23
5.26
6.72
8.94
5.32

17.2

18.5
4.40
19.4
16.6
7.13
6.51
14.4
6.9
13.4

16.8
13.6
18.9
8.27
14.6
8.60

23.5
2.72
11.39
21.3
5.78
11.3
16.9
5.55
12.4

18.6
13.4
18.9
II. 7
14.7
21.1
17.7
6.67
20.2

19.9

%FF
58.3
28.0
9.71
60.1
27.3
9.65

50.1

25.3
11.2
47.8
27.2
8.43
60.8
29.4
8.69
53.9

26.6
53.6
24.4
58.9
28.7
60.3

26.0
8.96
57.5
28.2
5.90
57.6
32.7
6.14
54.9

24.7
53.7
24.4
55.5
28.7
46.1
25.6
8.09
46.9

23.5
Table
Correlation
Index<">
0.667
0.596
0.372
0.730
0.556
0.363

0.405
^
0. 166
0.532
0.344
0.257'
0.105°
0.737
0.363
0.147
0.514

0.242'
0.507
0.136
0.675
0.346
0.661

0.050"
0.644
0.571
o. MO'
0.012'
0.575
0.321
0.063
0.544

0. 151
0.511
0. 137 ""
0.563
0.344
0.300
0.194*
0.121°
0.323

0.089°
(a)
  The correlation ii statistically significant at the 95% confidence level except where Indicated
  by an asterisk
                            4-32

-------
        Table 4-12.   Maximum Correlation Summary,
                        FTP Level II (CEV Fleet)
Short Test
Federal
Short Cycle

NY/NJ
Composite

Key Mode
(Laboratory)




Key Mode
(Garage)




Federal
Three-Mode
(Laboratory)




Federal
Three-Mode
(Garage)




2500 rpm Unloaded
(Laboratory)
2500 rpm Unloaded
(Garage)
N
39

39

40




40




31





40




40
40
Test
Mode




High
Low

Idle

High
Low

Idle

High

Low

Idle

High
Low

Idle



Pollutant
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
%EC
7.90
7.10
6.42
8.38
14.0
30.8
15.5
25.4
6.28
19.6
11.4
26.2
11.6
32.7
7.69
20.5
7.73
35.6
9.68
30.8
9.60
21.8
10.7
31.7
11.5
32.7
10.3
20. S
16.3
29.1
15.8
35.7
%Eo
8.31
2.36
6.70
2.50
15.2
7.58
16.8
7.03
6.50
6.35
12.3
7.01
12.5
7.64
8. 10
6.37
8.56
11.3
10.99
10.5
10.9
8.89
11.5
7.55
12.4
7.63
11.1
6.38
18.2
7.30
17.5
7.90
% FF
44.9
5.63.
46.5
5.48
37.7
9.51
36.0
10.1
46.3
10.7
41.1
9.80
40.9
9.18
45.4
10.4
47.8
12.2
45.4
13.0
45.4
14.6
41.9
9.27
41.1
9.18
42.4
10.4
35.3
9.52
35.9
8.91
Table
Correlation
Index(a)
0.675
0.510
0.737
0.467
0.415
0.142*
0.354
0.223*
0.744
0.321
0.523
0.209*
0.519
0.116*
0.683
0.304*
0.670
0.46
0.582
0.127*
0.585
0.297*
0.553
0.129*
0.521
0.116*
0.572
0.302*
0.309
0.166*
0.332
0.075*
'a'The correlation is statistically significant at the 95% confidence level except where indicated
  by an asterisk
                                4-33

-------
       Table 4-13.  Maximum Correlation Summary,
                      FTP Level HI (CEV Fleet)
Short Teat
Federal
Short Cycle
NY/NJ
Composite
Key Mode
(Laboratory)




Key Mode
(Garage)




Federal
Three -Mode
(Laboratory)




Federal
Three-Mode
(Garage)




2500 rpm Unloaded
(Laboratory)
2500 rpm Unloaded
(Garage)
N
39
39
40




40




31




40




40
40
Teat
Mode


High
Low

Idle

High
Low

Idle

High
Low

Idle

High
Low

Idle



Pollutant
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
%EC
8.49
1.16
6.79
1.57
16.2
29.4
18.1
20.7
6.61
6.98
12.8
21.9
13.0
32.5
8.26
14.1
8.46
42.8
10.8
33.4
10.7
18.7
11.9
30.8
12.9
32.4
11.4
14.2
19.2
26.4
18.5
37.9
*Eo
7.50
0.11
6.14
0.12
12.7
1.20
13.9
1.13
5.91
13.1
10.7
1.12
10.9
1.19
7.33
1.05
7.93
2.59
9.97
2.47
9.89
2.19
10.1
1.18
10.8
1.19
9.71
1.05
15.1
1.15
14.6
1.22
%FF
34.6
0.21
36.0
0.21
28.4
1.40
27.2
1.47
35.12
1.54
31.5
1.42
31.3
1.35
34.9
1.49
38.0
2.74
36.0
2.86
36.1
3.14
32.1
1.36
31.4
1.35
32.5
1.49
27.1
1.39
27.6
1.32
Table
Correlation
IndexW
0.673
0.312*
0.736
0.267*
0.413
0.082*
0.352
0.134*
0.742
0.207*
0.522
0.125*
0.515
0.066*
0.682
0.193*
0.670
0.031*
0.582
0.086*
0.586
0.214*
0.552
0.074*
0.519
0.067*
0.570
0.192*
0.308
0.097*
0.330
0.043*
''The correlation ia statistically significant at the 95% confidence level except where indicated
 by an asterisk
                               4-34

-------
        Table 4-14.   Maximum  Correlation Summary,
                         FTP Level IV (CEV Fleet)
Short Teat
Federal
Short Cycle
NY/NJ
Composite
Key Mode
(Laboratory)




Key Mode
(Garage)




Federal
Three-Mode
(Laboratory




Federal
Three-Mode
(Garage)




2500 rpm Unloaded
(Laboratory)
2500 rpm Unloaded
(Garage)
N
39
39
40




40




31




40




40
40
Teat
Mode


High
Low

Idle

High
Low

Idle

High
Low

Idle

High
Low

Idle



Pollutant
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
HC
CO
%EC
8.49
0.05
6.66
0.09
17.4
22.3
19.7
12.3
6.40
5.49
13.4
13.5
13.6
26.5
8.24
6.26
8.69
41.8
11.4
29.0
11.2
11.0
12.4
24.2
13.5
26.3
11.7
6.32
21.3
18.7
20.4
33.9
%Eo
6.28
0.0
5.22
0.0
9.85
0.08
10.7
0.08
4.93
0.07
8.69
0.07
8.79
0.08
6.14
0.07
6.89
0.30
8.51
0.29
8.44
0.26
8.23
0.08
8.73
0.08
7.94
0.07
11.8
0.08
11.4
0.08
% FF
25.3
0.00
26.4
0.00
20.2
0.09
19.4
0.09
25.1
0.10
22.9
0.09
22.8
0.08
25.4
0.09
28.9
0.31
27.3
0.32
27.4
0.35
23.3
0.09
22.9
0.08
23.6
0.09
19.8
0.09
20.1
0.08
Table
Correlation
Index!*)
0.665
0.138*
0.729
0. 104*
0.401
0.030*
0.341
0.053*
0.734
0.092*
0.512
0.048*
0.505
0.024*
0.674
0.083*
0.665
0.014*
0.576
0.041*
0.580
0.113*
0.542
0.027*
0.509
0.024*
0.561
0.083*
0.299*
0.036*
0.321
0.015*
'a*The correlation i« »tatistically significant at the 95% confidence level except where indicated
  by an aateriik
                                 4-35

-------
   501~
   45
   40
   35
\
UJ
o
   25
   20
   15
   10

                         '

                    I
                                    O HIGH SPEED
                                    A LOW SPEED
                                    D IDLE
                                    	FF
\N
                                          -O
1.4      0.5      0.6     0.7      0.8
              1975 FTP HC LEVEL, gm/mi
                                          0.9
              1.0
    Figure 4-14.  Variation of Ec, Eo, and FF with HC
     FTP Level; Federal Three-Mode Test; Maximum
     Correlation Method; Predicted Population of CEV
     Fleet
                        4-36

-------
                                O  HIGH SPEED
                                A  LOW SPEED
                                D  IDLE
                                	FF
        0.5     0.6     0.7      0.8
               1975 FTP HC LEVEL, gm/mi
Figure 4-15.  Variation of Ec, E0,  and FF
 with HC FTP Level; Key Mode Test;
 Maximum Correlation Method; Predicted
 Population of CEV Fleet
                     4-37

-------
LU
O
CC
UJ
Q.
    15 —
10
     5 -
                                                           O FSC
                                                           A NY/NJ
                                                           D 2500 rpm
                                                                UNLOADED
     345         678
                            1975  FTP CO LEVEL,  gm/mi


     Figure 4-16.  Variation of Ec, Eo, and FF with CO FTP Level;
      Maximum Correlation Method; Predicted  Population of CEV
      Fleet
                                                                   10
                                  4-38

-------
451-
O  HIGH SPEED
A  LOW SPEED
D  IDLE
                    5678
                       1975 FTP CO LEVEL, gm/mi
Figure 4-17.   Variation of Ec, Eo, and FF with CO FTP Level;
 Federal Three-Mode Test; Maximum Correlation Method; Pre-
 dicted Population of CEV Fleet
                            4-39

-------
    30
    25
<
UJ
UJ
O
OS
    20
    15
    10
                                                           O HIGH
                                                           A LOW SPEED
                                                           D IDLE
                                  678
                             1975 FTP CO LEVEL, gm/mi
10
      Figure 4-18.  Variation of Ec. Eo,  and FF with CO FTP Level;
       Key Mode Test; Maximum Correlation Method;  Predicted
       Population of CEV Fleet
                                   4-40

-------
   50

    40
                                      o
                                      FF
<  30
  o
o

UJ
D_
    20
    10
     0
     0.4
0.5
 0.6        0.7        0.8

1975  FTP HC LEVEL,  gm/mi
0.9
1.0
   Figure 4-19.   Variation of Ec,  Eo> and FF with HC FTP Level;
    Unloaded 2500 rpm Test; Garage Instruments; Maximum Cor-
    relation Method; Predicted  Population of CEV Fleet
                                 4-41

-------
   60
   50
   40
o
<
   30
h«M _^
o 20
   10

                                                      O HIGH SPEED
                                                      A LOW SPEED
                                                      D IDLE

                   k
                       'Ik
                                •\
                                                      	FF
                                             I
    0.4
0.5
0.6
0.7
0.8
0.9
                       1975  FTP HC LEVEL, gm / mi
1.0
     Figure 4-20.  Variation of Ec,  Eo,  and FF with HC FTP Level;
      Federal Three-Mode Test; Garage Instruments; Maximum Cor-
      relation Method; Predicted Population of CEV Fleet
                                4-42

-------
UJ
O
a:
    60
    50
    40
£  20
    10
R



                                   XX
O  HIGH SPEED
A  LOW SPEED
D  IDLE
                                                     ---- E
                                                     ^-6

0
0.4
                0.5      0.6        0.7       0.8
                         1975 FTP  HC LEVEL,  gm / mi
                                                 0.9
              1.0
    Figure 4-21.   Variation of Ec, E0, and FF with HC FTP Level;
     Key Mode Test; Garage Instruments; Maximum Correlation
     Method; Predicted Population of CEV Fleet
                                4-43

-------
o
O£
   401-
   30
 J" 20
   10
        \
                                               	FF
                            678
                        1975 FTP CO LEVEL, gm/mi
  Figure 4-22.   Variation of Ec,  Eo, and FF with CO FTP
    Level; Unloaded 2500 rpm Test; Garage Instruments;
    Maximum Correlation Method; Predicted Population of
    CEV Fleet
   40 r-
                                                  O  HIGH SPEED
                                                  A  LOW SPEED
                                                  D  IDLE
                    5678
                       1975 FTP CO LEVEL, gm/mi
10
  Figure 4-23.   Variation of Ec,  E0, and FF with CO FTP
   Level; Federal Three-Mode Test; Garage Instruments;
   Maximum Correlation Method; Predicted Population of
   CEV Fleet
                            4-44

-------
o
O£
UJ
O.
    10-
     C -.
                                                         O  HIGH SPEED
                                                         A  LOW SPEED
                                                         D  IDLE
                        5678

                            1975 FTP CO LEVEL, gm/ml
     Figure 4-24.   Variation of Ec, Eo, and FF with CO FTP Level;
      Key Mode Test; Garage Instruments; Maximum Correlation

      Method; Predicted Population of CEV Fleet
                                  4-45

-------
    201-
    15

u,°

§  10
    °
        KEY
        HIGH
\
F3M
HIGH
	|_
                       NY/NJ
                       NY/NJ
                                2500 rpm    IDL£
                                UNLOADED      •F3IV1
                                                 LOW
0
               10
            20
    30
PERCENT
40
50
60
   Figure 4-Z5.  Variation of Eo and Ec at NOX Level of 3. 1 gm/mi;
     CEV Fleet; Maximum Correlation Method; Predicted Population
     of CEV Fleet

-------
              By examination of the values of E  , E ,  and FF in Tables 4-11
 through 4-14, the ST can be seen to follow the correlation ratings of Table 4-8.
 For example, on the Key Mode (Laboratory) for HC

High Speed:
Low Speed:
Idle:
Ec
10.6
11.5
5.21
Eo
17.2
19.4
6.51
FF
50. 1
47.8
60.8
As the idle mode has the highest percent FF and lowest percent E  and E ,
it is a superior mode for HC.  This is consistent with the ratings in Table 4-8.
               The summary tables and the summary graphs do not clearly
favor a single FTP level as a design level for the CEV fleet.  However,  the
CO plots suggest that levels II,  III,  and IV are too high for CO as the percent
FF dips below  15 percent on all tests.
               The maximum correlation method does not admit a policy
decision after the FTP level has been set.  Thus its usefulness is restricted
for purposes of tradeoff analysis.
4.2.2.5        Bounded Errors of Commission Analysis Results
4.2.2.5.1      Single-Constituent Tests
               For the CEV fleet, the bound of errors of commission was
varied from 5 percent to 1 percent in 1 percent increments, with the values
0. 5 percent and 0. 1 percent also included. An analysis was made for each
of the FTP levels  of Table 4-10.  The results of the analysis are summarized
in the following three sections.  The data plotted are for the predicted CEV
population.  Since the exact FTP value is uncertain, only general observations
can be made.
                                   4-47

-------
4.2.2.5.1.1  Hydrocarbon Emissions
              The variation of E , E , and FF as a function of HC cut-point
is displayed in Figures 4-26 through 4-35 for each ST examined, and for the
range of HC FTP values selected in Table 4-10  (HC = 0.41 to 0.90).  The
figures correspond to the following ST/FTP level spectrum:
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode
(Laboratory)
Clayton Key Mode
(Garage)
Federal Three -Mode
(Laboratory)
Federal Three -Mode
(Laboratory)
Federal Three -Mode
(Garage)
Federal Three -Mode
(Garage)
2500 rpm Unloaded
(Laboratory)
2500 rpm Unloaded
(Garage)
FTP HC Level
0.41
X
X
X
X
X

X

X
X
0.60
X
X






X

0.75
X







X

0.90
X
X
X
X

X

X
X
X
Figure No.
4-26
4-27
4-28
4-29
4-30
4-31
4-32
4-33
4-34
4-35
      The graphical displays indicate the general nature of the tradeoff avail-
able for policy formulation.  Reducing the errors of commission (Ec) increases
the errors of omission (E ) and decreases the correct failures (FF).  To
illustrate specific values and trends among the STs, Tables 4-15 and 4-16
                                   4-48

-------
   60
   55
   45
   40
t  35
o
LU
a.  25
   20
   15
   10
   8.
           \
                 \
                  \


                       \
              \

              \\
                \\
                  \\
                                           	FF
                          0.60

                    4y
              / 1975
             / r- FTP
                 LEVEL \ i
         0.41
                       °- 60
                                            \\  /
                                             -

                                 0.75
                                         0.90
   5 -
0.4      0.5     0.6      0.7     0.8      0.9      1.0     1.1
                HC CUT-POINT, gm/mi
Figure 4-26.   Variation of Ec,  E0, and FF with HC  Cut-Point;
  CEV Fleet; Federal Short Cycle;  Bounded Errors of Commis-
  sion Method
                               4-49

-------
I
Ul
o
                                                                          ^M M» HV^B  C
                                                                                   0
                                                                                   FF
                        0
                         0
20        30       40
  HC CUT-POINT,  ppm
50
60
                        Figure 4-27.   Variation of Ec, Eo> and FF with HC Cut-Point;
                         CEV Fleet; NY/NJ Composite Test; Bounded Errors of Com-
                         mission Method

-------
60
                                                   	FF

                                                           1975  FTP
                                                   	 1975  FTP = 0.90
                                                           AT 0.90 FF <10%
                                                            FOR HIGH SPEED AND
                                                            LOW SPEED
0.41
                                                                  HIGH SPEED
                                                                  LOW  SPEED
                                                                  IDLE
                             400       500
                           HC CUT-POINT, ppm
       Figure 4-28.  Variation of Ec, Eo, and FF with HC Cut-Point;
        CEV Fleet; Key Mode Test; Bounded Errors of Commission
        Method

-------
ts>
                                                                              1975 FTP - 0.41
                                                                              1975 FTP - 0.9

                                                                              H • HIGH SPEED
                                                                              L • LOW  SPEED
                                                                              I • IDLE
                                         30        40       50        60        70
                                               HC CUT-POINT,  ppm
                        Figure 4-29-   Variation of Ec, E0,  and FF with HC Cut-Point;
                          CEV Fleet; Key Mode Test; Garage Instruments; Bounded
                          Errors of Commission Method

-------
                                                                  H - HIGH SPEED
                                                                  L - LOW SPEED
                                                                     IDLE
I
Ul
CO
                    0
                                                                                    ..--'
                                                                                           I
                     40
80
120
200    240     280
HC CUT-POINT,  ppm
320
360
400    440
                        Figure 4-30.   Variation of Ec,  E0, and FF with HC Cut-Point;
                          CEV Fleet; 1975 FTP HC Standard = 0.41 gm/mi; Federal
                          Three-Mode Test; Bounded Errors of Commission Method

-------
Ut
.*».
                                                         H • HIGH SPED
                                                         L - LOW SPEED
                                                             IDLE
              0
               200     250     300     350     400     450     500     550
                                          HC CUT-POINT,  ppm
600
650
                      Figure 4-31.   Variation of Ec, Eo, and FF with HC Cut-Point;
                        CEV Fleet;  1975 FTP HC Standard = 0.9 gm/mi; Federal
                        Three-Mode Test; Bounded Errors of Commission Method

-------
                                                                     H - HIGH SPEED
                                                                     L = LOW SPEED
                                                                     I = IDLE
                                                                                      ^
Ul
               LU
                            25
30
35
 40     45     50
HC CUT-POINT, ppm
55
60
65
70
                         Figure 4-32.   Variation of Ec> Eo, and FF with HC Cut-Point;
                          CEV Fleet; 1975 FTP Standard =  0.41 gm/mi; Federal Three-
                          Mode Test; Garage Instruments; Bounded Errors of Commis-
                          sion Method

-------
   40 r~
                                           H • HIGH  SPEED
                                           L • LOW SPEED
                                           I  • IDLE
               '\  '\     ^x  \
              '  \  V     v
                     50      60      70
                     HC CUT-POINT, ppm
Figure 4-33.  Variation of Ec, Eo, and FF with HC Cut-Point;
 CEV Fleet; 1975 FTP Standard = 0. 9 gm/mi; Federal Three-
 Mode Test; Garage Instruments; Bounded Errors of Commis-
 sion Method

-------
601
50)

40]
1975 FTP
LEVELX '
—  x
 0.41

301
201
101
              0.60
                     0.75
                          0.90
            0.60,
    0.41     0.60
                         0.90^
                                                         H = HIGH SPEED
                                                         L = LOW SPEED
                                                         I  - IDLE
                                                                         2200
                               UC
     Figure 4-34.   Variation of Ec,  EQ, and FF with HC  Cut-Point;
      CEV Fleet; Unloaded 2500 rpm Test; Bounded Errors of Com-
      mission Method

-------
                  55
                  50
40
in
oo
             1975 FTP LEVEL - 0.41
                                                  	FF
                                                  H  = HIGH SPEED
                                                  L  • LOW SPEED
                                                  I  - IDLE
                                                                    1975 FTP LEVEL • 0.9
                            100
                   120
140      160       180
   HC CUT-POINT, ppm
200
220
240
                          Figure 4-35.   Variation of Ec, Eo,  and FF with HC Cut-Point;
                           CEV Fleet; Unloaded 2500 rpm Test; Garage Instruments;
                           Bounded Errors of Commission Method

-------
Table 4-15.  Comparison of Selected ST Hydrocarbon Results.-
             CEV Fleet, Bounded Errors of Commission Analysis,
             HC FTP Level = 0.90 gm/mile (Ec = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
Idle
Low Speed
High Speed
Clayton Key Mode (Garage)
Idle
High Speed
Federal Three -Mode (Laboratory)
Idle
Low Speed
High Speed
Federal Three -Mode (Garage)
Idle
Low Speed
High Speed
2500 rpm Unloaded (Laboratory)
2500 rpm Unloaded (Garage)
Parameter, %
Eo
9
6.5

6
21
19.5

9.5
16

15
15
10

14
16
15
23
23
FF
22
25

22
9
10

22
16

21
21
25

17
16
17
8
9
                             4-59

-------
Table 4-16.  Comparison of Selected ST Hydrocarbon Results:
            CEV Fleet,  Bounded Errors of Commission Analysis,
            HC FTP Level = 0.41 gin/mile (E  = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
Idle
Low Speed
High Speed
Clayton Key Mode (Garage)
Idle
High Speed
Federal Three -Mode (Laboratory)
Idle
Low Speed
High Speed
Federal Three -Mode (Garage)
Idle
Low Speed
High Speed
Z500 rpm Unloaded (Laboratory)
2500 rpm Unloaded (Garage)
Parameter, %
E
o
11
7
7
35
30
10
21
17
17
11
18
22
20
38
37
FF
56
60
61
32
37
57
45
52
38
51
46
44
47
28
30
                            4-60

-------
summarize data from the figures at HC FTP levels of 0.41 and 0. 90.  On the
average, at both FTP levels, the bag tests (Federal Short Cycle and NY/NJ
Composite) have lower EQ and higher FF at the fixed E  = 5 percent than do
the volumetric tests.  However, the idle mode of the Clayton Key Mode (with
either laboratory or garage instruments) test produces similar results.  The
2500 rpm Unloaded test is very poor on a comparative basis.
4.2.2.5.1.2   Carbon Monoxide Emissions
              The variation of E , E  ,
                                      and FF as a function of CO cut-
point are displayed in Figures 4-36 through 4-43 for each ST examined,  and
for the range of CO FTP values selected in Table 4-10 (CO = 3.4 to 9.0).
The figures correspond to the following ST/FTP-level spectrum:
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode
(Laboratory)
Clayton Key Mode
(Garage)
Federal Three-Mode
(Laboratory)
Federal Three -Mode
(Garage)
2500 rpm Unloaded
(Laboratory)
2500 rpm Unloaded
(Garage)
CO FTP Level
3.4
X
X
X
X
X
X
X
X
5.0
X
X




X

7.0
X
X




X

9.0
X
X
X
X
X
X
X
X
Figure No.
4-36
4-37
4-38
4-39
4-40
4-41
4-42
4-43
                                   4-61

-------
                      30
                      25
                                                                                   FF
i
o*
ro
                      20
u" 15
                  LU
                  O
                   o.  10
X
                                                \
                                                  *
                                                    *
                                                       m
                     1975 FTP
                     LEVEL
                     -7.0

                     -9.0
                                                                                AT 5.0. 7.0. AND 9.0
                                                                                      FF<5%
                                                    CO CUT-POINT,  gm/mi
                          Figure 4-36.   Variation of Ec, Eo,  and FF with CO Cut-Point;
                           CEV Fleet; Federal Short Cycle; Bounded Errors of Commis-
                           sion Method

-------
i
<^
00
                         30 -
                                EQ AND  FF < 10% AT LEVELS  II,  III, AND  IV
                                                   CO CUT-POINT, ppm

                          Figure 4-37.   Variation of Ec,  E0, and FF with CO Cut-Point;
                            CEV Fleet; 1975 FTP CO Level = 3.4 gm/mi; NY/NJ Com-
                            posite Test; Bounded Errors of Commission Method

-------
   501—
	1975 FTP LEVEL =3.4   AT 3.4 FF < 10% FOR HIGH SPEED AND LOW  SPEED
	1975 FTP- 9.0        AT9.0FF<1%
   40
   30
 o
   20
UJ
O-
               	FF

               H • HIGH SPEED

               L • LOW SPEED

                 - IDLE
              300
    400
500
600      700      800
  CO CUT-POINT, ppm
900
1000
1100
1200
                 Figure 4-38.   Variation of Ec, Eo,  and FF with CO Cut-Point;
                   CEV Fleet; Key Mode Test; Bounded Errors of Commission
                   Method

-------
                                                               1975 FTP LEVEL - 3.4
                                                               EQ AND FF < 1% AT 1975 FTP
               9.0
            50 r~
            40
        o
        
-------
                                     «^-*"
               ^	L	_
                                                          HIGH SPEED
                                                          LOW SPEED
                                                          IDLE
       400     600
800    1000   1200
CO CUT-POINT, ppm
1400   1600    1800     2000
Figure 4-40.  Variation of Ec, E0, and FF with CO Cut-Point;
 CEV Fleet;  1975 FTP CO  Level = 3.4 gm/mi; Federal Three-
 Mode Test;  Bounded Errors of  Commission Method

-------
               50
               40
             .  30
             o
I
o
-J
           UJ
           UJ
           o
Ec
Eo
FF
                                                         H • HIGH SPEED
                                                         L = LOW SPEED
                                                         I  • IDLE

                                                                                    	H
                                              600    650     700
                                              CO CUT-POINT, ppm
                        Figure 4-41.  Variation of Ec,  Eo, and FF with CO Cut-Point;
                         CEV Fleet; 1975 FTP Level = 3. 4 gm/mi; Federal Three-
                         Mode Test; Garage Instruments;  Bounded Errors of Com-
                         mission Method

-------
                                                                  AT 5.0, 7.0 AND 9.0FF<3%



                                                                             	E
00
                      50
                      40
o


<  30


  "b
LU



UJ"


£  20

                                                       	3.4

                                                   ^•*"
                                                                    —5.0
                                            1975 FTP LEVEL

                                           500
                                  600       700


                                CO CUT-POINT,  ppm
800
900
                         Figure 4-42.  Variation of Ec, Eo, and FF with CO Cut-Point;

                          CEV Fleet; Unloaded 2500 rpm Test; Bounded Errors of Com-

                          mission Method
1000

-------
                                                    1975 FTP LEVEL
                                                    FF < 5%
                 3.4
   50
   40
-o30
   20
   10
  EQ AND FF< 5% AT 1975 FTP -  9.0
Ec
Eo
FF
                                  750       800       850
                                    CO CUT-POINT,  ppm
             900
950
1000
              Figure 4-43.  Variation of Ec, E0, and FF with CO Cut-Point;
                CEV Fleet; Unloaded 2500  rpm Test; Garage Instruments;
                Bounded Errors of Commission Method

-------
              As in the preceding case of hydrocarbon emissions, these
figures indicate the tradeoffs possible between E , E , and FF.  However,
for CO FTP levels above 3.4, the general or average CO levels of the CEV
fleet were sufficiently low, i.e., a very high percentage of the vehicles
were better than the 5.0, 7.0,  and 9.0 gm/mi  requirements,  that both
E and FF percentage values were very small  for all of the short test pro-
cedures.  This characteristic is summarized in Table 4-17 for the CO FTP
level of 9.0 gm/mi.
              At the 3.4 level,  however, as shown in Table 4-18, the bag
tests were sufficiently discriminatory to identify FF  values above 20 percent,
with E  values in the 14- to 16-percent range.  The volumetric tests, on
the other hand, all had high E  values (30- to 40-percent range) with very
low FF values (< 16).
4.2.2.5.1.3   Oxides of Nitrogen Emissions
              The variations of E , E , and FF as a function of NO   cut-
                                O    C                           X
point are  displayed in Figures 4-44 through 4-48 for  each ST examined, for
the single NO  FTP  values of 3. 1 gm/mi examined in the study.
              The significant results  at the  E  level of 5 percent are sum-
marized in Table 4-19 for comparative purposes.  As shown, the high-speed
mode of the volumetric tests (Clayton  Key Mode and Federal Three-Mode)
produced the highest FF values and the lowest  E values, and are thus indi-
cated to be superior for NO  discrimination purposes.
4.2.2.5.1.4  Weighted Three-Mode Tests
              In addition,  a bounded errors analysis was made for two
weighted Key Mode tests.  The first weighting  factors were based on the
multiple regression analysis of Sec. 4.2.1.2.  The second weighting factors
are suggested by Clayton Manufacturing  Co.    These latter factors were
developed for HC and CO, based on  1972  surveillance data.  The weighting
factors are given in Table 4-20.  The analysis was performed at FTP level I
 "Exhibit G, Short Tests Versus 1975 CVS Relatability Analysis/Correlation/
 Errors of  Commission and Omission," May 1973
                                   4-70

-------
Table 4-17.  Comparison of Selected ST Carbon Monoxide Results:
            CEV Fleet,  Bounded Errors of Commission Analysis,
            CO FTP Level =9-0 gm/mi  (E  = constant = 5%)
             Short Test
                                             Parameter,  %
                                                         FF
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
    Idle
    Low Speed
    High Speed
Clayton Key Mode (Garage)
    Idle
    Low Speed
    High Speed
Federal Three-Mode (Laboratory)
    Idle
    Low Speed
    High Speed
Federal Three-Mode (Garage)
    Idle
    Low Speed
    High Speed
2500 rpm Unloaded (Laboratory)
2500 rpm Unloaded (Garage)
                                                         < 1
                                            < 1
                                            < 1
< 1
                                                         < 1
                                            < 1
                                            < 1
                             4-71

-------
Table 4-18.  Comparison of Selected ST Carbon Monoxide Results:
            CEV Fleet,  Bounded Errors of Commission Analysis,
            CO FTP Level = 3.4 gm/mi (E = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
Idle
Low Speed
High Speed
Clayton Key Mode (Garage)
Idle
Low Speed
High Speed
Federal Three -Mode (Laboratory)
Idle
Low Speed
High Speed
Federal Three -Mode (Garage)
Idle
Low Speed
High Speed
2500 rpm Unloaded (Laboratory)
2500 rpm Unloaded (Garage)
Parameter, %
Eo
14
16

29
33
36

29
36
33

33
40
43

29
36
36
35
38
FF
22
20

15
11
8

14
7
10

16
8
7

14
7
7.5
8
6
                            4-72

-------
I
-4
Ul
I—
UJ
o
UJ
Q_
                       25
                       20
                   <   15
                   . o
10
                        0
                                                                        FF<5.4%
                                                                  I
                                             5                    6
                                             NOX CUT-POINT, gm/mi
                                                             J
                                                              7
                        Figure 4-44.   Variation of Ec, Eo, and FF with NOX Cut-Point;
                         CEV Fleet; Federal Short Cycle Test; Bounded Errors of Com-
                         mission Method

-------
                  20 r-
I
-J
                                                  NO  CUT-POINT, ppm
                                                    A
                        Figure 4-45.   Variation of Ec, Eo, and FF with NOX Cut-Point;
                          CEV Fleet; NY/NJ Composite Test; Bounded Errors of Com-
                          mission Method

-------
            25 i-
            20
i
Ul
          o  15
          j
          "L>
             10
                                     ON LOW SPEED FF < 2%
                                     ON IDLE       FF< 2.5%
	EQ
	FF
H = HIGH SPEED
i = LOW SPEED
I  = IDLE
200      400      600       800       1000      1200      1400      1600      1800
                                                    NOY CUT-POINT,  ppm
                                                      A
                                                                                               2000
                                                               2200
                          Figure 4-46.   Variation of Ec,  Eo, and FF with NOX Cut-Point;
                            CEV Fleet; Key Mode Test; Bounded Errors of Commission
                            Method

-------
I
-o
u. 30
u_
a
* 20
o
UJ
o
PERCENT E
=> 0
-
•••


1
t '
V I i i ''
0 400 800 1200
                                                   Ec
                                                   Eo
                                                   FF
                                                       1
H • HIGH SPEED
L • LOW SPEED
I  = IDLE
    I

                                              1600    2000    2400    2800    3200    3600   4000
                                               NO  CUT-POINT, ppm
                                                 A
                       Figure 4-47.   Variation of Ec,  EO)  and FF with NOX Cut-Point;
                         CEV Fleet; Federal Three-Mode Test; Bounded Errors of
                         Commission Method

-------
                             25
                             20



                         It

                         Q
                         -ZL  ...
                         <  15
-J
-4
10
                          UJ
                                               FF < 2%
                                                                        	FF
                                                            800
                                               NO  CUT-POINT, ppm
                                                 A
                                        900
1000
                        Figure 4-48.   Variation of Ec, E0, and FF with NOX Cut-Point;

                          CEV Fleet; Unloaded 2500 rpm Test; Bounded Errors of Com-

                          mission Method

-------
Table 4-19.  Comparison of Selected ST NOX Results:

            CEV Fleet, Bounded Errors of Commission Analysis,

            NO  FTP Level =3.1 gm/mi (E  = constant = 5%)
               X                          C
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)**'
Idle
Low Speed
High Speed
Federal Three -Mode (Laboratory)*31'
Idle
Low Speed
High Speed
2500 rpm Unloaded (Laboratory)'
Parameter, %
Eo
9.5
10

13
14
6.5

11
11
3
13
FF
5.5
5
•
2
<2
8.5

1
1
8.5
2
  (a)  Garage-type analyzers for NO were not available for ST

      evaluation.
                             4-78

-------
                Table 4-20.  Key Mode Weighting Factors
Origin
Regression Analysis



Clayton Report



Mode
High
Low
Idle
Constant
High
Low
Idle
Constant
Weights
HC
0.00025
0.00017
0.00174
0. 154
0.8736
0.8736
0.312
	
CO
6.67
- 1.79
72.65
2.07
0.66
0.66
0.33
	
NO
X
0.00116
0.00024
0.00204
0.929
	
	
	
	
only (see Table 4-10).  The results are depicted in Figures 4-49 to 4-51.
They clearly illustrate that the weighted volumetric tests are not signifi-
cantly better than the best single mode.
4.2.2.5.1.5  Variance Estimates
              As the plots in Figures 4-26 to 4-48 are predictions from the
data, the variability of these predictions should be addressed.  Referring to
Figures 4-7 and 4-9, the problem of estimating the  ST cut-point, for a fixed
FTP level,  is analogous to estimating the quantiles of a distribution function
(Ref. 4-5).  Thus, the  large sample standard deviation is  given by
                          PF
                         US)
                          dy
N
                                                                       (4-1)
                              y =
                                    4-79

-------
        o
        (XL
i
oo
o
            0
              0
                                                           W, • WEIGHTING  FACTORS BASED
                                                             1   ON REGRESSION COEFFICIENTS
                                                                CLAYTON  FACTORS
                                                                    H - HIGH SPEED
                                                                    L = LOW SPEED
                                                                    I  - IDLE
10
20
30
    40
PERCENT E
50
60
70
80
                      Figure 4-49.   Variation of Ec and Eo for Key Mode and Weighted
                        Key Mode Tests; CEV Fleet; 1975 FTP HC Level = 0.41 gm/mi;
                        Bounded Errors of Commission Method

-------
                                           I    L  H
oo
               0
                                                      w.
                                                               W, = WEIGHTING FACTORS  BASED
                                                                 1   ON REGRESSION COEFFICIENTS
                                                               W  - CLAYTON FACTORS
                                                                        H = HIGH SPEED
                                                                        L = LOW SPEED
                                                                        I  - IDLE
                 0
10
20
   30
PERCENT E.
40
50
60
                       Figure 4-50.   Variation of Eo and Ec for Key Mode and Weighted
                         Key Mode Tests; CEV Fleet; 1975 FTP CO Level = 3.4 gm/mi;
                         Bounded Errors of Commission Method

-------
                                   H
00
ro
                      0
                       0
                                                        W, - WEIGHTING FACTORS BASED
                                                          1   ON REGRESSION COEFFICIENTS
                                                                 H - HIGH SPEED
                                                                 L • LOW SPEED
                                                                 I •  IDLE
10       15
PERCENT E.
20
25
                       Figure 4-51.  Variation of Eo and Ec for Key Mode and Weighted
                        Key Mode Tests; CEV Fleet;  1975 FTP NOX Level = 3. 1 gm/mi;
                        Bounded Errors of Commission Method

-------
where
           PF = probability of passing the FTP
          P(y) = probability of an error of commission for ST
                cut-point set at value y
             •y = upper bound on probability of errors of commission
            N = sample size or number of cars in the data set
          LS = true cut-point  for the population.
As PF, p(y), and LSQ are unknown, they can only be approximated from the
data.  LSQ is, of course,  approximated by the  cut-point estimated from the
data.  PF is estimated on the percent passed by FTP divided by 100.  p(y) is
taken to be the locus of E  versus cut-point. |dp(y)/dy|T c  is taken as the
                          c                   I         l1-1^©
derivative of the E  versus cut-point curve evaluated at the cut-point of
interest (LS  J.  Equation (4-1)  will be used to discuss variablility of the pre-
dicted population.
              For a fixed FTP level, the  standard deviation  of the estimated
cut-point can be independently controlled by increasing the sample size.
Once the sample size is fixed,  this standard deviation varies inversely with
the magnitude of the derivative of the E  versus cut-point curve.   Thus,  in
regions where the curve is steep,  the variability of  the predictions will be
less  than in regions where the curve is flat.
              For example, at FTP level for HC =  0.41 and  Federal Short
Cycle cut-point of 0.4, Y * 0.045, N = 39,  PF =  0.33, and
                             dp(y)
                              dy
oc 0.25
                                  0.41
Thus, Eq.  (4-1) gives the approximate standard deviation of 0.044 gm/mi.
                                   4-83

-------
For the Federal Short Cycle point of 0.75,
                 dp(y)
                  dy
                      0.75
asO.033      ,       -ySO.003
and the standard deviation increases to 0.086 gm/mi.  Figure 4-52 illustrates
the effect of the cut-point uncertainty on the other computed quantities of EQ
and FF.  It shows that the uncertainty in the predicted results increases with
decreasing errors  of commission bounds.
4.2.2.5.2     Multiple-Constituent Tests
               In addition to analyzing each pollutant individually, an analysis
was made for multiple-constituent tests, using the contingency table  approach.
In a three-constituent test, a car fails the  ST if any of its HC,  CO, and NOx
measurements exceed  the previously determined cut-points.  These tests are
applicable to the bag tests, the unloaded test,  and the individual modes of the
three-mode volumetric tests.   Nine  constituent tests are applicable only to
the three-mode volumetric tests.  A car fails  the ST if any one of the modes
fails on its  three-constituent tests.  Only data analytic and parametric actual
results were computed.  A model for predicting population results was not
available.
               Shown in Figure 4-53 is the computational procedure followed
in determining the  multiple constituent tests.  Note that the cut-point selec-
tion policy is applied at the pollutant level and not at the multiple-constituent
test level.  For example, the percent E is bounded for individual pollutants
in the method of bounded errors of commission,  and this bound can possibly
be exceeded on a multiple-constituent test. In forming the multiple-constit-
uent test contingency table, the following definitions apply:
         Correctly passed (PP):       Car passes the ST and the FTP
         Correctly failed (FF):        Car fails the ST and the FTP
         Error of Commission (E ):   Car fails the ST and passes the FTP
         Error of Omission (E ):      Car passes the ST and fails the FTP
                                    4-84

-------
   70
   60
   50
 o40
UJ
£  30
LU
a.
    10


    0
     0
            T
            1
                                ERROR BAR INDICATES PLUS AND MINUS ONE
                                STANDARD DEVIATION OF THE ESTIMATE
                                           	FF

                       vk     A
234567
    HC FEDERAL SHORT CYCLE, gm/mi
      Figure 4-52.  Variability of Predicted Population Results
                               4-85

-------
     START
  Determine ST
  cut-points for
HC, CO, and NO,
 Write Results of
Contingency Table
   Analysis for
HC, CO, and
    Determine
  3-Constituent
  Test Results,
  Write Results
        YES
    Determine
  9-Constituent
  Test Results,
  Write Results
 Set
Policy
                      NO
                              -FINISH
 •FINISH
  Figure 4-53.  Computation Flow Chart
                4-86

-------
where FTP or ST failure occurs if any one of the test constituents exceeds
its respective cut-points.  A car is counted once in forming the table and
falls into one, and only one,  of the above categories.  The percent E  on a
multiple-constituent test may be larger than the largest individual pollutant
percent EC,  or may be smaller than the smallest individual pollutant percent
EC, depending upon the actual  data set and its particular mix of pollutant
failures.  A useful observation, using  the individual pollutant results, of the
actual percent E  for the multiple constituent test is
                c
             Percent E  =* (maximum pollutant ST percent FF)
                        X (minimum pollutant FTP percent PP)

Other useful relations for the multiple constituent tests are:

                 Percent FF > max (pollutant percent FF)
                 Percent PP < min (pollutant percent PP)

              The three-constituent test results for the Federal Short Cycle
and the Federal Three-Mode (high-speed and idle modes only) are summarized
in Figures 4-54 through 4-65.   The data plotted are the parametric population
results.  Both the laboratory and garage instrument results are displayed for
the Federal Three-Mode short test (Figures 4-58 through 4-65).
              The data were generated in the following manner: The method
of bounded errors of commission was used to determine an ST pass/fail cut-
point for each pollutant individually.   The three-constituent test results were
obtained by simultaneously comparing the observed emission  levels for HC,
CO, and NO  of a vehicle against the determined ST and the given FTP cut-
points .
                                    4-87

-------
          LU   o
          u   8
          at
              OK>—O
(
) <
> (
) (
_ <
)
1 2 3 4 5

oo
00
             801-
             60
             40
             20
                                                                 401-
                                                          30
                          PREDICTED PERCENT E.
                                                      "i 20
                                                             o
                                                          10
                                                                                    FF	
                                                                       PREDICTED PERCENT E
Figure 4-54.   Variation of Actual Ec, E0,
  and FF with Predicted Ec; Federal Short
  Cycle;  Three-Constituent Test; Bounded
  Errors of Commission Method; CEV Fleet;
  1975 FTP Level  I
                                                            Figure 4-55.   Variation of Actual Ec,  Eo,
                                                             and FF with Predicted Ec; Federal Short
                                                             Cycle; Three-Constituent Test; Bounded
                                                             Errors of Commission Method; CEV Fleet;
                                                             1975 FTP Level II

-------
           O
           O£
                0  0
                                            o
o
(J
LU
I 8
o
on
LLJ
Q_
i 4
«


—



r
IJ
c



>


5 <






) (






) (






) (




12345
I
oc
              401-
              30
20
              10
                                                               o
                                       FF
                401-
                30
                20
                10
                                                                                   I
                              2      3
                           PREDICTED PERCENT E
                         1       2       3
                            PREDICTED PERCENT E
        Figure 4-56.   Variation of Actual Ec, Eo,
          and FF  with Predicted Ec; Federal Short
          Cycle; Three-Constituent Test; Bounded
          Errors  of Commission Method; CEV Fleet;
          1975 FTP Level III
          Figure 4-57.   Variation of Actual Ec, Eo,
            and FF with Predicted Ec; Federal Short
            Cycle; Three-Constituent  Test; Bounded
            Errors of Commission Method; CEV Fleet;
            1975 FTP Level IV

-------
rfk

o
             ACTUAL PERCENT E. • 0 FOR IDLE MODE
   o
   et
c
0 HIGH SPEED
<
i (
> (
> (
) (
> <
      4 -
      0'	=t—"^
     80
     60
     40
     20
                             3       4

                                	FF
                    OHIGH SPEED
                    AIDLE
• HIGH SPEED

AIDLE
                  PREDICTED PERCENT EC

Figure 4-58.   Variation of Actual Ec,  Eo,
  and FF with Predicted Ec; Federal Three-
  Mode; Laboratory Instruments; Three-
  Constituent Test; Bounded Errors of Com-
  mission Method; CEV Fleet; 1975 FTP
  Level I
                                                               O HIGH SPEED
                                                               A IDLE
                                            RANGE OF VARIATION:
                                            - HIGH SPEED MODE
                                             - IDLE MODE
                          101-
                       o
                                                                           9
                                                            60
                                                            40
                          20
                                   OHIGH SPEED
                                   AIDLf
	FF

  • HIGH SPEED
  AIDlf
                                     PREDICTED PERCENT E


                     Figure 4-59.   Variation of Actual Ec>  Eo,
                      and FF with Predicted Ec; Federal Three-
                      Mode; Laboratory Instruments; Three-
                      Constituent Test; Bounded Errors of Com-
                      mission Method; CEV Fleet; 1975 FTP
                      Level II

-------
         O HIGH SPEED
         A IDLE
     20
     10
          0   0
    RANGE OF VARIATION:
    — HIGH SPEED MODE

    ~ IDLE MODE
      O
                                  0      C
  1'
  O
      40
      30
      20
      10
                   OHIGH SPEED

                   AIDLE
         	FF

           • HIGH SPEED
           AIDLE
               IDLE
               I	
                                    HIGH
I
I
              1      2      3
                  PREDICTED PERCENT
Figure 4-60.   Variation of Actual Ec, Eo,
  and FF with Predicted EC; Federal Three-
  Mode; Laboratory Instruments; Three-
  Constituent Test; Bounded Errors of Com-
  mission Method; CEV Fleet; 1975 FTP
  Level  III
                                                          LU
                                                          o
                                                          O
                                                          
-------
*.
sO
           O HIGH SPEED
           A IDLE
   I
   o:
   o
      10 -
       5 	
      80
      60
 o
      20
RANGE OF VARIATION:
— HIGH SPEED MODE
- IDLE MODE
-
— (


> (

(
> ,
i

7
(.
)
*
r /
i t
t j
i


01234
                    O HIGH SPEED
                    AI OLE
    	FF
       • HIGH SPEED
       A IDLE
                   PREDICTED PERCENT E_
                                                           o
                                                           u
                                                              20
                                   10
                                                              40
                                                           t
                                                         cj
30
                                                              20
                                                              10
     O HIGH SPEED
     A IDLE
                                                                                           RANGE OF VARIATION:
                                                                                           - HIGH SPEED MODE
                                                                                            - IDLE MODE
                                                                to—c

                                                                         i
                                                                                             4      5
                                                                                             	FF
                   OHIGH SPEED
                   AI OLE
                                                                                                        HIGH SPEED
                                                                                                        IDLE
                                                  ^"TA
                                     ••   ^     *     \
                                     **•*"             IDLE
                                                                                             I
                                            1234
                                                PREDICTED PERCENT E.
Figure 4-62.   Variation of Actual Ec, E0,
  and FF with Predicted Ec; Federal Three-
  Mode; Garage Instruments; Three-
  Constituent Test; Bounded Errors of Com-
  mission Method; CEV Fleet; 1975 FTP
  Level I
                              Figure 4-63.   Variation of Actual Ec,  Eo>
                               and FF with Predicted Ec; Federal Three-
                               Mode; Garage Instruments; Three-
                               Constituent Test; Bounded Errors  of Com-
                               mission Method; CEV Fleet;  1975 FTP
                               Level n

-------
  o
  ec.
  o
     10
          ACTUAL PERCENT L • 0 FOR IDLE MODE
                        c

                 OHIGH SPEED MODE
                                            O
     401—
                           ---- FF
                               HIGH SPEED MODE
                               IDLE MODE
               1       2       3
                  PREDICTED PERCENT E,
Figure 4-64.   Variation of Actual Ec, Eo,
  and FF with Predicted EC; Federal Three-
  Mode; Garage Instruments; Three-
  Constituent Test; Bounded Errors  of Com-
  mission Method; CEV Fleet; 1975 FTP
  Level III
      10

              O HIGH SPEED
              A IDLE
   RANGE OF VARIATION:
   — HIGH SPEED MODE
    ~ IDLE MODE
                                                                                     ft
     40
     30
                                                             20
                                                             10
                O HIGH SPEED
                AIDLE
	FF

   • HIGH SPEED
   AIDLE
                  PREDICTED PERCENT E.
Figure 4-65.    Variation of Actual Ec, Eo,
  and FF with Predicted Ec; Federal Three-
  Mode; Garage Instruments; Three-
  Constituent  Test; Bounded Errors of Com-
  mission Method; CEV Fleet; 1975 FTP
  Level IV

-------
              The results in the figures are presented as follows: for each
three-constituent test, the actual errors of omission,  actual correct failures,
and actual errors  of commission are plotted versus the predicted error of
commission that was the original bound on the individual pollutants.  For
example,  in Figure 4-54, if policy was set at a maximum of 2 percent errors
of commission on  each individual pollutant, the actual results of a three-
constituent test  are about 30 percent errors of omission, 36 percent correct
failures,  and 2.6 percent errors of commission.  For the actual errors of
commission, plus or minus one standard error of the estimate is indicated
by an error bar, with the value of the estimate  in the center of the bar.  In
the above example, plus one standard error of the estimate gives about
5 percent E ,  while the minus  side shows about 0. 1 percent E  ,  with the
observed  value being 2.6  percent E .
4.2.2.5.2.2  Variance Estimates
              For fixed ST and FTP cut-points, the cell counts in a 2 X 2
contingency table are binomially distributed when the observations are inde-
pendent (Ref. 4-6).  Since the ST cut-points are computed from data prior to
forming the contingency table,  there is statistical dependence between the ST
cut-points and the resulting table.  Hence,  the binomial distribution will be
an approximation to the true distribution. Thus,  the approximate standard
deviation  is
                                X (100 - X
                                     N
where
                         X = cell count in percent
                         N = total table count
                                   4-94

-------
For example,  if the percent errors of omission is computed to be 50 percent,
then, with 40 cars, the standard deviation is 7.9 percent.  Table 4-21 shows
the approximate cell standard deviations for the range of cell percentages
assuming N = 40.  This procedure was also used to calculate the standard
error of the estimate depicted in Figures 4-54 through 4-65 by the error
bars on the actual errors of commission.
4.2.2.5.2.3   Discussion of Results
              As the FTP cut-points increase from level Set I to  level Set IV,
the resulting actual errors of commission tend to increase for a given pre-
dicted level of errors of commission.  For example,  on the Federal Short
Cycle at 2 percent predicted errors of commission,  the actual errors are
2.5 percent, 5 percent, 5 percent, and 7.5 percent for FTP levels I, II,  III,
and IV.  respectively.  This trend is not present for the garage instrument
results as shown in Figures 4-62 through 4-65.
               Table 4-21.  Approximate Standard Deviation
                            for Three-Constituent Tests -
                            CEV Fleet, N = 40
Cell
Percentage
60
50
40
30
20
12.5(a)
10
5
-------
              A comparison of the modes on the Federal Three-Mode test
shows that,  for a fixed predicted percent EC, the high speed mode has a higher
percent FF and Lower percent E  than does the idle mode.  This is true
regardless of instrumentation or FTP level.  However,  the actual percent E
is generally lower on the idle mode than on the high-speed mode, but this
difference is not always significant.
              A comparison of different modes or ST should be made on a
fixed actual percent E basis.  This is, of course, difficult to do because of
the computational procedure followed.  It can be approximately performed,
however.  Consider comparing the Federal Short Cycle  to the Federal Three-
Mode. At FTP level I, the actual percentages of E are approximately the
same for the high speed mode and the Federal Short Cycle (statistically, they
are equivalent).  Now, comparing the percent FF  and percent E  curves,
percent FF and percent E  are both higher  on the  high-speed mode than the
Federal Short Cycle.  This difference is not statistically significant  at the
95 percent level,  and the two tests would have to be judged as equal.  Also,
at the 95 percent level, the high-speed mode is  superior to the  idle mode.
              The  differences  between laboratory and garage instruments
are quite predictable,  based upon the previous results from individual pollu-
tants.  For a fixed  predicted percent E ,  on their respective modes,
        a.    Actual percent E  is higher for garage instruments than
              for laboratory instruments
        b.    Actual percent FF is lower for garage instruments than
              for laboratory instruments
        c.    Actual percent E  is higher for garage instruments than
              for laboratory instruments.
                                   4-96

-------
4.3           REFERENCES FOR SECTION 4

              The following references are used in Section 4:

4-1.   W.  J. Dixon,  ed. ,  BMPD Biomedical Computer Program, University
       of California Press,  Berkeley (1974).

4-2.   D. F. Morrison, Multivariate Statistical Methods,  McGraw-Hill
       Book Co.,  Inc., New York (1967).

4-3.   F. A. Graybill, An Introduction to Linear Statistical Models, Vol. 1,
       McGraw-Hill Book Col,  Inc.,  New York (1961).

4-4.   T. W.  Anderson,  Multivariate Statistical Analysis,  John Wiley and
       Sons, Inc., New York (1958).

4-5.   H. Cramer,  Mathematical Methods of Statistics, Princeton University
       Press,  New  Jersey (1971).

4-6.   C. R. Rao, Linear Statistical Inference and Its Applications, John Wiley
       and Sons, Inc., New York (1965).
                                    4-97

-------
5.  IN-USE 1974 MODEL YEAR
      VEHICLE FLEET

-------
             5.  IN-USE 1974 MODEL YEAR VEHICLE FLEET
              This section summarizes the results of statistical analyses
conducted to determine the degree of correlation existing between the vari-
ous short tests (STs) and FTP tests conducted on a fleet of in-use 1974
model year vehicles.
              Several distinguishing features of the  1974 model year fleet
resulted in variations in focus and scope of the statistical analyses from
those reported for the CEV fleet in Section 4.  They  include:
         a.    The  1974 model year fleet was manufactured to known
              emission standard values, whereas  the CEV fleet was not.
         b.    The  1974 model year fleet population was stratified by
              three inertia test weight groups, whereas the CEV fleet
              was  at a. single inertia test weight value.
         c.    There was no substantial number of replicate test obser-
              vations for the 1974 model year fleet.
              The  appropriate  1975 FTP emission standards for the 1974
model year fleet were computed to be:
              HC = 3.02 gm/mi
              CO = 28.0 gm/mi
             NO  =3.1 gm/mi
The three inertia test weight  groups were designated as:
              Group A (4000-Ib class)
              Group B (2750-Ib class)
              Group C (5500-Ib class)
              For  analysis purposes, laboratory instrument test data were
available for 147 cars,  while  garage-type instrument data were available for
144 cars.  These test data had been processed by EPA and were  received
stored on magnetic tape. Correlation analysis results are summarized in
Sec.  5. 1; the contingency table  analysis results are  summarized in Sec. 5.2.
                                    5-1

-------
5. 1           CORRELATION ANALYSIS RESULTS
              A conventional  correlation analysis was made for the 1974
model year fleet.  The method was as described in Sec.  4.2. 1. 1.  The re-
sulting ST/FTP correlation coefficients are summarized in Table 5-1 for
the individual inertia test weight groups (A,  B, C) and for the pooled vehi-
cle population (combined groups A, B, C).  For N = 147 cars,  a  computed
correlation coefficient greater than 0. 16 indicates that the ST and FTP pol-
lutants are statistically correlated with 95% confidence.  For N = 48 to 50,
this threshold is approximately 0.29.
              In addition, a correlation analysis of FTP composite emis-
sions versus FTP bags 2  and 3 was made  (by the method outlined in
Sec. 4.2. 1. 1. 1).  The results  are shown in  Table  5-2.
              As  can be seen in Table 5-1,  no single ST performs con-
sistently well on all three individual groups, or on a pooled basis.  Gen-
erally, the STs  are unable to track HC and CO emission levels on Group C
(5500-lb Chevrolet vehicles).  This is also supported by the FTP composite
versus bags 2 plus 3  correlations of Table 5-2.  The low correlation for
NO  in Group C in Table 5-2 is the result of a single outlying point and,
thus,  does reflect an usually low relatability.   However, the HC  and CO
correlations for Group C  are significantly different (in the sense of a rigor-
ous statistical test) than those  of Groups A and B.   This would indicate that
"hot" procedures would not perform as well on Group C  as on Groups A
and B.
              The presence of one ST with  good NO  correlation across the
                                                  Ji
population is missing in the 1974 model year fleet. From a correlation view-
point, the garage analyzers are inferior to the laboratory analyzers.  ST
ratings using the scale established in  Sec. 4. 2. 1. 4. 4 for the CEV fleet are
given in Table 5-3.  As with the CEV  fleet (Table  4-8), the bag-type STs
have higher ratings than the volumetric tests.  The 2500 rpm Unloaded test
shows substantially higher correlation for the 1974 model year fleet than for
the CEV fleet.   The  extreme CO tracking deficiency for the CEV fleet data
is not evident for the 1974 model year fleet.
                                    5-2

-------
Table 5-1.  Correlation Coefficient Summary:
            1974 Model Year Fleet
Short Test
Federal
Short
Cycle

NY/NJ
Composite


Key Mode
(Laboratory)










Key Mode
(Garage)



Vehicle
Group *a)
Pooled
A
B
C
Pooled
A
B
C
Pooled

A


B


C


Pooled

A

Test
Mode






High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
N(b)
147
50
48
49
147
50
48
49
147

50


48


49


145

50

ST/FTP Correlation
Coefficient*0'
HC
0.932
0.933
0.897
0.383
0.906
0.911
0.920
0.513
0.757
0.776
0.793
0.590
0.595
0.723
0.812
0.868
0.825
0.238*
0.228*
0.460
0.528
0.545
0.455
0.228*
0. 151*
0.245*
CO
0.905
0. 972
0.897
0.476
0.890
0.950
0.857
0.498
0.518
0.769
0.739
0.514
0.827
0.704
0.262*
0.738
0.650
-0. 195*
0.435
0.757
0.507
0.472
0.470
0.563
0.652
0.372
NO
X
0.355
0.780
0. 104*
0.674
0.060*
0.733
0.005*
0.611
0. 521
0.419
0.463
0.562
0.495
0.381
0.731
0.635
0.548
0.555
0. 580
0. 571




                     5-3

-------
Table 5-1.  Correlation Coefficient Summary:
           1974 Model Year Fleet (Continued)

Short Test






Federal
Three -Mode
(Laboratory)










Federal
Three -Mode
(Garage)





Vehicle
Group'*'
B


C


Pooled

A


B


C


Pooled

A



Test
Mode
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle
High
Low
Idle

N(b)
46


49


147

50


48


49


145

50


ST/FTP Correlation
Coefficient^)
HC
0.478
0.765
0.692
0.191*
0. 198*
0. 100*
0.766
0.771
0.803
0.507
0.523
0.709
0.890
0.859
0.851
0.522
0.533
0.252*
0.474
0.531
0.632
0.138*
0.107*
0.660
CO
0.362
0.540
0.560
-0.221*
-0.091*
0.229*
0.604
0.729
0.734
0.717
0.801
0.724
0.278*
0.737
0.622
0.159*
0.592
0.733
0.387
0.409
0.476
0.533
0.597
0.397
NO
X






0.467
0.453
0.411
0.492
0.664
0.369
0.722
0.611
0.665
0.552
0.707
0.639





                   5-4

-------
              Table 5-1. Correlation Coefficient Summary:
                         1974 Model Year Fleet (Continued)
Short Test






2500 rpm
Unloaded
(Laboratory)

2500 rpm
Unloaded
(Garage)

Vehicle
Group la>
B


C


Pooled
A
B
C
Pooled
A
B
C
Test
Mode
High
Low
Idle
High
Low
Idle




N(b)
46


49


147
50
48
49
147
50
46
49
ST/FTP Correlation
Coefficient(c)
HC
0.536
0.763
0.717
0.095*
-0.008*
-0.060*
0. 809
0.832
0.865
0. 107*
0. 574
0.487
0.781
-0.064*
CO
0.268*
0.539
0.550
-0.083*
0.239*
0.392
0.740
0.812
0.724
0.350
0.447
0.676
0.684
-0.051*
NO
X






0.447
0.524
0. 577
0.679


(b)

(c)
;A = Chrysler (4000 Ib)
B = Ford (2750 Ib)
C = Chevrolet (5500 Ib)
Pooled = Groups A + B + C

Number of cars in the data set

The correlations are statistically significant at the 95 percent confidence
level  except where indicated by an asterisk.  ST and FTP uncorrelated
for correlations below 0.28.
                                  5-5

-------
       Table 5-2.  FTP Composite Versus Bag 2 + 3 Correlation
                  Coefficients:  1974 Model Year Fleet
Vehicle Group

Pooled
A
B
C
N(a)

147
50
48
49
FTP/FTP Bag 2 + 3
Correlation Coefficient^'
HC
0.992
0.987
0.998
0.965
CO
0.994
0.993
0.996
0.987
NOX
0.925
0.976
0.996
0.761
'
Number of cars in data set

The correlations are statistically significant at the 95% confidence level
                                 5-6

-------
Table 5-3.  ST Ratings:  1974 Model Year Fleet
Short Test
Federal Short Cycle



NY/NJ Composite



Key Mode
(Laboratory)


Key Mode
(Garage)


Federal Three -
Mode (Laboratory)


Federal Three -
Mode (Garage)


Vehicle
Group (a)
Pooled
A
B
C
Pooled
A
B
C
Pooled
A
B
C
Pooled
A
B
C
Pooled
A
B
C
Pooled
A
B
C
Ratings
HC
E
E
G
P
E
E
E
P
G (I)(c)
G (I)
G (L)
P (D
P (L)
U
G (L)
U
G (I)
G (I)
G (H)
P (L)
F (I)
F (I)
G (L)
U
CO
E
E
E
P
G
E
G
P
G (L)
G (L)
G (L)
G (I)
P (H)
F (L)
P (L)
U
G (I)
G (L)
G (I)
G (I)
P (D
P (D
P (D
P (D
N0x
P
G
U
F
U
G
U
F
P (H)
P (H)
G (H)
P (L)



P (H)
F (L)
G (H)
G (L)



                      5-7

-------
              Table 5-3.  ST Ratings:  1974 Model Year Fleet
                         (Continued)
Short Test

2500 rpm Unloaded
(Laboratory)


2500 rpm Unloaded
(Garage)


Vehicle
Group la'

Pooled
A
B
C
Pooled
A
B
C
Ratings (b)
HC

G
G
G
U
P
P
G
U
CO

G
G
G
P
P
F
F
U
NO
X
P
P
P
F




(b)
;A = Chrysler (4000 Ib)
 B = Ford (2750 Ib)
 C = Chevrolet (5500 Ib)
 Pooled = Groups A + B + C

 Rating scale as in Sec.  1. 1.1. 8
{C) I = idle
  L = low speed mode
  H = high speed mode
                                   5-8

-------
 5.2            CONTINGENCY TABLE ANALYSIS RESULTS
 5. 2. 1          Maximum Correlation Method
               Using the  method as defined in Sec.  4. 2. 2. 1, a maximum
 correlation analysis was made for the pooled sample population of the  1974
 model year fleet.  Table 5-4 summarizes  the analysis  results for the pre-
 dicted population.  For N =  147, a computed table correlation coefficient
 greater than 0.16 indicates  that the ST and FTP pollutants are statistically
 correlated with 95% confidence.
               Examination  of these results indicates that the correlation
 indices (Table 5-4) are quite similar to the relative ST ratings developed
 in Table 5-3.  NO  tracking difficulty is indicated by a high percentage of
 E relative to percent FF.  CO is the dominant variable in that it has the
 highest percent FTP failure rate.   (For the CEV fleet, the dominant variable
 was  HC.)
 5. 2. 2          Bounded Errors of Commission Method
               A contingency table analysis for the 1974 model year fleet
was made using the methods  described in Sec. 4. 2. 2. 1. 2 and 4. 2. 2. 2.  For
this analysis the bound on percent E  was varied from 5 percent to 1 percent
in 1 percent increments, with the values 0. 5 percent and 0. 1 percent included.
The results of the analysis are summarized below.  The data shown are for
the predicted  1974 model year fleet population.
5.2.2.1        Single-Constituent Tests
5.2.2.1.1      Hydrocarbon Emission
               The variations of E ,  E , and FF as a function of HC cut-point
                                 o    c
are shown in Figures 5-1 through 5-8 for each ST examined. The graphical
displays indicate the general  nature  of the  tradeoffs available for policy
formulation.  Reducing the errors of commission (Ec) increases the errors
of omission (E ) and decreases the correct failures (FF).   To illustrate
specific values and trends among the STs,  Table 5-5 summarizes data  from
the figures for the E  value of 5 percent.
                                     5-9

-------
Table 5-4.   Maximum Correlation Summary; 1974 Model Year
              Fleet,  Predicted Population

ST

Federal Short Cycle



NY/NJ Composite



Key Mode
(Laboratory)








Key Mode
(Garage)




Federal Three -Mode
(Laboratory)










Federal Three-Mode
(Garage)




2500-rpm Unloaded
I Laboratory)


2500-rpm Unloaded
(Garage)

N

147



147



147









144





147











144





147



144


Test Mode









High speed



Low speed

Idle



High speed

Low speed

Idle

High speed



Low speed



Idle



High speed

Low speed

Idle








Pollutant

HC
CO
NO
X
HC
CO
NO
X
HC
CO
NO
X
HC
CO
HC
CO
NO
X
HC
CO
HC
CO
HC
CO
HC
CO
NO
X
HC
CO
NO
X
HC
CO
NO
X
HC
CO
HC
CO
HC
CO
HC
CO
NO
X
HC
CO

% E
C
6.05
5.02
26.8

7. 20
5.37
38.56

12. 2
9.99
20.4

11.7
7.35
11. 17
7.74
22. 63

18. 1
10. 1
17. 4
10. 4
11 4
10.4
11.9
9.20
22. 5

11.8
7.86
23.0

6.80
7.80
24.6

18.9
11.1
17.7
10.«
12.2
10.3
10.7
7.72
23.3

16.5
10.6

% E

5.44
6.94
7.56

6.36
7. 60
8.77

10.0
20.3
6.77

9.64
12. 13
9.29
13. 16
7.06

13.9
20. 6
13. 5
21.7
14. 8
21.7
9. 83
17.5
7.04

9.74
13. 5
7. 11

4.08
13. 33
7.31

14.4
24.3
13.7
23.7
15. 4
21.4
8.94
13. 1
7. 14

12.9
22.3

% FF

35.4
65.5
10. 5

34.5
64.8
9.24

30.8
52. 1
11.24

31.2
60.3
31.5
59.3
10.9

27.2
51.7
27.6
50. 62
26.3
50.7
31.0
54.8
11.0

31. 1
58.9
10.1

28. 6
59. 1
10.70

26.7
48.0
27.41
48.7
28.9
50.9
31.9
59.3
10.9

28.2
50.0
Table
Correlation
Index* a>
0.763
0.708
0.201

0.720
0.684
0.033*

0. 544
0.330
0.310

0.563
0.641
0. 580
0. 511
0. 270

0.351
0.322
0.371
0.297
0.305
0. 298
0. 553
0. 394
0.273

0. 558
0.502
0.263

0.759
0. 506
0. 236

0.326
0. 239
0.363
0. 254
0. 437
0.303
0. 597
0.513
0.259

0.402
0.283
   The correlation is statistically significant at the 45*o confidence level except where
   indicated by an asterisk.
                                  5-10

-------
   30
   25
t  20
a
o
OC
UJ

°-  10
                   \

                     *
'.HC

\
 \
 \
                        \/

                         )(
                        /  \
                    /HC

                                                         0

                                                         FF
                                       NO,
     0
                  HC AND NOX CUT-POINT,
     Figure 5-1.   Variation of Ec, Eo» and FF with HC

      and NOX Cut-Point;  1974 Model Year Fleet; Fed-

      eral Short Cycle Test; Bounded Errors of Com-

      mission Method
                           5-11

-------
I
t-*
IS)
                  35
                  30
                  25
  20
j°
"o
               O
               oc
   15
                  10
                                         0
                                   	FF
                         •^  NO,
                                                     \
                                        \
                                          \

                                                                A
  /
/HC
                                                        \
*\HC
  •
                                      100
                                125       150       175
                              HC AND NOX CUT-POINT,  ppm
            225
                                    250
               Figure 5-2.  Variation of Ec,  Eo, and FF with HC and NOX Cut-Point; 1974 Model
                Year Fleet; NY/NJ Composite Test; Bounded Errors of Commission Method

-------
(Jl
I
(JO
                  30 -
                  25
                  20
o
Q£
LU
Q_
                   10
1500       2000       2500      3000      3500      4000
                               HC  CUT-POINT,  ppm
4500
                                                                        H = HIGH SPEED

                                                                        L = LOW SPEED

                                                                        I  - IDLE
                                                                                       5000
                                                                                  5500
                         Figure 5-3.   Variation of Ec, Eo,  and FF with HC Cut-Point;

                           1974 Model Year Fleet; Key Mode Test; Bounded Errors of

                           Commission Method

-------
Ol
I
                  40
                  30
u°
 -.,  20
             O
                  10
                    ~~
                        H
                                                                 	E
                                                                          o
                                                                         FF
                                          H « HIGH SPEED
                                          L - LOW SPEED
                                          I - IDLE
                   60
100
                                  120       140       160
                                   HC CUT-POINT,  ppm
                        Figure 5-4.   Variation of Ec, Eo, and FF with HC Cut-Point;
                         1974 Model Year Fleet; Key Mode Test; Garage Instruments;
                         Bounded Errors of Commission Method

-------
    35
    30 -
    25
<   20
. .o
o±
LU
D_
    15
    10
                                        /
                                      .'
                     1            V  '
              H    /     L      >C         /
               \  '        \   /  \   /
              A
             H7     \
V        v
 \       X  *         	En
f \     /  \               0
    \     X^      ^              f-r
                      *
                                     \
                        «
                          \
                            *
\         Xs
    \
               H
     1500      2000      2500      3000     3500     4000
                                  HC CUT-POINT, ppm
                                                          4500
                          H = HIGH SPEED
                          L = LOW SPEED
                          I  = IDLE
                           5000
                                               5500
         Figure 5-5.   Variation of Ec,  Eo, and FF with HC Cut-Point;
           1974 Model Year Fleet;  Federal Three-Mode Test; Bounded
           Errors of Commission Method

-------
              40
               30
J°
J° 20
Ul
I
o
oc
LU
OU
               10
                     H

                                                                    o
                                                                    FF
                                                                  H - HIGH SPEED
                                                                  L - LOW SPEED
                                                                  I  - IDLE
                60
                        100
                                 120       140       160
                                   HC CUT-POINT, ppm
180
200
220
                        Figure 5-6.   Variation of Ec, Eo,  and FF with HC Cut-Point;
                          1974 Model Year Fleet; Federal Three-Mode  Test; Garage
                          Instruments; Bounded Errors of Commission  Method

-------
01
I




25

fc 20
1
L«J^
o15
LU
UJ
O
s 10

t


n
tc /
rL /
	 E /
0 y
	 FF HC /
— \ / 40
\ /
X t 30
XN°x / \ 1
/ / \ _
/ / • LU°
* f \ ULJ
1 ' \ - ?n
/ HC ^ upa
\ %
\ 0
\ § 10
\
\
NOX VHC X
\ x 0
~ /
/
_ /
/ c
	 E
w
	 FF
\
_ \
\
\A , , ,
NOA V 0 100 200 300 400
V ^^^^ HC CUT-POINT, ppm
V i i i • — i— -^ i
 "0       500     1000      1500     2000     2500     3000
                HC AND NOX CUT-POINT,  ppm


Figure  5-7.   Variation of Ec,  Eo, and FF with HC
 and NOX Cut-Point;  1974 Model Year Fleet;
 Unloaded 2500 rpm Test; Bounded  Errors of Com-
 mission Method
                                                                       Figure 5-8.   Variation of Ec, E0,
                                                                         and FF with HC Cut-Point; 1974
                                                                         Model Year Fleet; Unloaded
                                                                         2500 rpm Test; Garage Instru-
                                                                         ments; Bounded Errors  of Com-
                                                                         mission Method

-------
Table 5-5.  Comparison of ST Hydrocarbon Results: 1974 Model
           Year Fleet, Bounded Errors of Commission Analysis
           (E  = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
Idle
Low Speed
High Speed
Clayton Key Mode (Garage)
Idle
Low Speed
High Speed
Federal Three -Mode (Laboratory)
Idle
Low Speed
High Speed
Federal Three -Mode (Garage)
Idle
Low Speed
High Speed
2500 -rpm Unloaded (Laboratory)
2500 -rpm Unloaded (Garage)
Parameter, %
E
o
6.5
8.5

16
17
18

11.5
14
13

15.5
17.5
18

17
14
12
16
16
FF
34.5
32

24.5
23. 6
22.5

29
27
28

> 25
23
23

24
27
29
26
26
Figure No.
5-1
5-2

5-3
5-3
5-3

5-4
5-4
5-4

5-5
5-5
5-5

5-6
5-6
5-6
5-7
5-8
                            5-18

-------
              The bag tests (Federal Short Cycle and NY/NJ Composite)
have lower EQ and higher FF at the fixed E  = 5 percent condition than do
the volumetric tests.  There  is little difference shown between the various
volumetric STs.
5.2.2. 1.2    Carbon Monoxide Emission
               The variations of E , E  , and FF as a function of CO cut-
                                 o   c
point are shown in Figures 5-9 through 5-16 for each ST examined.  As in the
preceding area of hydrocarbon emission, these figures indicate the possible
tradeoffs between EC, EQ, and FF.  To illustrate specific values and trends
among the STs,  Table 5-6 summarizes data from the figures for the E
value of 5 percent.
               The bag-type STs (Federal Short Cycle and NY/NJ Composite)
exhibit  excellent CO tracking characteristics; the E values are considerably
better (lower) than the volumetric tests, and the  FF values are the highest.
When garage-type instruments  are used, the E  values are essentially doubled
(over laboratory instrument values) and FF values  are  significantly reduced.
5.2.2.1.3     Oxides of Nitrogen Emission
               The variations of E , E  and FF as  a function of NO  cut-point
                                 O   C                          3C
are shown in Figures  5-1, 5-2,  5-7, 5-17, and 5-18 for each ST examined.
The significant results at the E  level of 5  percent  are  summarized in Table
5-7 for  comparative purposes.  As can be noted, all STs identified very low
percentages of correctly failed vehicles (FF), <5 percent, while having
significant errors of  omission, ~15 percent.
5.2.2.1.4     Variance Estimates
               The general variance trends discussed in Sec.  4. 2. 2. 5. 1. 5
for the CEV fleet are also applicable in this case.  However,  the actual
magnitude of the  standard  deviation is different for the  1974 model year
fleet.  For the example illustrated in Sec.  4. 2. 2. 5.1.5, the  Federal Short
Cycle HC cut-point is 2. 6  gm/mi at \ = 0.045 (Figure 5-1).  PF is 0.048 and
                                    5-19

-------
I
M
O
65



60



55
u_
u.
H—
8«
g50
0.

45



40

yj
r~ 30



25

o
UJ
-Q 20
*0
z
rAjl_ ^|J «t
ce
UJ
CL
10



5
n
	 •v
*N
X /
%v /
- \ /
\
\ /
ft jf
\ /
A
/ \
mLrri j ^
\
\
_ / \
s'
s
s
— x.^
1 1 l^^T 	 4-
W15 20 25 30 35 4t
                                                                                       0
                                                                                      FF
                                                                         «.
                                                                            \
                                                                                \
45
                                                                                       50
                                                  CO CUT-POINT,  gm/mi
                        Figure 5-9.   Variation of Ec, Eo, and FF with CO Cut-Point;
                         1974 Model Year Fleet; Federal Short Cycle Test; Bounded
                         Errors of Commission Method

-------
                      351-
I
t\>
              60 f~
              3( —
0
600
                                 700       800      900      1000     1100     1200      1300      1400      1500
                                                        CO CUT-POINT,  ppm
                          Figure 5-10.   Variation of Ec,  Eo, and FF with CO Cut-Point;
                            1974 Model  Year Fleet; NY/NJ Composite Test; Bounded
                            Errors of Commission Method

-------
fc
Q
   70 r-
   60
   50
   40
   20
    10
                                  H
           \
         \     /
           ><        /

         X   \
         '        \         \
                     \
                                                      	E.
                                                                  H - HIGH SPEED
                                                                  L = LOW SPEED
                                                                  I  = IDLE
3000      10.000    15.000    20.000
                                        25.000    30.000    35.000
                                            CO CUT-POINT, ppm
40.000    45.000     50,000    55.000
                Figure 5-11.   Variation of Ec,  Eo, and FF with CO Cut-Point;
                  1974 Model Year Fleet;  Key Mode Test;  Bounded Errors of
                  Commission Method

-------
UJ
    60 -
    50 -
    40
    30
    20
     10
                                                                   H • HIGH SPEED
                                                                   L • LOW SPEED
                                                                   I  • IDLE
0
0.5
1.0
1.5
                                          2.0      2.5
                                        CO CUT- POINT, %
              Figure 5-12.  Variation of Ec,  Eo, and FF with CO Cut-Point;
               1974 Model Year Fleet; Key Mode Test; Garage Instruments;
               Bounded Errors of Commission Method
4.5

-------
U1
            70 •—
            60
            50
            40
         g  30
         o
            20
            10
     Lx     /
       \  /
H      ,X
\  /   \  /
  V        v
A   ,/\
*      \   /       '»
H      y        \



                                             \
                                             0
                                             FF
H = HIGH SPEED
L - LOW SPEED
I • IDLE
            5000     10.000    15.000   20.000    25.000    30.000    35.000    40.000
                                               CO CUT-POINT, ppm
                                                            45,000   50,000   55.000
                       Figure 5-13.  Variation of Ec,  E0.  and FF with CO Cut-Point;
                         1974 Model Year Fleet; Federal Three-Mode Test; Bounded
                         Errors of Commission Method

-------
Ul
I
            60
            50-
            40
          o30
            20
            10
             0
                    H'
                          L/
-     "I     L\
                                                N





H • HIGH SPEED
L - LOW SPEED
1 • IDLE
        o
        FF
0       0.5       1.0       1.5      2.0      2.5      3.0
                                CO CUT-POINT, %
                                                                            3.5
       4.0
                        Figure 5-14.  Variation of Ec, Eo,  and FF with CO Cut-Point;
                          1974 Model Year Fleet; Federal Three-Mode Test; Garage
                          Instruments; Bounded Errors of Commission Method
                                                                               4.5

-------
             60
             50
         fc  40
171
i
ro
S

S 20
             10
                                                        	FF
6000
7000
8000
                                        9000     10.000
                                           CO CUT-POINT,
                                                                   11.000
                                                                 ppm
12.000    13.000    14.000    13,000
                          Figure 5-15.   Variation of Ec,  Eo> and FF with CO Cut-Point;
                            1974 Model Year Fleet; Unloaded 2500  rpm Test; Bounded
                            Errors of Commission Method

-------
01
I
            o
            oc
                60 -
                50
                40
                 30
                 20
                 10
                                                                                    c
                                                                           	FF
                  0
                  0.5
0.6       0.7      0.8       0.9      1.0
                    CO CUT-POINT, %
1.1
1.2
1.3
                         Figure 5-16.  Variation of Ec, Eo,  and FF with CO Cut-Point;
                          1974 Model Year Fleet;  Unloaded 2500 rpm Test; Garage
                          Instruments; Bounded Errors of Commission Method

-------
Table 5-6.  Comparison of ST Carbon Monoxide Results:  1974
           Model Year Fleet,  Bounded Errors of Commission
           Analysis (E  = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
Idle
Low Speed
High Speed
Clayton Key Mode (Garage)
Idle
Low Speed
High Speed
Federal Three -Mode (Laboratory)
Idle
Low Speed
High Speed
Federal Three -Mode (Garage)
Idle
Low Speed
High Speed
2500 -rpm Unloaded (Laboratory)
2500 -rpm Unloaded (Garage)
Parameter, %
E
0
7
8

19
18
35

35
35
37

20
20
29

35
31
30
19
33
FF
65
64

53
54
38

38
38
35

53
52
43

37
41
42
53
40
Figure No.
5-9
5-10

5-11
5-11
5-11

5-12
5-12
5-12

5-13
5-13
5-13

5-14
5-14
5-14
5-15
5-16
                             5-28

-------
                     25 -
                     20
                                                FF < 5% ON ALL MODES
                      15

ro
                      10
H • HIGH SPEED
L • LOW SPEED
I  • IDLE
                                                     H L
500      1000      1500      2000
                NOX CUT-POINT, ppm
                                                                      2500
      3000
3500
                         Figure 5-17.  Variation of Ec,  Eo, and FF with NOX Cut-Point;
                           1974 Model Year Fleet; Key Mode Test; Bounded Errors of
                           Commission Method

-------
                      30
                      25
                   ht  20
                   o
u>
o
                   o
                   az
                      10
                       0
                          H
                          L
                          I
                     HIGH SPEED
                     LOW SPEED
                     IDLE
                                              FF < 5% ON ALL MODES
                                 500
1000
1500
2000
2500
3000
3500
                                                 NOX CUT-POINT, ppm
                         Figure 5-18.   Variation of EC, Eo,  and FF with NOX Cut-Point;
                          1974 Model Year Fleet; Federal Three-Mode Test; Bounded
                          Errors of Commission Method

-------
Table 5-7.  Comparison of ST NOX Results:  1974 Model Year
            Fleet, Bounded Errors of Commission Analysis
            (E  = constant = 5%)
Short Test
Federal Short Cycle
NY/NJ Composite
Clayton Key Mode (Laboratory)
Idle
Low Speed
High Speed
Federal Three -Mode (Laboratory)
Idle
Low Speed
High Speed
2500 -rpm Unloaded (Laboratory)
Parameter, %
E
0
14. 5
16.5

13.5
14
13.5

14
14
14
14
FF
3
1.5

<5
<5
<5

<5
<5
<5
4
Figure No.
5-1
5-2

5-17
5-17
5-17

5-18
5-18
5-18
5-7
                           5-31

-------
                            dp(y)       *0.o2
                             *y|y=2.6

With N = 147, Eq. (4-1) yields a standard deviation of 0.35 gm/mi.  At
v = 0.005,  the standard deviation is about 0.95 gm/mi where the cut-point
is 3. 2 gm/mi.  The standard deviation on CO (Figure 5-9) is estimated at
2. 4 gm/mi for the cut-point  at 21. 5 gm/mi.   For CO at v = 0.005, the cut-
point is 38 gm/mi and the standard deviation is approximately 4.9 gm/mi.
              These estimates  show that the  standard deviation is on the
order of 10 percent to 15 percent of the estimated cut-point.
5.2.2.1.5     Instrument Comparisons
              For comparing the instruments used in the test program, plots
of the type shown in Figures  5-19 through 5-32 are informative. Here per-
cent E  and percent FF have been plotted against percent E  for HC and CO
with each modal ST.  If a policy decision is given in terms of percent E
allowable,  then the percent FF and percent E  can be compared.  To illustrate, , ,
suppose percent E  is fixed at 3 percent. For CO on the low speed Key Mode
(Figure 5-22), the laboratory instruments (dashed lines) give 48 percent FF
and 24 percent E , while the garage instruments give 26 percent FF and
46 percent E .
              If policy is stipulated in terms of percent rejected by the ST,
then percent E  can be compared.  For the CO low-speed mode on the Key
Mode test, suppose ST percentage rejection is to be approximately 30 percent.
Then, for percent FF equal to 30 percent (percent E   equals 42 percent),
percent E  is 0.6 percent for the laboratory instruments and 3.9 percent
for the garage instruments.
5.2.2.1.6     Discussion of Results
              On the average, the  bag-type tests have lower E  and higher
FF for a fixed rate of E  than do the volumetric tests.   However, FF rates
                      c
in the 30 percent range can be achieved with any of the tests.  For a fixed
percent FF,  the percent E  is determined since the sum of FF and E  is the
                                    5-32

-------
OO
                 0
                                                                        GARAGE
                                                                        INSTRUMENT

                                                                        LABORATORY
                                                                        INSTRUMENT
                                                                MAXIMUM DIFFERENCE
                                                                Eo
                                                                FF
                                         10%
                                         10%
                   0
10
20
30
40
50
60
70
                                             PERCENT EQ AND FF
                       Figure 5-19-  Variation of Ec, Eo, and FF with Instrument Type;
                        HC; 1974 Model Year Fleet; Key Mode Test; High Speed Mode;
                        Bounded Errors of Commission Method

-------
en
i
o
UJ  o
o_  2
               0
                 0
                                                                    	  GARAGE
                                                                            INSTRUMENT

                                                                    	LABORATORY
                                                                            INSTRUMENT

                                                                    MAXIMUM DIFFERENCE

                                                                    E0<1%

                                                                    FF <
               10
20
30
40
50
60
                                          PERCENT EQ AND FF
70
                      Figure 5-20.   Variation of Ec, Eo, and FF with Instrument Type;
                       CO; 1974 Model Year Fleet; Key Mode Test; High Speed Mode;
                       Bounded Errors of Commission Method

-------
(Jl
I
(Jl
                  0
                                                                        GARAGE
                                                                        INSTRUMENT
                                                                        LABORATORY
                                                                        INSTRUMENT
                                   MAXIMUM  DIFFERENCE
                                   EQ ~ 10%
                                   FF ~ 10%
                    0
10
20
30
40
50
60
70
                                             PERCENT EQ AND FF
                      Figure 5-21.   Variation of Ec, Eo, and FF with Instrument Type;
                        HC; 1974 Model Year Fleet; Key Mode Test; Low Speed Mode;
                        Bounded Errors of Commission Method

-------
Ul
I
OJ
            0
                                                                     	 GARAGE
                                                                             INSTRUMENT
                                                                     	 LABORATORY
                                                                             INSTRUMENT
                                              MAXIMUM DIFFERENCE
                                              E -17%
                                               o
                                              FF~17%
              0
10
20
30
40
50
60
70
                                        PERCENT EQ AND FF
                      Figure 5-22.   Variation of Ec, Eo, and FF with Instrument Type;
                       CO; 1974 Model Year Fleet; Key Mode Test; Low Speed Mode;
                       Bounded Errors of Commission Method

-------
Ul
I
OO
o_   2
                 0
                                                                      GARAGE
                                                                      INSTRUMENT
                                                                      LABORATORY
                                                                      INSTRUMENT
                                                               MAXIMUM  DIFFERENCE
                                                               EQ~13%
                                                               FF ~ 13%
                   0
               10
20
30
40
50
60
70
                                            PERCENT EQ AND FF
                       Figure 5-23.   Variation of Ec,  Eo, and FF with Instrument Type;
                       HC; 1974 Model Year Fleet; Key Mode Test; Idle Mode; Bounded
                       Errors of Commission Method

-------
i
OJ
00
              0
                0
                                                                       	 GARAGE
                                                                               INSTRUMENT

                                                                       	LABORATORY
                                                                               INSTRUMENT

                                                                       MAXIMUM DIFFERENCE
                                                                       FF ~ 18%
10
20
30
40
50
60
70
                                          PERCENT EQ AND  FF
                      Figure 5-24.   Variation of Ec, E0,  and FF with Instrument Type;
                       CO; 1974 Model Year Fleet, Key Mode Test; Idle Mode; Bounded
                       Errors of Commission Method

-------
               UJ
               O
               UJ
               0.
U)
                    0
                                                                      GARAGE
                                                                      INSTRUMENT
                                                              	LABORATORY
                                                                      INSTRUMENT

                                                              MAXIMUM DIFFERENCE
                      0
10
20
30
40
50
60
70
                                                PERCENT EQ AND FF
                      Figure 5-25.   Variation of Ec, Eo, and FF with Instrument Type;
                        HC; 1974 Model Year Fleet; Federal Three-Mode Test; High
                        Speed Mode; Bounded Errors of Commission Method

-------
Lul
    0
                                     	 GARAGE
                                             INSTRUMENT
                                     	 LABORATORY
                                             INSTRUMENT

                                     MAXIMUM DIFFERENCE

                                     EQ~13%

                                     FF~13%
              10
20
30
40
50
60
70
                               PERCENT EQ AND FF
             Figure 5-26.   Variation of Ec, Eo»  and FF with Instrument Type;
              CO; 1974 Model Year Fleet; Federal Three-Mode Test; High
              Speed Mode; Bounded Errors of Commission Method

-------
0
                                                    GARAGE
                                                    INSTRUMENT
                                            	 LABORATORY
                                                    INSTRUMENT

                                            MAXIMUM DIFFERENCE
                                            E
                                             0
                                             FF ~ 9%
  0
10
20
30
40
50
60
70
                           PERCENT EQ AND  FF
      Figure 5-27.   Variation of Ec, Eo,  and FF with Instrument Type;
       HC; 1974 Model Year Fleet; Federal Three-Mode Test; Low
       Speed Mode; Bounded Errors of Commission Method

-------
Ul
I
*.
is)
               0
                                                                     	 GARAGE
                                                                              INSTRUMENT
                                                                     	LABORATORY
                                                                              INSTRUMENT
                                                                      MAXIMUM DIFFERENCE
                                                                      EQ~21%
                                                                      FF~21%
10
20
30
40
50
60
70
                                        PERCENT EQ AND FF
                      Figure 5-28.   Variation of Ec, Eo, and FF with Instrument Type;
                       CO; 1974 Model Year Fleet; Federal Three-Mode Test; Low
                       Speed Mode; Bounded Errors of Commission Method

-------
                 5r-
ui
i
oo
o
UJ
°-  2
                  0
                                                                     GARAGE
                                                                     INSTRUMENT
                                                                     LABORATORY
                                                                     INSTRUMENT
                                                             MAXIMUM DIFFERENCE
                                                             FF ~ 8%
                   0
               10
20
30
40
50
60
70
                                           PERCENT EQ AND FF
                      Figure 5-29.   Variation of Ec, Eo, and FF with Instrument Type;
                        HC; 1974 Model Year Fleet; Federal Three-Mode Test; Idle
                        Mode; Bounded Errors of Commission Method

-------
         o
         a:
Ul
i
•*»•
                                                               f
                                                                               GARAGE
                                                                               INSTRUMENT
                                                                               LABORATORY
                                                                               INSTRUMENT
                                                                        MAXIMUM DIFFERENCE
                                                                        EQ~17%
                                                                        FF~17%
                        10
20
30
40
50
60
70
                                         PERCENT EQ AND FF
                      Figure 5-30.   Variation of Ec, E0> and FF with Instrument Type;
                        CO; 1974 Model Year Fleet; Federal Three-Mode Test; Idle
                        Mode; Bounded Errors  of Commission Method

-------
Ul
*•
\jn
            o
            UJ
            0-
                 0
                                                                     GARAGE
                                                                     INSTRUMENT
                                                                     LABORATORY
                                                                     INSTRUMENT
                                                                 MAXIMUM DIFFERENCE
                                                                      10%
                                                                 FF ~ 10%
                                      Eo~
                   0
10
20
30
40
50
60
70
                                         PERCENT EQ AND FF
                      Figure 5-31.   Variation of Ec,  Eo, and FF with Instrument Type;
                        HC; 1974 Model Year Fleet; Unloaded 2500 rpm Test; Bounded
                        Errors of Commission Method

-------
         O
UI
I
                                                              I
                                                     GARAGE
                                                     INSTRUMENT

                                                     LABORATORY
                                                     INSTRUMENT
                                                                      MAXIMUM DIFFERENCE

                                                                      EQ -19%
                                                                      FF ~19%
               0
10
20
30
40
50
60
70
                                       PERCENT EQ AND FF
                      Figure 5-32.  Variation of Ec, E0i and FF with Instrument Type;
                       CO;  1974 Model Year Fleet; Unloaded 2500 rpm Test; Bounded
                       Errors of Commission Method

-------
FTP rejection rate.  Thus, the "best" test for fixed percent FF is the one
with the lowest percent E  .  In general,  the bag-type STs are better in this
respect.  However, the actual level of percent E  on the volumetric tests is
still quite low.  For example, at 30 percent FF on the CO  Federal Short
Cycle (Figure  5-9), the percent EC is essentially zero.  For CO on the Key
Mode low-speed mode, percent EC is 0. 65 percent for laboratory instruments
(Figure 5-11) and 3.85 percent for garage instruments (Figure 5-12).
5.2.2.2       Multiple-Constituent Tests
               In addition to analyzing each pollutant individually, an analysis
was made for multiple-constituent tests.  The method of analysis and com-
putational procedures were the same as for the  CEV fleet, as discussed in
Sec. 4.2.2.5.2.
5. 2. 2. 2. 1      Bounded Errors of Commission Results
               Three-constituent test results for the Federal Short Cycle
and the Federal Three-Mode (high-speed and idle modes only) are displayed
in Figures 5-33 through 5-35.  Both laboratory and garage instrument results
are included.   The data plotted are the parametric results.  For a detailed
discussion of the plot presentation, see Sec.  4. 2. 2. 5. 2. 1.
5. 2. 2. 2. 2     Variance Estimates
               Table 5-8 shows the approximate cell standard deviation for a
range of cell percentages, assuming N = 147.  See Sec. 4. 2. 2. 5. 2. 2 for a
detailed discussion of the approximation procedures.
5.2.2.2.3     Discussion of Results
               A comparison of modes on the Federal Three-Mode ST indi-
cates that the idle mode may be more favorable.  Using laboratory instru-
ments,  the idle mode has fewer errors of commission while maintaining a
superior percent FF and percent E  relation over the high-speed mode for
most of the range of predicted percent EC shown  in Figure 5-34.  Using
garage instruments (Figure 5-35), no statistical  difference  between the modes
is observed.
                                    5-47

-------
      8
  o
  a:
  <


  o
      0
                                                      I
I
      80i-
      60
0 tt
a: "-
o
<
      20
      0
                                    	FF
        o
                      PREDICTED PERCENTE
  Figure 5-33.   Variation of Actual Ec,  Eo, and FF with

   Predicted Ec; Federal Short Cycle; Three-Constituent

   Test; Bounded Errors of Commission Method; 1974

   Model Year Fleet
                           5-48

-------
    ,» 4
   o
   ce.
   tt
       0
             ACTUAL PERCENT Er - 0 FOR IDLE MODE
                           * c
      80
      60
      40
o
      20
       0
                                              	FF
                    HIGH SPEED
                                    1
         0
     2        3
PREDICTED PERCENT
  Figure 5-34.    Variation of Actual Ec, E0, and FF with
   Predicted Ec; Federal Three-Mode; Three-Constituent
   Test; Laboratory Instruments; Bounded Errors of Com-
   mission Method; 1974 Model Year Fleet
                          5-49

-------
    ACTUAL PERCENT EQ - 0 FOR  IDlf MODE
    ACTUAL PERCENT EQ = 0.68 FOR HIGH SPEED MODE ±0.67
                                 	FF
 801-
 60
O U_

of Q
—I <

l-<
O
<
-40
 20
  0
                       OHIGH SPEED

                       AIDlf
• HIGH SPEED
AIDL£
   0
                 1         2        3
                       PREDICTED PERCENT E
           Variation of Actual Ec,  Eo,  and FF with
   Figure 5-35.
    Predicted Ec; Federal Three-Mode; Three-Constituent
    Test; Garage Instruments; Bounded Errors of Commis-
    sion Method; 1974 Model Year Fleet
                     5-50

-------
          Table 5-8.  Standard Deviation for Three-Constituent
                      Tests: 1974 Model Year Fleet,  N = 147
Cell
Percentage
60
50
40
30
20
10
5
3.5
Cell
Standard
Deviation, %
4.04
4.12
4.04
3.78
3.30
2.45
1.80
1.52
              Comparison of the Federal Short Cycle and the Federal Three-
Mode can be made over a limited range of the results.  For the actual per-
cent E  less than 2 percent,  the laboratory results of the Federal Three-
      c
Mode and the Federal Short Cycle are comparable.  Table 5-9 indicates the
minimum and maximum for percent FF and percent EQ, while percent EC is
less than 2 percent. There is little difference between the idle  mode and the
Federal Short Cycle. Over this range of percent EC> the idle mode would
appear favorable to the  Federal Short Cycle due to the low value of percent
E  on the idle mode.
 Q
              A comparison of instrument  types shows that the laboratory
instruments are generally preferable.
                                   5-51

-------
Table 5-9.  ST Comparison: 1974 Model Year Fleet;
            Multiple Constituent Tests (E  S 2%)
Short Test
Federal Short Cycle
Federal Three -Mode:
Idle
High
% FF
Min
25

22
5
Max
36

38
42
%EO
Min
44

42
38
Max
55

58
75
                       5-52

-------
5.3           RELATIVE IMPACT ON AIR QUALITY
5.3.1         By Individual Pollutant
              The FTP standards, or cut-points,  can be interpreted as
establishing the desired impact on air  quality in that the FTP cut-points fix
the percent of the population classified as high- polluting vehicles.  If the
FTP were used  as the test procedure in an inspection/ maintenance program
which tested all vehicles (i. e. , as the  ST),  the  relative impact on  air quality
would ideally be 100 percent; that is, all the vehicles that are failures are
in fact identified as such.
              Similarly,  the effectiveness of the various STs can  also be
used as a measure of impact on air quality,  where "ST effectiveness" is
defined as :

             c_   ,,   ..          __ % FF for the  short test           .    .
             ST  effectiveness =
                                 FTP faiiures in same population
                                   % FF
                             ~ % FF + % E
                                          o

Thus,  on this basis,  the ST is always less effective than the FTP, in
proportion to the percent of errors of omission (E ) associated with a given
ST.  Table 5-10 shows the ST effectiveness values for the 1974 model year
fleet for an E  rate of 5 percent.  These  values indicate the relative impact
on air  quality of the ST as compared with the impact of the FTP on air quality,
for the E  conditions shown.
         c
               Actual benefit or impact is dependent upon the user's needs
and constraints.  One measure of benefit would be the tons of pollutant re-
moved from  the atmosphere on an annual basis in a given region by the use
of an ST in an inspection/maintenance program.  This can be approximated
by the  relationship:

        Tons removed = ST effectiveness X A pollutant  to be removed
                        in population X % population sampled          (5-2)
                                    5-53

-------
         Table 5-10.  Short Test Effectiveness; E  = 5%
                      1974 Model Year Fleet
Short Test
Federal Short Cycle
NY/NJ Composite
Key Mode
Laboratory
Garage
Federal Three -Mode
Laboratory
Garage
2500 rpm Unloaded
Laboratory
Garage
ST Effectiveness**'
HC
0.83
0.78

0.58
0.34

0.61
0.41

0.61
0.39
CO
0.90
0.88

0.76
0.51

0.72
0.48

0.73
0.47
NOX
0. 17
0.06

0.28


0.22


0.22

%FF
HC
34
32

24 (I)(b)
14 (L)

25 (I)
17 (I)

25
16
CO
65
64

55 (L)
37 (H)

52 (I)
35 (I)

53
34
NOX
3
1

5 (I)


4(H)


4

(a)
  ST Effectiveness =
  where
                          FF
                     FTP Fails
(b)
      FTP HC Fails = 41. 09%
      FTP CO Fails = 72. 35%
      FTP NOX Fails = 17. 8%

   I = idle mode
  L = low speed mode
  H = high speed mode
                              5-54

-------
       where
       ST effectiveness  . ^ r      £
                                     o
       and
       A pollutant to be removed in population = average value for the
                                                 population of HC, CO,
                                                 or NOX, in tons/year,
                                                 in excess of that per-
                                                 mitted by the FTP
                                                 standard; it is based
                                                 on the FTP failures
                                                 and  corresponding
                                                 emission values ob-
                                                 served in the popula-
                                                 tion, and vehicle-miles-
                                                 traveled characteristics
This relationship ignores those additional benefits likely to occur if the failed
vehicles were  repaired and achieved emission levels  below the FTP standards
after repair.
               Equation (5-2)  indicates areas of tradeoff that should be ex-
amined prior to the implementation of a specific inspection/maintenance
program.  Figure 5-36 depicts one aspect of such tradeoffs.   This figure is
an illustrative  plot of Eq. (5-2) for two different ST (Federal Short Cycle,
and Unloaded 2500 rpm with garage instruments)  as used for CO  emissions.
As indicated in Table 5-10, their effectiveness values are 0.90 and 0.47,
respectively; i. e. , as  compared with the CO discrimination capability of the
FTP procedure,  they are 90 and 47 percent  as effective as the FTP in iden-
tifying vehicles which fail the  FTP test on CO.  Thus, to achieve the  same
benefit in total CO pollutant removal, the percentage  of the population that
must be sampled by the Unloaded 2500 rpm ST is approximately double that
which must be  sampled with the Federal Short Cycle ST.  Alternatively stated,
for any given percent sampling of the population,  the  use of the Federal Short
Cycle ST  would result  in approximately double the amount of CO  removed.
               The complexity of program implementation can be measured
in annual  cost.   The  cost components would  include such items as annual
                                    5-55

-------
o
o
                             FEDERAL
                             SHORT
                             CYClf
                                             2500 rpm
                                             UNLOADED
             PERCENT VEHICLE POPULATION SAMPLED
    Figure 5-36.   Impact of Percent Population Sampled on CO
       Removed (Illustrative Example Only)
                            5-56

-------
operating expenses, maintenance expenses,  and amortized initial development
and installation expenses.  The ST requiring laboratory instrumentation would
have substantial initial procurement costs, and higher annual maintenance and
operating expenses than those using garage instruments.  The bag-type ST
requires more skilled personnel and a CVS station.  The bag ST and multi-
mode tests also require a dynamometer.   Thus, the ST can be ranked
according to cost as follows:
         •     Federal Short Cycle, NJ/NY  Composite
         •     Three-Mode volumetric with  laboratory instruments
         •     Three-Mode volumetric with  garage instruments
         •     2500 rpm  Unloaded with laboratory instruments
         •     2500 rpm  Unloaded with garage instruments
For those inspection/maintenance programs targeted to 100 percent inspection
of all vehicles,  the above ranking  of ST by cost would appear valid.  However,
if less than 100 percent inspection is envisioned for some reason,  then addi-
tional factors should be considered.  For example, the unit cost of a program
(per vehicle) would be expected to decrease  as the  percent of the population
sampled increases. Thus, in the  example of Figure 5-36,  if the program were
targeted to a defined level of CO removal, a cost-benefit analysis might be an
appropriate method to select the ST and the percentage sampled for minimum
cost purposes.  The type of constraint normally imposed on a tradeoff study
would typically be total annual cost; however, additional constraints on per-
cent E  or percent rejected (E  plus FF)  are also admissible under this
approach.  Other areas of consideration are effective sampling and site
selection, importance of the  pollution source as a function of geographic
location, social impact,  etc.
5.3.2         Multiple Constituent Tests
              Short test effectiveness is  also a useful measure of test quality
for the multiple-constituent test,  although the pollutant removal implications
of Eq.  (5-2) must apply on an individual pollutant basis.  Shown in Table 5-11
                                    5-57

-------
          Table 5-11.  Short Test Effectiveness Values for Multipl
                      Constituent Tests; 1974 Model Year Fleet
                                                           •Pie
                                                            la)
Short Test
Federal Short Cycle


Federal Three-Mode
(Laboratory Instruments)
Idle
High
Federal Three-Mode
(Garage Instruments)
Idle
High
ST Effectiveness
0.77
0.373
0.314

0.483
0. 568

0.330
0.374
Percent EC
Predicted(b)
5
0.05
0.01

5
5

5
5
Actual
8.84
2.04
0.68

0.00
2.72

0.00
0.69
(a)
(b)
FTP failures = 80%
Using bounded errors of commission method of analysis
are the effectiveness values for the Federal Short Cycle and the Federal
Three-Mode.  Comparison of the test-to-test effectiveness  values should,
of course,  be made at points where the actual percent E  is equal; however,
this can be only approximated with the existing  data.
              The technical favorability of the  Federal Short  Cycle  is
diminished when comparing on the  basis of equivalent percent E . Although
the Federal Short  Cycle effectiveness is 0.77 at actual percent E  equal to
8. 84, it is reduced to 0. 373 and 0. 314 for actual percent E  values of 2. 04
and 0.68, respectively.  However, as shown in Table 5-11, the effectiveness
values of the high-speed mode of the Federal Three-Mode ST  with laboratory
                                    5-58

-------
and garage instruments are 0. 568 (actual percent E  =  2. 72) and 0. 374 (actual
percent E  = 0. 69),  respectively.  Comparable effectiveness values for the
idle mode with laboratory and garage instruments are 0. 483 and 0. 330,  re-
spectively, both with actual percent E  equal to 0.  Thus, in the actual per-
cent E  range below approximately 3, the Federal Three-Mode ST with
garage instruments (idle  or high-speed  mode) is  essentially equivalent to the
Federal Short Cycle in effectiveness while the Federal Three-Mode ST with
laboratory instruments has a higher  effectiveness than the Federal Short
Cycle.
              Although the favorability of the laboratory instruments over
the garage instruments persists under this method of comparison,  considera-
tion of program complexity could bias test desirability in favor of the Federal
Three-Mode with garage  instruments.
                                    5-59

-------
6.  DEFECT DATA FROM CATALYST-EQUIPPED
         EXPERIMENTAL VEHICLE FLEET

-------
             6.  DEFECT DATA FROM CATALYST-EQUIPPED
                    EXPERIMENTAL VEHICLE FLEET
               Upon completion of the FTP and ST tests performed on the
CEV fleet as described in Sections 3 and 4,  95 defect tests were performed
on 5 of the vehicles from the 40-vehicle CEV fleet.
               The 95 defect tests simulated a wide variety of malfunctions
that could occur in typical passenger cars.  The general categories of
defects are defective ignition components, changes in ignition timing,  dwell,
and spark advance,  faulty carburetion, defective valves, clogged air filter,
and faulty emission control components.  The defects were introduced
individually and mixed.  The Appendix lists  the defect test runs on the five
cars.   These test data were analyzed to (a) determine the statistical char-
acter of the defect tests, and (b) to examine the ability of the STs  to detect
defective vehicles of this nature.  The results are discussed below.
6.1            STATISTICAL ANALYSIS OF DEFECT TESTS
               Listed in Table  6-1 are the estimated ST/FTP correlation
coefficients for the ungrouped defect data and the original 40-car catalyst-
equipped fleet (first good data only),  using the method defined in Sec.
4.2.1.1. The HC correlations are consistently higher, over 0.9, among
the defect data than the previous 40-car CEV fleet.  Addition of all defect
data to the original CEV fleet data will significantly distort the population
characteristics with regard to  HC.  CO and  NO  distortion will also occur,
                                              3C
although not as pronounced as with HC.
               This distortion is also evident when examining elementary-
statistics.  Table 6-2 compares  statistics on the FTP data for the two
groups. Clearly the data are different and need to be analyzed as distinct
groups since the proportion of  defect cars to normally operating cars in the
true population is unknown.
                                    6-1

-------
Table 6-1.  ST/FTP Correlation Coefficient Comparison:
             Defect Test Vehicles vs Original CEV Fleet
             (laboratory instruments)

Test
Federal
Short Cycle

NY/NJ
Composite

Key Mode







Federal
Three -Mode







2500 rpm
Unloaded


N
-------
        Table 6-2.  Elementary FTP Statistics:  Defect Test Vehicles
                   vs Original CEV Fleet (gm/mi)
Pollutant
HC
CO
NO
X
Defect
Mean
4.35
10.04
3.23
Standard
Deviation
6.00
11.81
1.42
Original
Mean
0.64
2.86
2.48
Standard
Deviation
0.54
1.52
0.59
               Many of the defect tests are either replications or produce
similar data.  The defect tests for each car were grouped according to
similarity of defect (see  Appendix, under the column denoted Group No.).
Group No. 1 is the baseline group and represents the normally operating
vehicle.   A test for a significant difference in the FTP average values of
the defect group and the base group was made for each defect group on
each car.  Defect groups that have no significant difference cannot be
statistically distinguished, on the basis of their FTP values,  from the
baseline group.  The defect group contains at least one test distinguishable
from normal operation, if there is a  significant difference.  The distinguish-
able defect groups were further analyzed  for similarity among themselves.
               The result of this analysis is a smaller  set of defect tests,
on each car, that  are statistically different from one another. These test
data are then taken to represent observations on independent vehicles.
Thus,  the 95 tests on 5 cars were reduced to approximately  24 defect test
observations representing 24 distinct vehicles each with a defect.  The
results of the analysis are shown in Table 6-3.
                                    6-3

-------
           Table 6-3.  Groups Distinguishable from Baseline
                        Operation:  Defect Test Fleet
             Di stingui shable
Car ID         Group No.
                                Description of Defect
  162
       4             Lean main fuel system
       6             EGR circuit reduced flow
       8             Valves defective (exhaust)
       9             Valves defective (intake)
Groups 4 and 9 are statistically similar for Car 162
  164
        6
        7
        8
        9
Inefficient catalyst
Inefficient catalyst and 10% misfire
Inefficient catalyst and 5% CO idle
Baseline after leaded fuel use
  165
        3             Early power circuit activation
        4             No secondary air injection
        6             Rich idle and 10% misfire
        7             No EGR and 6° timing advance
        8             Reduced secondary air and oversize fuel jets
Groups 4 and 8 are statistically similar for Car 165
 169
        2
        3
        8
        9
       10
Timing under-advanced
Timing over-advanced
Rich idle and no secondary air
Rich idle and PCV closed
Defective  spark plug
 170
        3             Rich idle 8% CO
        4             10% intermittent misfire
        5             3% intermittent misfire
        6             No EGR
        8             10% misfire and rich idle
        9             10% misfire and lean idle
       10             10% misfire and no EGR
       11             Rich idle and no EGR
       14             Rich idle and rich main
Groups 6 and 11 are statistically similar for Car 170
                                    6-4

-------
6.1.1         Data Selection Procedures
              The statistical procedure used to test for differences between
                                                ife
groups was a multivariate linear hypothesis test.    The likelihood ratio
statistic which has an equivalent F-statistic was used to make the test of
significance at the 95% level.  The analysis was conducted on the  FTP data,
as these are most representative of the true  state of the vehicle.  The
conclusion of this analysis is shown in Table 6-3.
              To establish a data base for further analysis, actual data
from the individual groups were selected according to the following rules:
         a.   One run (testing sequence) may be selected from each
              distinguishable group.   If distinguishable groups are
              similar,  only one run may be  selected from the
              similar groups.
         b.   Run preferences are:
              1.    More acceptable ST data
              2.    Less ambiguity in the  run
              3.    Lowest run number
              As the  assumption of independence of the observations is
crucial to  contingency table analysis, the 95  defect tests were statistically
pruned to 24 tests representing  24 independent defective vehicles.  These
data are considered to represent a population distinct from the  original
40-car population. Of these 24, 6 have no Federal Three-Mode (laboratory)
data and 5  have  no Key Mode  (laboratory) data.
*T. W. Anderson, An Introduction to Multivariate Statistical Analysis,
 John Wiley and Sons, Inc., New York (1958).
                                    6-5

-------
6.2           CONTINGENCY TABLE ANALYSIS OF DEFECT
              DATA
              The analysis proceeded in two stages.  The original CEV
fleet population was first analyzed,  using first good  data.  The analysis
method was the bounded errors of commission procedure, which established
the ST cut-points (see Sec. 4.2.2.1.2).  Percent E  was  varied from 10%
to 1% in  1% increments, with the addition of points at 0.5% and 0. 1%.
Immediately following analysis of the  original CEV fleet,  the defect popula-
tion was analyzed.   The contingency table  results were calculated for this
population,  using the cut-points previously determined from the original
CEV fleet population.  The computations were performed at each of the E
settings.  Thus the analysis  is merely an assessment of how well a  test
constructed using an unknown mix of normal and defect operation data will
perform on a population of defective vehicles known  to represent extreme
departures from normal operation.
              A summary of the analysis  on each constituent is given in
Table 6-4.  The ST cut-points were established for E  less than or  equal to
5%, and the FTP level was level I (HC = 0.41 gm/mi,  CO = 3.4 gm/mi,
NO  = 3. 1 gm/mi).
   3t
              Sample plots are shown in Figures 6-1 through 6-6 for the
Clayton Key Mode  (laboratory data).  Comparing Figures 6-1 and 6-2, which
represent the analysis for HC, at E  equal to 0. 1%,  the original fleet has
approximately 33% E , and 35% FF.  The  defect data show E  at 5%, E  at
8%, and FF at 66%. As the loci of Figure 6-2 are relatively flat, the defect
discrimination qualities of the Key Mode on HC appear virtually insensitive
to policy decisions of 10% E   or less.
              The results of three- and nine-constituent  tests  for the Key
Mode (laboratory)  are shown in Table 6-5.  These results are typical for
all the multi-constituent tests.
                                    6-6

-------
       Table 6-4.  Defect Analysis Comparison Summary:
                    Predicted Population [% E   = 5^a),
                    FTP Level I
-------
i
00
   70


   60



it  50
o

\40
LU
 ~O
LU
z  30

LU
fc  20


   10


    0
                                     H * HIGH SPEED
                                     L • LOW SPEED
                                       • IDLE
0    100    200    300    400    500    600    700
              HC CUT-POINT,  ppm




it
0
«c
"o
LU
0
LU
^_

LLJ
O
0£
LLJ
CL


80
70

60

50
40


30

20
10
0

- —.^JBO**,*^
1 H — — — L
—

r ii - HIPH Trrn
Fc L • LOW SPEED
fco 1 • IDLE
	 FF

—

—
i "i 1 r 	 1 	 1-
                                                                     100    200   300    400    500
                                                                              HC CUT-POINT, ppm
                                                                                              600    700
  Figure 6-1.   Variation of Ec,  EO, and FF
    with HC Cut-point; Original CEV Fleet;
    Key Mode Test; 1975  FTP Level = 0.41
    gm/mi; Bounded Errors of Commission
    Method                    ^
                                                             Figure 6-2.   Variation of Ec, Eo,  and FF
                                                               with HC Cut-point; Defect Tests Only; Key
                                                               Mode Test;  1975 FTP Level = 0.41 gm/mi;
                                                               Bounded Errors of Commission Method

-------
                      0
                 	FF
H = HIGH  SPEED
L = LOW SPEED
I  = IDLE
0
 50 ~


 40 -


 30
»


 20


 10
                                                          70
                                                          60
50
                                                          40
                                                          30
                                                       LU
                                                       O
                                                       a:

                                                       £  20
                                                          10
                                                                                     Ec
                                                                                     Eo
                                                                                     FF
                                    H = HIGH SPEED
                                    L = LOW SPEED
                                    I  = IDLE
                                                                   \
                                                                     \
                  400    600     800
                  CO CUT-POINT,  ppm
                                    1000    1200
                                   200    400    600    800
                                         CO CUT-POINT, ppm
                                   1000   1200
  Figure 6-3.   Variation of EC, E0, and FF
    with CO Cut-point; Original CEV Fleet;
    Key Mode Test;  1975 FTP Level = 3.4
    gm/mi; Bounded Errors of Commission
    Method
                                                      Figure 6-4.   Variation of EC, EQ,  and FF
                                                        with  CO Cut-point; Defect Tests Only; Key
                                                        Mode Test;  1975 FTP Level  =  3.4 gm/mi;
                                                        Bounded Errors of Commission Method

-------
I
»-*•
o
70


60


50
.  40
 o
   30


   20


   10


    0
                               c
                              Eo
                              FF
                                                                    H  • HIGH  SPEED
                                                                    L  = LOW SPEED
                                                                    I  = IDLE
    200    400    600     800    1000    1200    1400    1600    1800   2000   2200
                                NO  CUT-POINT, ppm
                                  A


Figure 6-5.  Variation of Ec,  Eo, and FF with NOX Cut-point; Original CEV
             Fleet; Key Mode Test; 1975 FTP Level = 3. 1 gm/mi; Bounded
             Errors of Commission Method

-------
   70
   60
   50
O
o:
   30
   20
   10
    0
                   	FF
                                                      H = HIGH SPEED

                                                      L = LOW SPEED

                                                      I  = IDLE
200    400    600    800    1000    1200   1400    1600    1800    2000

                            NO CUT- POINT,  ppm
                               A
                                                                            2200
 Figure 6-6.  Variation of Ec, Eo,  and FF with NOX Cut-point;  Defect Tests

               Only; Key Mode Test; 1975 FTP Level - 3. 1 gm/mi; Bounded

               Errors of Commission Method

-------
         Table 6-5.  Key Mode Composite Test*a' (laboratory data)
Test Type
Three- constituent:
High Speed
Low Speed
Idle
Nine - constituent
Original
CEV Fleet
% FF

27.5
22.5
60.0
62.5
%Ec

5.00
5.00
5.00
12.50
%Eo

37.5
42.5
5.00
2.50
Defect
Fleet
% FF

89.5
73.7
89.5
94.7
%tr
c

0
0
0
0
%TT*
Eo

10.5
26.3
10.5
5.26
  (a)
    % Ec < 5; FTP Level I (HC = 0.41 gm/mi, CO = 3.4 gm/mi,  NO  =
    3. 1 gm/mi)                                                  x
6.3
CONCLUSIONS
              A review of the typical results illustrates that the short tests
perform well at isolating a population of defective cars.  This is noted by
the general tendency for percent FF to increase and percent E   to decrease
in the defect population.  Although percent E  decreased for HC, this was
not generally true for CO and NO .
                                JL
              The sources of the errors of commission  and omission are
two-fold.  The first and usual source  is that of the test procedures, i.e.,
measurement errors.   The second source is due to mixing of defects.  An
observation was classified as a defective car if any component of this vehicle
was defective. Hence,  all the NO  data analyzed are not representative of
                                j£
NO defects, for example.  The multiple-constituent tests (which tend to
eliminate mixing errors), show a very high probability,  greater than 70%,
of detecting defect vehicles (note that  all the defective  cars  failed the FTP
at Level I).
              In conclusion,  the ST/FTP tracking of defective vehicles  is
very good.
                                  6-12

-------
         APPENDIX






DEFECT TEST DESCRIPTIONS

-------
              Table  A-l.   Defect Vehicle Test Schedule and Defect Description
Step
No.

1

2


3
4

5

6
7


8

9

10
-
11

12

Type of Defect

Baseline

Idle system lean


Baseline
Idle system low
rpm
Idle

Baseline
Lean main fuel
system

Baseline

Carburetor
power circuit
Baseline

EGR circuit
reduced flow
Baseline

Car Setup Procedure
Car 1754162
Check CO, timing, dwell, etc., and record. Perform one
baseline test on the vehicle.
Lean idle system to either 0. 5% CO before catalyst with
secondary air disconnected or 100-rpm drop lean from
lean best idle. Do not allow excessive misfire, however.
Return idle setting to original setting.
Decrease idle rpm 75 to 100 rpm while holding all other
parameters at manufacturer's specifications.
Decrease idle rpm by 150 rpm, providing misfire is not
encountered.
Return car to original setting.
Install main fuel jets that are two sizes (0. 002 in. )
smaller than original fuel jets. Fuel float level remains
same as with original jets.
Remove jets and reinstall original jets. Run one baseline
test.
Disable carburetor power circuit so that the vehicle
receives no power circuit operation.
Return vehicle to original condition by reactivating power
circuit.
Reduce EGR flow in EGR circuit by approximately 50% by
blocking EGR tube to carburetor baseplate.
Restore full EGR flow and return vehicle to original
condition.
Number of
Tests This
Step

1

2


0
2

1*

0
2


1

2

0

2

0

Group
No.

1

2



3

3


4


1

5



6



Olson Labs
Run
Numbe r

(A07752)1
A07905
A07947
A07961


A07974
A07984
A08125X


A08141
A08172

A08191

A08242
A08254


A08260
A08264X


1 = Baseline replicate
* = Tests that require temperature and flow measurements
X = Runs with temperature and flow measurements

-------
                    Table A-l.   Defect Vehicle Test Schedule and Defect Description (Continued)
ro
Step
No.

13
14
15






16




17


1

2



3
4


Type of Defect

Fuel pump low
Baseline
Valves defective






Valves defective
intake



Baseline


Baseline

Advanced basic
ignition timing


Baseline
Insufficient
secondary air

Car Setup Procedure
Car 1754162 (Continued)
Reduce fuel pump pressure by 25% and test vehicle once.
Restore full fuel pump pressure and run one baseline test.
Remove cylinder head from vehicle. Obtain one replace-
ment exhaust valve from a Ford dealer and cut a wedge
in the face of the valve which has an area removed cor-
responding to 5 to 10% of the total valve face area. In-
stall valve in the front cylinder and reinstall head. Main-
tain the same valve lash as for the original valve
removed.
Remove cylinder head and defective exhaust valve. Obtain
the corresponding intake valve for this vehicle and also
take a wedge of 5 to 10% of the total valve face from the
intake valve. Install the front cylinder. Install original
nondefective exhaust-valve.
Remove cylinder head and defective valve. Reinstall
original valve. Run one baseline test.
Car 2104164
Check CO, timing, dwell, etc., and record. Perform one
baseline test on the vehicle.
Using a distributor with vacuum and centrifugal advance
characteristics representative of the five cars under test,
advance the idle timing by 6° (not to exceed audible knock
during first large acceleration on FTP when engine is hot).
Return timing to original setting.
Modify the secondary air supply system (larger pulley,
air leak, etc.) so as to obtain approximately a 50% reduc-
tion in secondary air injection.
Number of
Tests This
Step

1*
1
2






2




1*


1

2



0
2


Group
No.

7
1
8






9




1


1

2




3


Olson Labs
Run
Number

A08278X
A08293
A08371
A08377





A08431
A08445



A08477X


(A07751)1
A07812
A07950
A07960



A07972
A07983

            1 = Baseline replicate
            * = Tests that require temperature and flow measurements
           X = Runs with temperature and flow measurements

-------
                    Table A-l.  Defect Vehicle Test Schedule and Defect Description  (Continued)
>

(JO
Step
No.

5
6


7
8


9

10
11






12


13




Type of Defect

Baseline
Over-rich main
fuel system

Baseline
High rpm idle


High rpm idle

Baseline
Inefficient
catalyst





Inefficient cata-
lyst plus inter-
mittent misfire
Inefficient cata-
lyst plus rich
idle


Car Setup Procedure
Car 2104164 (Continued)
Return secondary air injection system to normal.
Install main fuel jets that are three sizes (0.003 in.)
larger than original fuel jets, e. g. , 47F to 50F jet
sizes. Fuel float level remains as previously set.
Return main fuel jets to original size.
Increase engine idle speed by 150 rpm to approximately
800 rpm. All other parameters remain as at lower idle
speed.
Increase engine idle speed by 75 to 100 rpm to between
725 and 750 rpm.
Set all parameters to original baseline levels and test.
Drain the zero -lead fuel from the vehicle and refuel with
leaded regular gasoline. Operate the vehicle so as to
consume the tank of gasoline. Replenish the gasoline sup-
ply and test the vehicle once. Remove the leaded fuel and
replace with unleaded (30% of tank volume). Repeat the
test. Fill the vehicle with leaded fuel. Test again.
Note: The following tests contain two or more defects:
With the catalyst operating inefficiently, as in step No. 11,
introduce a 10% intermittent misfire rate and test on
leaded fuel.
Set idle CO at 5% (without secondary air). Ignition sys-
tem operating normally. Test using leaded fuel. Return
all components to normal and operate the car on unleaded
fuel at high loads and speed so as to reactivate the
catalyst.
Number of
Tests This
Step

0
2


0
1


2

1
3






2*


1*




Group
No.


4



5


5

1
6






7


8




Olson Labs
Run
Number


A07918
A08051


A08066


A08101X
A08110X
A08128
A08155
A08170
A08183




A08214X
A08231X

A08253X




            * = Tests that require temperature and flow measurements
            X = Runs with temperature and flow measurements

-------
        Table A-l.  Defect Vehicle Test Schedule and  Defect Description (Continued)
Step
No.

14







1

2


3
4



5
6

7
8



9
Type of Defect

Baseline







Baseline

Retarded tim-
ing (basic)

Baseline
Early power
circuit
activation

Baseline
No secondary
air injection
Baseline
Timing over-
advancing
(vacuum)

Baseline
Car Setup Procedure
Car 2104164 (Continued)
Test the car on unleaded fuel. If the emissions have
returned to the original baseline level, proceed with the
next step. If the emissions have not returned to "nor-
mal," operate for one additional tank of unleaded fuel.
If the emissions have still not normalized, the remainder
of this vehicle's tests will be performed on another
vehicle.
Car 2364165
Check CO, timing, dwell, etc. , and record. Perform one
baseline test on the vehicle.
Using a distributor with vacuum and centrifugal advance
characteristics representative of the five cars under
test, retard the idle timing by 6°.
Return car to original condition.
Search the Ford Motor Company parts specifications and
determine the power value part number that is designed to
"come in" soonest, i.e., about 10 in. Install this part in
the carburetor.
Return car to original condition.
Deactivate the secondary air injection system.

Return car to original condition.
Modify the vacuum advance mechanism so as to give early
advancing without impacting the maximum advance ob-
tained. Modify so as to obtain the same advance at 10 in.
as would normally be obtained at 15 in.
Return the car to original condition.
Number of
Tests This
Step

2







1

2


0
2



0
2*

0
2



1
Group
No.

9







1

2



3




4


5



1
Olson Labs
Run
Numbe r

A08259
A08279






(A07906)1
A07934
A07948
A07963


A08003
A08052



A08100X2
A08180X

A08193
A08215


A08230
1 = Baseline replicate
2 = No cat bed roll
* = Tests that require temperature and flow measurements
X = Runs with temperature and flow measurements

-------
       Table A-l.  Defect Vehicle Test Schedule and Defect Description (Continued)
Step
No.


10
11
1Z
13
14
15
16
17
18
Type of Defect


Rich idle plus
intermittent
misfire of
spark plugs
Baseline
EGR not work-
ing plus ignition
timing advanced
Baseline
Reduced flow
from secondary
air system plus
over -rich main
fuel system
Reduced second-
ary air flow plus
lean main fuel
system
Baseline
Retarded igni-
tion timing plus
high idle rpm
Baseline
Car Setup Procedure
Car 2364165 (Continued)
Note: The following tests contain two or more common
defects:
Richen idle system to either 5% CO before catalyst with
secondary air disconnected or 100 rpm drop rich from
lean best idle plus introduce intermittent misfire at a
10% misfire rate.
Return the car to original condition.
Deactivate EGR system plus advance the idle timing by 6°
(no audible knocks).
Return the car to original condition. Run one baseline
test.
Modify secondary air supply system to obtain approxi-
mately a 50% reduction in secondary air injection plus
install main fuel jets that are three sizes larger than
original fuel jets.
Remove oversize jets and install undersize jets (two sizes
smaller) and retest with reduced secondary air flow
(reduction same as step No. 14).
Return the car to original condition.
Increase idle by 100 rpm and retard idle basic timing
by 6°.
Return the car to original condition. Run one baseline
test.
Number of
Tests This
Step


1
0
2
1
2
1
0
1
1
Group
No.


6

7
1
8
9

10
1
Olson Labs
Run
Numbe r


A08240X

A08256
A08258
A08267
A08295
A08307
A08320

A08432
A08444
X = Runs with temperature and flow measurements

-------
        Table A-l.  Defect Vehicle Test Schedule and Defect Description (Continued)
Step
No.

1

2



3
4





5
6




7

8



9
10

Type of Defect

Baseline

Timing under -
advancing
(vacuum)

Baseline
Timing over-
ad vane ing
(centrifugal)



Baseline
Timing under -
advancing
(centrifugal)


Baseline

Vacuum line
leaking


Baseline
PCV valve
stuck closed
Car Setup Procedure
Car 2544169
Check CO, timing, dwell, etc., and record. Perform one
baseline test on the vehicle.
Modify the vacuum advance mechanism so as to give late
advancing without impacting the maximum advance ob-
tained. Modify so as to obtain the same advance at 10 in.
as would be obtained at 5 in.
Return car to original condition.
Modify the centrifugal advance mechanics so as to give
early advancing without impacting the vacuum advance
circuit and without increasing the maximum centrifugal
advance possible. Modify so as to obtain the same
advance at 1500 rpm (distributor) as would be obtained
at 2000 rpm normally.
Return car to original condition.
Modify the centrifugal advance mechanism so as to give
late advancing without impacting the vacuum advance cir-
cuit or the maximum amount of centrifugal advance.
Modify so as to obtain the same advance at 2000 rpm
(distributor) as would be obtained at 1500 rpm normally.
Return car to original condition. Perform one baseline
test.
Remove one of the non-emission control device vacuum
lines from the "Christmas tree." Meter if necessary to
prevent excessive lean misfire which could cause engine
stalling.
Return car to original condition.
Remove PCV valve and plug PCV line, allowing no positive
crankcase ventilation.
Number of
Tests This
Step

1

2



0
2





0
2




1

2



0
1

Group
No.

1

2




3






4




1

5




6

Olson Labs
Run
Number

(A07922)1
A07935
A07973
A07987



A08020
A08050





A08065
A08083



A08124

A08132
A08140



A08182

1 = Baseline replicate

-------
        Table A-l.  Defect Vehicle Test Schedule and Defect Description (Continued)
Step
No.

11

12

13


14



15



16




17


18

19

20
Type of Defect

PCV valve
stuck open
Baseline

Vacuum spark
disconnect not
working
Baseline



Idle system too
rich plus sec-
ondary air
disconnected
Idle system too
rich plus vac-
uum spark dis -
connect not
working
Idle system too
rich plus PCV
valve blocked
Baseline

One defective
sparkplug
Baseline
Car Setup Procedure
Car 2544169 (Continued)
Remove blockage in PCV line and reconnect with PCV
valve in circuit but locked open.
Return to original condition by reinstalling good PCV
valve .
If the vehicle is equipped with a vacuum spark disconnect
circuit, render it inoperative.

Restore VSD circuit and return to original condition.
Perform one baseline test.
Note: The following tests (steps 15 through 18) contain
two or more defects:
Richen idle system to 5% CO before catalyst with second-
ary air disconnected.


With idle CO at 5% CO, disconnect vacuum spark discon-
nect circuit (secondary air system in operation during
testing).


With idle CO at 5%, plug PCV system so that there is no
flow into the intake manifold.

Return vehicle to original condition. Perform one baseline
test.
Disconnect the high tension lead to one spark plug to simu-
latfe a bridged plug or failed lead.
Perform one baseline test.
Number of
Tests This
Step

1

0

Defect i


1



1*



Defect i




1*


1

1

1*
Group
No.

7



Olson Labs
Run
Number

A08192



lot available


1



8





A08217



A08241X



ot available




9


1

10

1




A08266X


A08294

A08321

A08357X
* = Tests that require temperature and flow measurements
X = Runs with temperature and flow measurements

-------
                     Table  A-l.   Defect Vehicle Test Schedule and Defect Description (Continued)
00
Step
No.

1

2


3

4
5

5A

6

7
8

9

10





Type of Defect

Baseline

Rich idle


Rich idle

Baseline
Intermittent
misfire
Intermittent
misfire
Intermittent
misfire
Baseline
No EGR

Baseline

Clogged air
filter




Car Setup Procedure
Car 1614170
Check CO at idle with secondary air disconnected
upstream of the catalyst. Reconnect secondary air.
Richen idle system to either 5% CO before catalyst with
secondary air disconnected or 100 rpm drop due to en-
richment from lean best idle. Reconnect secondary air.
Richen idle system to 8% CO before catalyst with second-
ary air disconnected. Reconnect secondary air.
Return idle mixture to original setting.
Introduce intermittent misfire (electronically short cylin-
ders at random) at 10% misfire rate.
Introduce intermittent misfire (electronically short cylin-
ders at random) at 10% misfire rate.
Introduce intermittent misfire at 3% misfire rate.

Return ignition system to original condition and setting.
Deactivate EGR system.

Set all parameters (CO, ignition, and EGR) to original
baseline values and test.
Using a new air filter element, mask 95% of its flow area
or sufficient to cause a 10-fold increase in Ap and then
test vehicle. Leave the open zone of the element in two
quadrants of the circumference. Ap to be read at 50-mph
Key Mode loading. (Ap to be measured across element
only — do not include Ap across air horn).
Number of
Tests This
Step

1

Z


1

0
1*

1*

Z*

0
Z

1

1*





Group
No.

1

Z


3


4

4

5


6

1

7





Olson Labs
Run
Number

(A07907)1
A07933
A07949
A0796Z

A08037


A08156X

A08190X

A08Z3ZX
A08478X2

A08243
A08255
A08257

A08265X





            1 = Baseline replicate
            2 = Run was made out of order; just prior to run A08504
            * = Tests that require temperature and flow measurements
            X = Runs with temperature and flow measurements

-------
                    Table A-l.  Defect Vehicle Test Schedule and Defect Description (Continued)
 i
vD
Step
No.

11




12

13



14



15


16

17


18



Type of Defect

Clogged air
filter



Baseline

Intermittent
misfire plus
idle system too
rich
Intermittent
misfire plus
idle system too
lean
Intermittent
misfire plus
EGR plugged
Baseline

Idle system too
rich plus EGR
not •working
Idle system too
rich plus igni-
tion timing
advanced
Car Setup Procedure
Car 1614170 (Continued)
Mask or otherwise chock the flow of air through the air
filter element so as to obtain a 5-fold increase in Ap
across the air filter at 50-mph Key Mode loading. (Ap
to be measured across element only — do not include Ap
across air horn).
Return the car to the original condition.
Note: The following tests contain two or more defects:
Introduce intermittent misfire at 10% misfire rate as in
step No. 5 plus richen up the idle system to 5% CO before
catalyst with secondary air disconnected.

Introduce intermittent misfire at 10% misfire rate as
in step No. 5 plus lean out the idle system to 0. 5% CO
(or lowest CO level possible without misfire) before
catalyst with secondary air disconnected.
Deactivate the EGR system plus introduce intermittent
misfire at 10% rate as in step No. 5.

Return the vehicle to original condition. Run one base-
line test.
Deactivate EGR system plus richen idle system to 5%
before catalyst with secondary air disconnected.

With 5% idle CO, advance basic idle timing 6°. EGR
system operating normally.


Number of
Tests This
Step

2




0

1*



1*



1


1

1


2*



Group
No.

7






8



9



10


1

11


12



Olson Labs
Run
Number

A08280
A08292





A08306X



A08319X



A08343


A08376

A08430


A08443X
A08446X


             * = Tests that require temperature and flow measurements
            X = Runs with temperature and flow measurements

-------
                    Table A-l.  Defect  Vehicle Test Schedule and Defect Description (Concluded)
Step
No.

19
20
11
Type of Defect

Idle system too
rich plus igni-
tion timing
retarded
Idle system too
rich plus main
fuel system too
rich
Baseline
Car Setup Procedure
Car 1614170 (Continued)
With 5% idle CO, retard basic idle timing by 6*.
Install main fuel jets that are three sizes too large as
per car No. Z, step 6 and set idle CO at 5% level with
secondary air disconnected.
Return the vehicle to original condition. Run one base-
line test.
Number of
Tests This
Step

1
1
1*
G roup
No.

13
14
1
Olson Labs
Run
Number

A08457
A08470
A08504X
I
N^,
O
            * = Tests that require temperature and flow measurements
            X = Runs with temperature and flow measurements

-------
                                 TECHNICAL REPORT DATA
                           (Please read Instructions on the reverse before completing)
 . REPORT NO.
 EPA-460/3-76-011
2.
                            3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE

 Federal Test Procedure and Short Test
 Correlation Analyses
                            5. REPORT DATE
                             April 1976
                            6. PERFORMING ORGANIZATION CODE
 7. AUTHOR(S)

 M. G. Hinton,  J.  C. Thacker, and W.  B. Lee
                            8. PERFORMING ORGANIZATION REPORT NO.

                             ATR-76(7353)-l
 9. PERFORMING ORGANIZATION NAME AND ADDRESS
 The Mobile Systems Group
 Environment & Energy Conservation Division
 The Aerospace Corporation
 El Segundo,  California 90245
                            10. PROGRAM ELEMENT NO.
                            11. CONTRACT/GRANT NO.

                              68-01-0417
 12. SPONSORING AGENCY NAME AND ADDRESS
  EPA Office of Air and Waste Management
  Office of Mobile Source Air Pollution Control
  Emission Control Technology Division
  Ann Arbor, Michigan 48105
                            13. TYPE OF REPORT AND PERIOD COVERED
                             Final
                            14. SPONSORING AGENCY CODE
 15. SUPPLEMENTARY NOTES
 16. ABSTRACT
 A series of statistical analyses was performed to determine the degree of "correla-
 lation" that exists between five specific short tests (STs) and the federal emission
 certification test procedure (FTP) for new vehicles.  This work was performed to
 determine  if "reasonable correlation with certification test procedures" exists; this
 is a condition precedent to the promulgation of regulations that impose the in-use
 warranty provisions of Sec. Z07 (b)  of the Clean Air Act of 1970 upon the motor
 vehicle manufacturers.
 The basis for the analyses was ST and  FTP test data from three vehicle fleets:
 (a) a catalyst-equipped experimental vehicle fleet (40 vehicles),  (b) an in-use 1974
 model year vehicle fleet (147 vehicles), and (c) a catalyst-equipped defect test fleet
 (5 vehicles). Each of the  vehicles in these fleets was tested by the FTP and the
 following STs:  (a) Federal Short Cycle, (b) New York/New Jersey (NY/NJ) Com-
 posite, (c) Clayton Key Mode, (d) Federal Three-Mode,  and (e) Unloaded 2500 rpm.
 Hydrocarbon (HC) and carbon monoxide (CO) measurements were recorded with
 both laboratory analyzers and garage-type instruments for most of the volumetric
 tests.  All oxides of nitrogen (NOX)  measurements were made with laboratory ana-
 lyzers.  Two different statistical analysis methods were used to assess  "correla-
 tion"--a conventional correlation analysis and a contingency table  analysis.
 17.
                              KEY WORDS AND DOCUMENT ANALYSIS
                 DESCRIPTORS
                                            b.lDENTIFIERS/OPEN ENDED TERMS
                                           COS AT I Field/Group
 Air Pollution
 Emission Testing
 Short Test Procedures
 Test Correlations
                Air Pollution Control
                Conventional Correlation
                   Analysis
                Contingency Table
                   Analysis
                Laboratory and Garage
                   Instruments
   13 B
   14 B
 8. DISTRIBUTION STATEMENT

 Unlimited
                19. SECURITY CLASS (ThisReport)
                Unclassified
21. NO. OF PAGES
    276
                20. SECURITY CLASS (Thispage)
                Unclassified
                                         22. PRICE
EPA Form 2220-1 (9-73)

-------