United States       Air and Radiation      EPA420-P-99-007
           Environmental Protection              March 1999
           Agency                    M6.IM.001
vvEPA     MOBILES Inspection/
           Maintenance Benefits
           Methodology for
           1981 through 1993 Model
           Year Light Vehicles

           DRAFT
                              > Printed on Recycled Paper

-------
                                                                        EPA420-P-99-007
                                                                              March 1999
                                                                               for
      1981

                                M6.IM.001

                                  DRAFT
                               Edward L, Glover
                                Dave Brzezinski

                        Assessment and Modeling Division
                            Office of Mobile Sources
                       U.S. Environmental Protection Agency
                                   NOTICE

   This technical report does not necessarily represent final EPA decisions or positions.
It is intended, to present technical analysis of issues using data which are currently available.
         The purpose in the release of such reports is to facilitate the exchange of
      technical information and to inform the public of technical developments which
       may form the basis for a final EPA decision, position, or regulatory action.

-------
                                                                  EPA420-P-99-007

                                       - Draft -
           MOBILE6 Inspection / Maintenance Benefits Methodology
                for 1981 through 1993 Model Year Light Vehicles
                            Report Number M6.IM.001

                             Last Revised February 22, 1999
                                   Edward L. Glover
                                    Dave Brzezinski

                       U.S. EPA Assessment and Modeling Division
                                        NOTICE

    This technical report does not necessarily represent final EPA decisions or positions. It is intended to present
    technical analysis of issues using data which are currently available. The purpose in the release of such
    reports is to facilitate the exchange of technical information and to inform the public of technical
    developments which may form the basis for a final EPA decision, position, or regulatory action.
M6IM001.WPD DRAFT                 1                          Mar 24, 1999

-------
                                                                   DRAFT

                               EXECUTIVE SUMMARY

                  List of Issues, Key Points, Assumptions and User Inputs
                             Regarding MOBILE6 I/M Credits
           The methodology described in this document (M6.IM.001) covers 1981-93 model
    year cars and light-duty trucks, and 1994 and 1995 model year cars and trucks which were
    not certified to Tierl or later standards.  It calculates separate I/M credits for running and
    start emissions.  I/M credits are based on a simple distribution model in which every
    vehicle in the fleet is either a high emitter (FTP emission greater than 2 times HC or NOx
    standards or 3 times CO standards) or a normal emitter.  The emission levels of the high
    and normal emitters are based on FTP data collected independently by EPA, AAMA and
    API as part of the organizations' in-use  vehicle emission assessment programs.  The
    frequency  and distribution of high and normal emitters in the fleet is based  on a large
    database of EVI240 data collected in Dayton, Ohio in 1996 and 1997. The basic emission
    levels used in the model are a function of vehicle mileage, vehicle technology,  and model
    year.

           The basic assumption behind I/M is that a fraction of the high emitters  in the fleet
    are identified and repaired down to lower emission levels during the I/M process.  This
    process reduces the average emission level of the fleet. It is modeled using a mathematical
    model which resembles  a 'sawtooth'.  The bottom of the "teeth" are the after repair
    emission levels immediately  following I/M, and the top of the "teeth" are the levels to
    which the  fleet deteriorates  after one periodic inspection cycle, or a six  month RSD /
    change of ownership cycle.

           MOBILE6 will allow various I/M scenarios to be modeled. Some of these are new
    to the MOBILE  model series.  The others have all been changed or revamped in a
    significant manner. MOBILE6 will allow for some new features.

    New Features:

    1.      Internal operation  - No external I/M credit files to attach to the main program for
           1981 and  later model year vehicles.

    2.      I/M credits given for the EVI240 test, the ASM tests, the Idle tests and OBD testing.

    3.      Custom user supplied cutpoints for EVI240 can now be entered directly in the
           program.  For example, the combination (1.5 g/mi HC, 55 g/mi CO, and 3.2 g/mi
           NOx) can be entered for an EVI240 scenario.

M6IM001.WPD DRAFT                  2                          Mar 24, 1999

-------
                                                                  DRAFT
    4.      Annual, Biennial, Triennial, and Change of Ownership I/M testing frequency can
           now be modeled.

    5.      Ability to model up to five different I/M programs simultaneously.

    6.      Remote Sensing of High emitters can now be modeled.

    7.      Ability to model  the exemption of the first "n" model years / ages in an I/M
           program. The "n" can be up to the first 20 model years / ages.

    8.      User input and default values for non-compliance with testing requirements, and
           cost waivers on failures can be specified.

    9.      I/M credits given for cost waivered vehicles.

    10.     Ability to model RSD Clean Screening and High Emitter Profiling exemptions from
           an I/M program.


    Development of Important Parameters

    1.      The I/M  methodology and associated parameters presented in this document are
           heavily based on two other EPA documents. These are "Determination of Running
           Emissions as a Function of Mileage for 1981-93 Model Year LDV and LDT
           Vehicles" - M6.EXH.001, and "Determination of Start Emissions as a Function of
           Mileage  and Soak Time  for 1981-93 Model  Year Light Duty Vehicles." -
           M6.STE.003. The reader is encouraged to obtain these documents from the EPA
           Web site and review them.

    2.      Grouping Parameters - Most of the grouping of the data was done by model year
           and technology groups.  Ported fuel injection (PFI) technology was split from
           throttle body injection (TBI) and carbureted technology. Model year groups were
           chosen based on engineering judgement regarding technology changes, or were
           grouped based on similar certification emission standards.

    3.      Basic emission rate and I/M analyses were done for both cars  and light trucks
           separately.  The same analysis approach was used for each vehicle type; however,
           different model year grouping were selected for cars and trucks because of the
           different certification standards which were in effect.
M6IM001.WPD DRAFT                 3                          Mar 24, 1999

-------
                                                                   DRAFT

    4.     Basic Emission Rates - FTP emission factor data comes from significant EPA and
           industry testing (3,000+ FTP tests), and was corrected for recruitment bias based
           on IM240 testing from Dayton, Ohio (211,000 IM240 tests).

    5.     Average emissions of Normals and Highs for start and running emissions - FTP
           data were used.

    6.     Identification Rate of High emitters - These are based on a sizeable database (900
           vehicles) which received both the FTP and EVI240 tests at an EPA contractor
           facility.

    7.     After I/M Repair Effects for running emissions  - These are based on thousands of
           EVI240 tests from Arizona on vehicles which were repaired to pass I/M.

    8.     After I/M Repair Effect for start emissions - These are based on FTP data collected
           by EPA.

    9.     Sawtooth Methodology - It is  from MOBILES.  It assumes that vehicles repaired
           as part of the I/M process deteriorate at the same rate as a fleet which does not have
           an I/M program.  However, unlike previous MOBILE models, the deterioration
           varies over the entire mileage range of 0 to 300,000 miles.

    10.    Waiver Repair Levels -  In MOBILE6, cost waivered I/M failures will get some
           repair benefit.  A value of a 20  percent reduction has been chosen. This value may
           change between draft and final versions, if real data provides another value.
           Stakeholders are encouraged to comment on this assumption, and provide any data
           or rational for an alternative default value.

    11.    High Emitter Non-Compliance Rate  -  Set to a  default value of  15 percent.
           MOBILE6 will offer users the ability to enter alternative values. This is a generous
           default which is based on extensive analysis of Arizona and Ohio I/M vehicles.
           The analysis suggested higher  rates (> 20 percent).  It also includes high emitters
           which do not show up for the  initial I/M test.  The fact that 15 percent has  been
           selected for use in the absence  of user input does not constitute a policy by EPA to
           allow the use of this value for SIP purposes. EPA will propose a policy  on this
           issue separately from this document.  Stakeholders are encouraged to comment on
           this assumption, and provide any data or rational for an alternative default value.

    12.    High Emitter Waiver Rate -  Selected to be 15 percent of failures,  and loosely
           based on analysis  of Arizona and Ohio  I/M vehicles. The user will also have the
           ability to enter an alternative value  into MOBILE6.  Also, the fact that 15 percent
M6IM001.WPD DRAFT                 4                          Mar 24, 1999

-------
                                                                  DRAFT

           has been selected for use in the absence of user input does not constitute a policy
           by EPA to allow the use of this value for SIP purposes. EPA will propose a policy
           on this issue separately from this  document.   Stakeholders are  encouraged to
           comment on this assumption, and provide any data or rational for an alternative
           default value.

    13.    Remote Sensing Parameters  - These are based on two reports published by EPA.
           One report was on RSD identification of high emitters and the other was on RSD
           clean screening effectiveness. RSD and Change of Ownership modeling is new to
           the MOBILE model series, and requires several new inputs. However, its impact
           is relatively minimal on the overall I/M credits or basic emission level rates.

    14.    Assume on average for the fleet that one RSD inspection to identify high emitters
           is done per year on each vehicle in the I/M program. Field experience with RSD
           suggests that this is an ambitious goal, and may require many vehicles to get dozens
           of RSD tests per year; however, programs which manage to test more frequently
           than this rate will not get additional credit. The user will also be allowed to enter
           a specific RSD fleet coverage fractions for RSD high emitter identification and
           RSD clean screening. The range  of these fractions will be from 0  percent to 100
           percent.

    15.    The default RSD or High Emitter Profile clean screening loss of credit is five
           percent.  However, the user of MOBILE6 is strongly encouraged to develop their
           own estimate and use it as an input to the model. Stakeholders are encouraged to
           comment on this assumption, and provide any data or rational for an alternative
           default value.

    16.    Change of Ownership - Data from Wisconsin suggests that roughly 16 percent of
           the testing annually is change of ownership testing. This translates into 8 percent
           every six months, and is built into the change of ownership "sawtooth" algorithm.

    17.    MOBILE6 will assume that the ASM tests will have the same relative performance
           to the EVI240 that they did in MOBILES. This is necessary because no new ASM
           I/M test data matched with FTP data are available since  MOBILES was released.
           New Idle and 2500RPM/Idle test data are available and new performance estimates
           have been computed, and will be installed in the MOBILE6 model.  The ASM and
           Idle I/M test performance in comparison to the EVI240 will be computed in the
           MOBILE6 model by adjusting the I/M test identification rate (IDR) factors.
M6IM001.WPD DRAFT                 5                         Mar 24, 1999

-------
                                                                   DRAFT

    18.    Both the ASM and Idle tests assume the same after I/M repair emission levels as the
           IM240 tests.  Only the IDR rates are changed. This assumption is currently under
           review for the Idle and Idle/2500 RPM tests. The most likely change will be to
           adopt the MOBILES repair effects for Idle tests rather than assume the Idle test has
           the same repair reduction as the EVI240 test.  The ASM test will continue to use the
           same repair effect as the EVI240 test.
    General Statement

           This document and the important parameters mentioned  in it are currently in
    DRAFT status, and will likely remain in that status until mid-1999. This document will
    also likely receive  some revision following peer review and stakeholder review. As a
    result, the I/M model, the basic emission rates, and the underlying parameters are all subj ect
    to possible future revision.    Comments regarding the modeling approach, important
    parameters and assumptions are encouraged from stakeholders and other interested parties.
M6IM001.WPD DRAFT                 6                         Mar 24, 1999

-------
                                                                  DRAFT

    1.0    INTRODUCTION

           This document describes EPA's new methodology for estimating exhaust emission
    Inspection / Maintenance (I/M) credits. This includes the methodology for various tests
    such as the IM240, the Idle test, the 2500 RPM/Idle test, and the ASM test. It includes the
    methodology used for all cars and light trucks for model years 1981 through 1993, and for
    non-Tier 1 cars and trucks for model years 1994 and 1995. The I/M credit methodology for
    the pre-1996 model year will also be used for 1996 and later model years which receive
    only exhaust I/M tests. This document also describes how credits and debits for the remote
    sensing device (RSD) testing will be incorporated into MOBILE6. The I/M credits for the
    pre-1981 model years are not being revised for MOBILE6.  The I/M credits for post-1995
    model years with OBD systems, and the evaporative emission  I/M test credits will be
    discussed in a separate documents "Determination of Emissions, OBD, and I/M Effects for
    Tierl, TLEV, LEV, and ULEV Vehicles" - EPA document M6.EXH.007, and "Inspection
    /  Maintenance Credits  for Evaporative  Control  System Tests"  -  EPA document
    M6.IM.003.

           MOBILE6 will handle I/M credits differently than previous MOBILE models. One
    major difference is the discontinuation of the TECHS model.  The TECHS model was a
    complex external FORTRAN program which calculated and exported the exact I/M credit
    values. These credit values were then built into the MOBILES block data code or read as
    an external file. The new credit methodology will instead be built into the MOBILE6 code,
    and will operate automatically every time an I/M program is called by the MOBILE6
    program.  This change will give the MOBILE6 user the ability to vary the effect of
    cutpoints and other program parameters through changes to the MOBILE6 input file.  No
    longer will it be necessary to  develop special I/M credits using  the TECHS model, and
    attach them to the MOBILE program.

           The new I/M credit methodology will also be updated to reflect the new basic
    emission rates (see "Determination of Running Emissions as a Function of Mileage for
    1981-1993 Model Year Light-Duty Vehicles-Report Number M6.EXH.001"). In addition
    to being lower in magnitude, the new emission rates separate start and running emissions.
    MOBILE6 will account for these emissions separately, and produce separate start and
    running I/M credits.

           This document is structured into five primary sections, and an Appendix section.
    Section 2 briefly describes the databases used in the analysis and development of the
    credits.  Section 3 describes the methodology for development of the running exhaust I/M
    credits based on the EVI240 test.  Section 4 describes the periodic I/M methodology -
    "sawtooth methodology", Section 5 describes the methodology for development of the start
    exhaust I/M credits based on the EVI240 test.  Section 6 describes the methodology for the


M6IM001.WPD DRAFT                 7                          Mar 24, 1999

-------
                                                                   DRAFT

    development of credits for the other types of I/M tests (Idle, 2500/Idle, and ASM).  The
    document also contains  an  extensive Appendix section which is listed A through H.
    Appendices A and C contain illustrative examples of the modeling approach. Appendix
    C contains some sample calculations.  Appendix D contains the programmer's explanation
    and adoption for coding purposes of the algorithm described in this document. Appendices
    E through H contain statistical diagnostics for many of the parameters used in this model.
    2.0    DATA
           Four databases were utilized to develop the EVI240 based credits. The first database
    was a large emission factor database which contained over 5,000 initial FTP tests on 1981
    through 1993 model year cars.  It was used in the I/M credit analysis to determine the
    average emissions of the "Normal" emitting vehicles and the "High" emitting vehicles.
    This is the same database which was used in generating the basic emission rates.  It is
    described  in greater detail in "Determination of Running Emissions as a Function  of
    Mileage for 1981-1993 Model Year Light-Duty Vehicles" - report number M6.EXH.001.

           The second database was a smaller I/M database.  It was used to determine the high
    emitter identification rates for the EVI240 test. It contained 910, 1981 and later cars and
    trucks which had both an EVI240 test and a running LA4 test (derived from the FTP test).
    It contained data from EPA emission factor testing in Ann Arbor, Indiana and Arizona in
    which vehicles were randomly recruited and tested on both the FTP test and the EVI240 test.

           This second vehicle emission database contains many of the same FTP / lane EVI240
    test pairs that  were used for the MOBILES I/M credits. In an attempt to update the
    MOBILE6 credits with newer model year data, additional vehicle data with FTP / lab
    EVI240 test pairs were added where FTP / lane EVI240 were not available. Use of a lab
    EVI240 versus a lane EVI240 for I/M credit purposes introduces some additional uncertainty
    in the analysis since a lab EVI240 test is less similar to an actual state conducted EVI240 I/M
    test than a lane EVI240. However, inclusion of the FTP / lab test data, enabled the analysis
    to include some post 1991 model year vehicles and additional light trucks rather than
    extrapolate these points.  Thus, EPA concluded that these benefits outweighed the slight
    increase in uncertainty caused by using lab EVI240 data.

           The third database was the Arizona EVI240 database from official state testing.  It
    contained several thousand before-and after-repair EVI240 tests, and was used to determine
    the repair effects for the running LA4 EVI240 credits. It contains data from a special test
    program that the State  of Arizona conducts on  a continuous basis to evaluate the
    performance of their I/M program. In this program, vehicles are randomly selected  to


M6IM001.WPD DRAFT                 8                          Mar 24, 1999

-------
                                                                  DRAFT

    receive the full IM240 test both initially, and if they fail, after all subsequent repair cycles.

           The fourth database of about 970 EPA tested vehicles contained both EVI240 and
    FTP data before and after repair.  It was used to calculate the effects of repair on start
    emissions.  It is documented in EPA document M6.EVI.002.
           The RSD credits and coverage parameters described in this document are based on
    extensive RSD testing at many locations. Details regarding the RSD data, and the analysis
    performed to determine the RSD credits can be found in EPA documents: "User Guide and
    Description for Interim Remote Sensing Program Credit Utility - EPA420-R-96-004", and
    "Draft Description and Documentation for Interim Vehicle Clean Screening Credit Utility -
    EPA420-P-98-008".
    3.0    I/M ALGORITHM FOR RUNNING EMISSIONS

    3.1     Definition of Categories

           The basic purpose of I/M is to identify and repair high emitting vehicles with broken
    emission control systems. These types of vehicles are termed "high" emitters, and typically
    have average emission levels which are considerably higher than the overall mean emission
    levels.  The remainder of the fleet is considered to be the "normal" emitters. These are low
    and average emitting vehicles, and their emission control systems are generally functioning
    properly. The overall fleet emission factor is assumed to be a weighted average of the high
    and normal emitters. For  comparison, the use of two emitter classes differs from the
    methodology used in the previous TECHS and MOBILES models. In those models, there
    were four emitter classifications (Normal, High, Very High, and Super).

           The MOBILE6 model will generate  specific I/M credits based on pollutant, model
    year group, and technology type. Credits for the three pollutants HC, CO, and NOx will be
    produced.  Also, credits for the 1981 through 1993 model years will be stratified into seven
    separate groups.  These are: 1988-93  (PFI), 1988-93  (TBI),   1983-87 (FI),  1986-89
    (CARB), 1983-85 (CARB), 1981-82 (FI), and 1981-82 (CARB).  PFI means ported fuel
    injection, TBI means throttle body fuel injection, (FI) means all closed-loop fuel injected,
    and (CARB) means closed-loop carbureted and all open-loop vehicles combined together.

    3.2     General I/M Algorithm

           Figure 1 is a general graphical view of the I/M algorithm for running emissions.
    Specific algorithms for each of the model  year / technology / pollutant groups will be


M6IM001.WPD DRAFT                 9                         Mar 24, 1999

-------
                                                                    DRAFT

    programed into the MOBILE6 model. Four lines are shown in Figure 1 which show the
    basic emission rate, the normal emitter emission rate, the high emitter emission level, and
    the after repair emission levels of the high emitters which were identified and repaired. The
    basic emission rate is shown as Line A. This line represents the average emissions of the
    fleet without an  I/M test.  It includes both the normal vehicles and the high emitting
    vehicles.

           Line B in Figure 1 represents the average emissions of the normal vehicles.  These
    are the vehicles which are very unlikely to fail any EVI240 test cutpoint in the range used
    by I/M programs, and should not require any significant emission related repair if they did
    fail.  The line is shown as a linear function of mileage to reflect the gradual deterioration
    that normal vehicles experience due to general wear.   In the data analysis these vehicles
    were defined as normal emitters for a specific pollutant if their FTP HC emissions were less
    than twice the applicable new car certification standard, or their FTP CO emissions were
    less  than three times the applicable new car certification standard, or their FTP NOx
    emissions were less than twice times the new car certification standard. In MOBILE6, it
    is assumed that these vehicles never fail I/M; no repair adjustment are made to them.

           Line C in Figure 1 represents the average emissions of the high vehicles. These are
    the vehicles which likely have "broken" emission control systems, and that should fail the
    EVI240 test cutpoint, and receive repair. In the data analysis these vehicles were defined as
    high emitters for a specific pollutant if their FTP HC emissions or FTP CO emissions
    exceeded twice or three times the applicable new car certification standard, respectively,
    or their FTP NOx emissions were two times the new car certification standard. Because
    high NOx emissions often  occur with low HC and/or low CO emissions, and sometimes
    even HC can be  high and CO normal, the three categories were kept separate.  Thus, a
    vehicle could be  a high HC emitter, but a normal CO and NOx emitter.

           The selection of twice or thrice FTP certification standards for the boundary level
    between normals and highs is an engineering choice based on the literature on I/M and
    repair.  Other reasonable boundary levels could also have been chosen. No formal analysis
    was  done to prove that these levels were optimum.  One of the reasons they were chosen
    is because they were used in MOBILES, and have generally been shown in the past to be
    a good dividing point between high emitting broken vehicles and lower emitting vehicles
    which  are not broken. Simple  statistical  analysis done on the data indicate that the two
    means  are statistically different.

           Line D represents the average emissions of the portion of high emitting vehicles that
    are identified and repaired because of the I/M process. This line is calculated as a function
    of vehicle age, and is a percentage (e.g., 150%) of Line B. The portion of the fleet which
    is identified by I/M will be repaired to a lower level on average. However, this level is not
M6IMOO1.WPD DRAFT                  10                         Mar 24, 1999

-------
                                                            DRAFT

    as low on average as the average of the normal vehicles.  The justification for this
    assumption was an analysis of Arizona EVI240 before and after repair data collected during
    1995 and 1996. (See EPA report EPA-420-R-97-001 "Analysis of the Arizona EVI240 Test
    Program and Comparison with the TECHS Model" for a description of this dataset).
    3.3    Calculation of Basic Running LA4 Emission Rates

          Line A in Figure  1  represents the basic non-I/M emission rate for a given
    combination of vehicle type / pollutant / model year group / technology group. The units
    represented in Figure 1 are running LA4 emissions in grams / mile.  The calculation
                                    FIGURE 1
                    GENERAL I/M CREDITS SCHEMATIC
             LA4
          EMISSIONS
                                               ------  D
                                     MILEAGE
M6IM001.WPD DRAFT
11
Mar 24, 1999

-------
                                                                   DRAFT

    methodology and databases used to determine these emission rates are fully documented
    in the report "Determination of Running Emissions as a Function of Mileage for 1981-1993
    Model Year Light-Duty Vehicles,"  report M6.EXH.001.  The reader is encouraged to
    review this  document  for more details.   Selected emission rates were  taken from
    M6.EXH.001 and used in this current report as examples.
    3.4    Calculation of Running LA4 Emission Rates for Normal Emitters

           Line B in Figure 1 represents the average emission rates for Normal emitters. These
    are the low emitting vehicles in the fleet which should not fail an I/M program. Line B was
    calculated by least squares regression of the emissions  of the normal emitters versus
    mileage in the FTP dataset. Sample sizes were satisfactory in all cases. The regression was
    done for each pollutant / model year / technology group.  The regression coefficients for
    cars are shown in Table la and light trucks in Table Ib.  The column labeled ZML contains
    the zero mile coefficients, and the column DET contains the deterioration coefficients
    (slope) from the regressions (units are grams per mile per IK miles). A sample scatterplot
    of the car data and the regression line is shown in Figure A-l through A-3 in Appendix A.
Table la
Regression Coefficients for RUNNING LA4 Emissions from Normal Emitter Cars
MY
Group

1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group

PFI
TBI
FI
Carb
Carb
FI
Carb
HC Coefficients
ZML
0.0214
0.0042
0.0942
0.0774
0.1266
0.0970
0.1539
DET
0.001385
0.001701
0.001439
0.000812
0.001214
0.002250
0.001271
CO Coefficients
ZML
0.4588
0.0000
1.4448
0.5666
0.7276
1.5762
1.3932
DET
0.02293
0.01990
0.01959
0.01371
0.01691
0.02150
0.01389
NOx Coefficients
ZML
0.2006
0.2253
0.4798
0.4960
0.5555
0.4597
0.5834
DET
0.00376
0.00381
0.00188
0.00170
0.00273
0.00633
0.00233
M6IM001.WPD DRAFT
12
Mar 24, 1999

-------
                                                                    DRAFT
Table Ib
Regression Coefficients for RUNNING LA4 Emissions from
Normal Emitter Light Trucks
MY
Group

1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group

PFI
TBI
FI
Carb
Carb
HC Coefficients
ZML
0.02989
0.04664
0.13384
0.26835
0.49182
DET
0.002376
0.002998
0.003280
0.002701
0.006485
CO Coefficients
ZML
0.4927
0.7663
1.6222
1.3553
7.4202
DET
0.02678
0.03442
0.04311
0.06660
0.03293
NOx Coefficients
ZML
0.3024
0.3150
0.3150
1.2872
1.6159
DET
0.003904
0.003171
0.003171
0.0001
0.000025
    3.5    Calculation of Running LA4 Emission Rates for High Emitters

           Line C in Figure 1 represents the average emission rates for High emitters.  These
    are the vehicles in the fleet which likely have problems with their emission control systems,
    and have emission levels which are considerably higher than the vehicles which do not have
    problems. In the analysis they were defined as those vehicles exceeding either twice FTP
    standards for HC or three times FTP standards for CO  or twice NOx standards. The line
    is a flat horizontal line because the emissions of a high emitter is not likely to be a strong
    function of mileage. Regressions of the high emitter emissions versus mileage were done.
    However, the relatively small sample sizes of high emitters make regression determined
    mileage coefficients unreliable indicators of actual behavior. Various  analyzes of failing
    cars in I/M programs also support the  use of a flat emission rate for high emitters.

           Instead, on many new vehicles, if something goes seriously wrong with the emission
    control system that is likely to immediately lead to high emissions, it is likely to be fairly
    random in occurrence (i.e., not mechanical wear in the carburetor that creates large numbers
    of high emitters over time, or built-in obsolescence at a particular mileage).  However, one
    weakness of this simplified approach is that a certain percentage (extremely small) of the
    brand new vehicles will be modeled as being high emitters.  This result occurs because at
    zero miles, the regression developed estimate of normal emitter's emission level is below
    the FTP and Ohio data developed estimate of the corresponding mean fleet emission level.
M6IM001.WPD DRAFT
13
Mar 24, 1999

-------
                                                                   DRAFT

           Table 2a shows the average emissions of the high emitters (cars only) for the 21
    pollutant / model year / tech groups.  Trucks are shown in Table 2b. Because of the small
    sample size of high emitters in most groups, some model year / technology groups were
    combined into another model year group, and an overall mean was computed for the
    combined group. For the cars and for each pollutant, the 1986-89 Carb and the  1983-85
    Carb were combined and averaged together. Likewise the 1981-82 Carb and 1981-82 FI
    Car groups were combined and the emissions from  the  high emitters were averaged
    together. For the trucks, all of the fuel injected trucks were combined together and a
    common mean  high  emitter emission level  was  computed for each pollutant.  This
    combination had the effect of producing more consistent means across groups. The high
    emitter HC emission level for the 1988-93 MY PFI group is also a special case.  For this
    group it was thought that the average high emitter emission level was too low because it
    caused the average high emitter level to be lower than the normal emitter level at fairly low
    mileages. It was increased from 1.10 g/mi HC to 1.74 g/mi HC by adding one very high
    emitting 1987 model year vehicle to the 1988-93 model year PFI group.

           The impact of this approach of averaging between groups and adding  selected
    vehicles to particular groups is that some high emitting vehicles contribute to the average
    high emitter level of their own model year group, and to another model year group.  This
    does not affect the non-I/M running emission estimates because the normal and high emitter
    split is not used to calculate the non-I/M estimates.  However, it does affect the  I/M
    emission rate and I/M benefits because it changes the portion of a particular model year
    group's emission  distribution between normals and  highs.   This  changed emission
    distribution will affect the fraction of fleet emissions in MOBILE6 which are identified and
    repaired by  I/M.  It is difficult to predict the size of the emission impact  because it
    simultaneously increases the average high emitter average, but decreases the fraction of
    high emitters in the fleet.  This change will also impact the start emissions and the start I/M
    credits because it changes the fraction of high start  emitters in the fleet (fraction of start
    high emitters is equal to the fraction of running LA4 high emitters), but does not affect the
    average start high emitter level.

           An analysis of the Ohio EVI240 data was also  done to try and estimate the high
    emitter levels for running LA4 and start emissions.  This was done because of the small
    numbers of  high emitters in the EPA and AAMA FTP (running LA4 and Start) data
    samples. In this analysis, a large sample of Ohio vehicles were segregated into normal and
    high emitters, and the average high emitter emission levels were determined and compared
    with the FTP based estimates.  They compared favorably. However, the analysis was
    plagued with uncertainties such as how to separate the normals from the highs when FTP
    data are not available, the inability to split PFI from TBI in  the Ohio EVI240 data, a
    questionable transformation of EVI240 results into running LA4 and start emissions, and
    unknown and possibly inconsistent conditions between lab testing and EVI240 lane testing.
M6IMOO1.WPD DRAFT                 14                         Mar 24, 1999

-------
                                                              DRAFT

    Because of these problems the Ohio IM240 data were not used to estimate the average high
    emitter emission levels.
Table 2a
Mean RUNNING Emissions of High Emitter Cars
MY Group
1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group
PFI
TBI
FI
Carb
Carb
FI
Carb
HC Mean
1.740
3.394
2.372
1.845
1.845
2.372
2.372
CO Mean
36.106
46.527
37.933
27.653
27.653
37.933
37.933
NOx Mean
2.846
2.872
2.951
2.872
2.872
2.951
2.951
Table 2b
Mean RUNNING Emissions of High Emitter Light Trucks
MY Group
1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group
PFI
TBI
FI
Carb
Carb
HC Mean
2.120
3.241
2.446
2.012
3.710
CO Mean
33.283
33.283
43.870
39.415
80.726
NOx Mean
2.846
2.846
2.846
4.988
5.014
M6IM001.WPD DRAFT
15
Mar 24, 1999

-------
                                                                    DRAFT

    3.6    Calculation of After Repair Percentages and Emission Levels

           Line D in Figure 1 represents the average after repair emission level of high emitters
    that are properly identified and repaired. In comparison, Line C represents those high
    emitting vehicles that are not identified and repaired properly, or belong to owners who
    evade the program after failing the initial test.  Line D is calculated by scaling up the
    normal emitter  emission level (Line B) using a multiplicative factor process which is a
    function of age, pollutant and cutpoint level. The normal emitter emission level is the basis
    for the after repair emission level, and is the lowest emission level to which high emitting
    vehicles can be repaired after adjustment for age and mileage.  This assumes that the I/M
    process on average does not turn aged vehicles into brand new ones. However, the process
    will allow an I/M program to claim full credit for fixing vehicles with definitive problems
    such as a bad oxygen sensor.


    3.6.1   After I/M Repair Multiplicative Adjustment Factor

           The after I/M repair multiplicative adjustment factor is a function of vehicle age and
    I/M cutpoint. It is calculated using a two step process. The first step is to calculate the
    multiplicative adjustment factor for the standard set of EVI240 cutpoints which the State of
    Arizona used in its EVI240 program.  These are the phase-in cutpoints of 1.2 g/mi HC / 20
    g/mi CO and 3.0 g/mi NOx. The second step involves computing and applying another
    ratio which is a  function of EVI240 cutpoint. It will allow the MOBILE6 program to assign
    a different after repair emission level as a function of EVI240 cutpoint. The combined after
    I/M repair multiplicative adjustment factor is multiplied by the normal emitter emission
    level to calculate the after repair emission levels.

    Phase-in Cutpoints

           Equations 1 through 3 are the multiplicative adjustment factors used to calculate the
    after repair emission level for HC, CO and NOx under phase-in cutpoints.  They were
    calculated from a large sample of Arizona EVI240 data.  The same coefficients are used for
    both cars and light trucks. The percent after repair I/M emission levels for the high emitters
    which were identified by I/M and repaired were developed by: (1) Stratifying the  sample
    by age into 15 groups (ages 1 through 15); (2) Computing for each age group the average
    emission level of the vehicles passing their initial Arizona I/M test; (3) Computing for each
    age group the  after  repair passing emission values of the  Arizona I/M failures; (4)
    Computing for each age group the ratio of the emissions of the repaired high emitters over
    the emissions of the initial passing vehicles;  (5) Regressing the ratios versus age for each
    of the three pollutants to produce Equations 1 through 3.
M6IMOO1.WPD DRAFT                  16                         Mar 24, 1999

-------
                                                                    DRAFT

           Equations 1 through 3 are used to produce Line D for the phase-in outpoints
    (1.2/20/3.0) by following the two steps.
           First, Line D is calculated as a percentage of Line B using Equations 1 through 3.

           HC ratio      =     2.2400 - 0.07595 * (vehicle age)                   Eqn 1
           CO ratio      =     2.1582 - 0.07825 * (vehicle age)                   Eqn 2
           NOx ratio     =     1.6410 - 0.04348 * (vehicle age)                   Eqn 3
           In these equations, vehicle age ranges between 1 and 15 years, and the percentage
    value at 15 years is used for all ages greater than 15.

           Second, the percentage values calculated in Eqns 1 through 3 (i percentage in Eqn
    4) are transformed into emission units by multiplying the percentage values by the emission
    values in Line B (average emission of the normal emitters) using Eqn 4. The emission level
    of the Normals is a function of mileage.

    After repair emissions pollutant i = i percentage * Emissions of Normals         Eqn 4
    Other Cutpoint Combinations

           Equations 1 through 4 are used to produce the after repair emission levels for an
    EVI240 program  which uses the phase-in cutpoints of 1.2/20/3 for HC, CO, and NOx
    respectively. Another adjustment factor is used to compute after repair emission levels for
    other cutpoints. It is a multiplicative factor which proportionally increases or decreases the
    after repair emission level computed for the 1.2/20/3 phase-in cutpoints to account for
    tighter or looser cutpoints.

           The factor used to compute the after repair emission level for cutpoints other than
    1.2/20/3 phase-in cutpoints is based on a limited amount of vehicle repair data collected
    by EPA in past testing programs. It was utilized to overcome the limitation of repair data
    collected at only one set of cutpoints in Arizona. This dataset was the same one used to
    develop MOBILES repair effects and technician training I/M credits.  The repair effects
    dataset which  was used consists of 273  vehicles from model years 1981 through 1992
    tested by an EPA contractor in South Bend, Indiana and at the EPA lab in Ann Arbor, MI.
    All of these vehicles had before and after repair EVI240 and FTP tests.  The sample of
    vehicles were  repaired to various FTP emission level targets.  None of the after repair
    results included a catalyst replacement.
M6IMOO1.WPD DRAFT                  17                         Mar 24, 1999

-------
                                                                   DRAFT
           The principal goal of the data analysis was to determine as a function of IM240
    cutpoint, the FTP after repair emission levels of vehicles which initially failed the EVI240
    tests and were repaired to pass the EVI240 test. For MOBILES, this analysis was done for
    seven  different HC/CO cutpoint  combinations and for five  NOx cutpoints.   These
    combinations are repeated in this document because they are the only after repair FTP data
    for a variety of cutpoints which currently exists.  These cutpoint combinations are shown
    in Tables 2c and 2d. Also, shown in Tables 2c and 2d are the after repair emission levels
    for each cutpoint combination group,  and the ratio of a given after repair emission level to
    the after repair emission level at 1.20 g/mi HC / 20 g/mi CO.  For NOx, the individual
    cutpoint groups are ratioed to the 3.0 g/mi NOx group.
M6IMOO1.WPD DRAFT                  18                         Mar 24, 1999

-------
                                                                DRAFT
Table 2c
FTP After Repair HC and CO Emission Levels and Ratios
versus EVI240 HC/CO Cutpoint Combination
HC Cutpt
(g/mi)
1.2
0.8
0.6
0.6
0.6
0.4
0.4
CO Cutpt
(g/mi)
20
15
15
12
10
10
15
After
Repair
HC (g/mi)
1.26
1
0.88
0.87
0.86
0.78
0.74
After
Repair
CO (g/mi)
13.46
11.85
11.94
11.15
10.50
11.30
11.71
HC Ratio
1.00
0.79
0.70
0.69
0.68
0.62
0.59
CO Ratio
1.00
0.88
0.89
0.83
0.78
0.84
0.87

Table 2d
FTP After Repair NOx Emission Levels and Ratios
Versus NOx EVI240 Cutpoint
NOx Cutpt
(g/mi)
1
1.5
2
2.5
3.0






After Repair
NOx (g/mi)
0.91
1.22
1.48
1.68
1.86






NOx Ratio
0.489
0.656
0.796
0.903
1.000






          For MOBILE6, the ratios data in Tables 2c and 2d were regressed versus HC, CO
    and NOx outpoint to produce an after repair emission level ratio for any HC, CO or NOx
    cutpoint (within the range allowed by MOBILE6) which the user may enter in MOBILE6
    (the MOBILE6 user is no longer restricted to a set of seven cutpoint combinations). A least
M6IM001.WPD DRAFT
19
Mar 24, 1999

-------
                                                                   DRAFT

    squares linear regression was used to produce the relationships for both HC/CO and NOx.
    The regression coefficients are shown in Table 2e.  The equation form for the HC Ratio
    and the CO Ratio are:
           Ratio  =
    For NOx it is:
           Ratio  =
A*HCCut + B*COCut
B * NOCut + C
           Eqn3b
           Eqn3c
    A linear regression was used instead of some other functional form because it produced
    high r-squared values (0.99 for HC and NOx and 0.95 for CO).  Also, note that the highest
    IM240 cutpoint for HC and CO are 1.2 and 20 g/mi. Repair effects at cutpoints higher than
    these will be linear extrapolation.
Table 2e
Regression Coefficients for Repair Effects Ratios
Ratio
HC Ratio
CO Ratio
NOx Ratio
A
0.4990
0.0249

B
-l.Olle-04
0.0168
0.2538
C
0.398
0.620
0.2613
rA2
0.996
0.950
0.993
    3.6.2   Application of the After Repair Adjustment Factors

           The ratio equations are used in MOBILE6 to compute the after repair emission
    levels for cutpoints which are different from the standard 1.2 / 20 / 2.0 cutpoints used by
    Arizona. This is done by multiplying Equations 1 or 2 or 3 by Equation 3b or 3c to produce
    the repair effects ratio for the non standard (1.2/20/2.0) cutpoint.  The final repair level is
    obtained by multiplying this ratio by the appropriate normal  emitter emission level line
    (Line B). The normal emitter emission level is used as the final after repair emission level
    if it is larger than the calculated after repair emission.

           The following example calculation of the after repair HC emission level for an
    HC/CO cutpoint  combination of 0.80g/mi HC and 15 g/mi CO is shown below for clarity.
M6IM001.WPD DRAFT
                20
Mar 24, 1999

-------
                                                                    DRAFT

    Aft Repair HC = (2.24-0.07595*age) * (0.4990*0.8g/mi - 1.01e-04*15.0g/mi + 0.398) * Norm_ave

    where

    Norm_ave is the average emissions of the normal emitters.  It is a function of mileage and
    technology/model year group. For an eight year old 1990 PFI vehicle at 100,000 miles it
    is:  0.0214 + 0.001385 * 100 = 0.159 g/mi Running HC.

    0.8g/mi HC is the HC cutpoint; 15.0g/mi is the CO cutpoint.

           Substituting the value of 0.159 g/mi and 8 years old  into the After Repair HC
    equation produces an after repair emission level ofO.206 g/mi running HC at a cutpoint of
    0.80 g/mi  HC and 15 g/mi CO for an eight year old vehicle with 100,000 miles.  This
    compares with an after repair emission level for the same age and mileage of 0.260 g/mi
    running HC  at a cutpoint combination of 1.2/20 g/mi HC/CO. In this example, the after
    repair emission level (0.206 g/mi HC) is above the value of the normal emitter (0.159 g/mi
    HC).  However, if the calculation produced a value  which was lower, then the normal
    emitter value would be used.


    3.6.3   Discussion of the After Repair Adjustment Factors

           This approach attempts to utilize the large sample of before and after repair EVI240
    data collected in Arizona. These data are an improvement over the MOBILES assumptions
    since they are a large sample, and are representative of the actual I/M experience.  The in-
    use data reflects the fact that regular  commercial mechanics performed the repairs under
    actual cost conditions. Also, the repairs were targeted to passing the actual state EVI240
    test. Many of these technicians also received some training and orientation to the EVI240
    program provided or encouraged by the State of Arizona prior to its implementation. The
    principal assumption underlying this  approach is the ratio between the  after repair EVI240
    emission level and the emission level of the vehicles passing the state EVI240 test is the
    same as the ratio of the after repair running LA4 emission level and the normal emitter
    running LA4 emission level.  This is not an unreasonable assumption; however, there are
    potential differences between the unpreconditioned EVI240 and the preconditioned running
    LA4 test.

           One drawback to the approach is that the Arizona data (and other states' data) were
    available at  only  one cutpoint level (phase-in cutpoints).   This made it impossible to
    determine the sensitivity of repair levels to the EVI240 cutpoint.  To overcome this obstacle
    the previous FTP databases used for MOBILES were used to make the after repair effects
    a function of cutpoint. A drawback to the use of these FTP data is that they are a relatively
    small sample, the repairs were often performed by expert emission control system

M6IMOO1.WPD DRAFT                  21                         Mar 24,  1999

-------
                                                                   DRAFT

    technicians rather than commercial technicians, cost was usually not a factor in the repairs,
    and specified numerical repair targets based on the FTP test were used. Also, running LA4
    were not available so the FTP data were used directly under the assumption that the ratio
    between cutpoints is same for the FTP and the running LA4.
    3.6.4   Technician Training Effects

           MOBILES had built-in I/M credits available for EVI240 programs which conducted
    some form of technician training for people involved in I/M repairs. In MOBILE6, the after
    repair emission levels discussed previously in Section 3.6 already include the effects of
    technician training. This is because Arizona conducted a technician training program prior
    and during implementation of their EVI240 program from which the repair effects data are
    based.

           Thus, it is proposed for MOBILE6 that the default after repair emission levels are
    those 'with technician training'. For I/M programs which do not conduct a technician's
    training program - 'w/o technician training', the after I/M repair emission levels will be
    increased by the percentages shown in Table 2f.

           The percentages shown in Table 2f are based on a limited study done by EPA to
    evaluate technician training in  an EVI240 program.  In the program, eleven experienced
    technicians in Arizona were trained on the eve of the EVI240 implementation in 1995 to
    repair emission failures using a training program developed by Aspire, Inc., and taught by
    an expert emission control system technician/trainer under EPA contract. Each participant
    received the training and three  vehicles to repair following the training. Unfortunately,
    budget limitations prevented a good pre-training baseline  of the technicians' performance
    to be established. The study is fully documented in SAE Paper 960091.

           The emission results shown in columns 2 and 3 of Table 2f are EVI240 test results
    in units of grams per mile.  The Student Tech column  shows two numbers.  The first
    number is the before any repair emission level.  It is shown  for comparison only, and to
    demonstrate that the technicians made sizeable emission reductions from repairs.  The
    second number is the average after repair EVI240 emission levels  of the vehicles after the
    students completed their work.  The Master Tech column shows the average after repair
    EVI240 levels after the instructor completed any additional repairs which were needed to
    bring the vehicle into complete compliance. On a few vehicles this included a new catalytic
    converter.

           The % Difference column is the percent difference between the after repair student
    tech and the after repair master tech emission results with the after repair master tech results
M6IM001.WPD DRAFT                 22                        Mar 24, 1999

-------
                                                                    DRAFT

    as the basis.  It demonstrates the potential difference in performance between a master tech
    and a trainee (journeyman) tech. It is proposed for MOBILE6 to calculate the 'w/o tech
    training' after repair levels (w/o means without) by increasing the 'with tech training' values
    by the % Difference values in Table 2f.
Table 2f
Technician Training Emission Effects
Pollutant
HC
CO
NOx
Master Tech
IM240 (g/mi)
0.38
3.00
1.11
Student Tech
IM240 (g/mi)
2.16/0.68
26.4/8.21
3.66/1.54
% Difference
78%
174%
39%
           Use of these limited data in MOBILE6 for technician training effects requires two
    important assumptions. First, that the after repair levels developed in the previous sections
    already contain the effects of technician training. This is a reasonable assumption since
    Arizona did institute a technician training program, and the after repair emission levels are
    at relatively low levels. Second, that the difference on a percentage basis between the
    master tech performance and the student tech performance is the same as the percentage
    difference between the with and w/o technician  training in  the overall fleet.   This
    assumption is a little tenuous since the performance of typical trained  technician is not as
    high as the master tech in this study.  This would  have a tendency to produce a larger
    percentage increase than in  actuality.  On the other hand, the student tech results were
    collected after the training rather than before the training, and do not strictly represent un-
    trained technicians.  This factor would have a tendency to produce a smaller percentage
    increase than in actuality.

           Overall, the two factors discussed above might tend to cancel each other out.
    However, because of these problems, the MOBILE6 program will allow optional user input
    of 'w/o technician training' emission increase percentages. These will be for users who do
    not have or plan to have a technician training program as part of their I/M program, but can
    nevertheless estimate the  detrimental impact of not having  one through engineering
    judgement, use of data from other I/M programs or other methods.
M6IM001.WPD DRAFT
23
Mar 24, 1999

-------
                                                                   DRAFT

    3.7    Waiver Repair Line

           Not shown in Figure 1 is the waiver vehicle repair line.  However, this line falls
    between the high emitter level and the after proper repairs line. These are failing vehicles
    which received a waiver from program requirements because a minimum amount of money
    was spent on unsuccessful or only partially successful repairs.  Typically, in most I/M
    programs this means that between $200 and $450 was spent on the vehicle, and it still fails
    the I/M test.  The waiver repair line is below the high emitter line, despite the vehicle's
    failing status, because even some limited or ineffective  repair translates into reduced
    emissions on average.

           Because no  analysis has yet been conducted on data from operating EVI240
    programs to estimate the after I/M emission level of vehicles which were waived from the
    requirement to pass the test, an assumed reduction percentage will have to be used, or the
    individual user will have to provide a value.  The default value will  be a 20 percent
    reduction from the high emitter line for all pollutants.  The user will also have the option
    of providing their own value(s) separately for each pollutant based on program data.  If
    EPA completes such an analysis between draft and final versions of this portion  of
    MOBILE6, or receives one in the comment period, the default value may be changed  to
    another number.

    3.8    Percentage of High and Normal Emitters in the Fleet

           Figure 1 shows in a general  sense the overall fleet average emission level, the
    average emissions of the  normal emitters, and the average emissions of the high emitters.
    The fleet average emission level was developed independent of the I/M credits, and the
    methodology for its development is documented in EPA document M6.EXH.OO 1.  In-order
    to compute the I/M credits, the percentage of high emitters and normal emitters in the fleet
    must also be calculated.  Fortunately, this is an easy task since the average emission rate
    is a weighted average of the normal emission rate and  the high emission rate.   The
    weighting factors are simply back calculated to make this true at all odometers.

           The fraction of High and Normal emitters is calculated  for each combination  of
    vehicle type / pollutant  / model  year /  technology group using the following general
    equations.

    Where:

    Highs = fraction of High emitters at each age point
    Normals = fraction of Normal emitters at each age point
    LA4 is the average emission rate at each age point (determined in M6.EXH.001)
M6IM001.WPD DRAFT                 24                         Mar 24, 1999

-------
                                                                   DRAFT

    High_ave is the high emitter emission average at each age point
    Norm_ave is the normal emitter emission average at each age point

           Highs + Normals = 1                                                 Eqn 5

    and
           LA4 = High_ave * Highs + Norm_ave * Normals                       Eqn 6

    Solving for the variables Highs and Normals produces:

           Highs = (LA4 - Norm_ave) / (Highave - Norm_ave)                    Eqn 7

           Normals = 1 - Highs                                                 Eqn 8
           For the model year groups of 1981-82 and 1983-85 HC and CO emissions, it was
    found that the base emission factors at higher mileage levels become higher than the
    average emissions of the high  emitters.  It occurs because at high mileages the basic
    emission factors are data extrapolations. However, under the structure of the model, this
    is not possible, and it implies that the fleet contains more than  100 percent high emitters.
    To overcome this inconsistency, it was assumed that the average base emission factors
    could not continue to rise after it reaches the average of the high emitters, and that it would
    be set to the  average of the high emitters.  Typically, the cross-over point is between
    150,000 and 200,000 miles, and after this point is reached, it is assumed that the percentage
    of highs in the fleet for this model year group / technology is 100 percent. This flattening
    of the emission factor line at very high mileages is consistent with some remote sensing
    studies. A physical explanation would be that while some surviving vehicles continue to
    deteriorate, the worst emitters are progressively scrapped out of the fleet in the high mileage
    range.
    3.9    High Emitter Identification Rates

           The high emitter identification rate (DDR) represents the ability of an I/M test to
    identify (fail) vehicles which are high emitters. It is represented as the percentage of the
    total sum of emissions from the high emitters in the fleet. For example, the DDR would be
    100 percent if it identified all of the running LA4 emissions from the high emitters in the
    fleet. For the HC and CO I/M credits, the IDR is a function of the EVI240 HC and CO
    cutpoints. For NOx I/M credits, it is a function of the NOx cutpoints only. In MOBILE6,
    the user will be able to supply the exact EVI240 cutpoints which are desired, and  the
    program will automatically calculate the IDR and the credits. The EVI240 cutpoints will
M6IM001.WPD DRAFT                 25                         Mar 24, 1999

-------
                                                                   DRAFT

    need to be in the ranges: HC: 0.50 to 5.0 grams/mile; CO: 5.0 to 100.0 grams/mile; and
    NOx: 1.0 to 5.0 grams/mile.

          The I/M IDRs equations  were calculated from the 910 vehicle  database that
    contained vehicle emission data from both running LA4 tests (FTP tests) and EVI240 tests
    on lane fuel on cars and trucks. Cars and trucks will have the same DDR rates in MOBILE6
    at a given cutpoint. However, separate cutpoints will be allowed for cars and trucks and
    for each model year in a given MOBILE6 run. The analysis to develop the IDRs consisted
    of several steps:

          (1) The sample was split into two groups - the high HC and CO emitters, and the
    high NOx emitters. There was some overlap between the groups. These two groups were
    kept separate throughout the rest of the IDR analysis.   (2) The total HC, CO, and NOx
    emissions from all of the High emitters in the sample was calculated.  (3) A total of 75 HC
    / CO cutpoint combinations were developed. These ranged from (0.5g/mi HC / 5g/mi CO)
    to (5.0g/mi HC / lOOg/mi CO).  For NOx,  eight cutpoints were used that ranged from 1.0
    g/mi to 5.0 g/mi. (4) The runningLA4 emissions identification rate (DDR) was determined
    for each cutpoint combination.  For example, the  strict cutpoint combination of 0.5 g/mi
    HC / 5.0 g/mi CO might identify  90 percent of the  total emissions of the high emitters
    whereas the lenient cutpoint combination of 5.0 g/mi HC / 100 g/mi CO might identify only
    10 percent of the total emissions.  (5) The identification rate (IDR) were calculated for 75
    HC/CO cutpoint combinations, and these points were least squared regressed versus the
    natural logarithms of the HC and CO cutpoint. Natural log regressions were used because
    they produced better fits, and better satisfied the inherent assumptions behind least squares
    linear regression. The logarithm form also makes sense physically given the skewed
    distribution of emissions.  For example, a change of the HC cutpoint from 1.0 to 1.5 g/mi
    has a larger effect on IDR than a change from 4.0 to 4.5 g/mi. The regression coefficients
    are shown in Equations 9 and 10. (6) The NOx emission identification rate (IDR) were also
    calculated for eight cutpoints  and fitted to a cubic equation. The cubic form was chosen
    because it provides a very good fit, and does not create anomalous results such as an IDR
    decrease as the cutpoint gets more stringent (See Appendix C).

          In MOBILE6, the IDRs for all 1981 and later cars and light trucks are represented
    by Equations 9 through 11. Where In(HCcut), In(COcut), and ln(NOcut) are the cutpoints
    transformed into natural logarithm space.
           HC IDR= 1.1451 - 0.1365*ln(HCcut) - 0.1069*ln(COcut)              Eqn 9

           COIDR= 1.1880-0.1073*ln(HCcut)-0.1298*ln(COcut)              Eqn 10



M6IM001.WPD DRAFT                 26                        Mar 24, 1999

-------
                                                                   DRAFT

    The NOx IDR equation is a cubic form:

           NOxIDR = 0.5453 + 0.7568*NOcut - 0.3687*NOcut2 + 0.0406*NOcut3  Eqn 11

    3.10   I/MNon-Compliance Rates

           One potential problem in I/M is that of non-compliant vehicles.  By definition,
    these are the  high emitting vehicles which fail the initial test, but drop out of the I/M
    process prior to receiving a passing test or a cost waiver.  This type of non-compliant
    vehicle is assumed to remain a high emitter at the average high emitter emission level (no
    reduction is given like in the case of cost waived vehicles).   In the MOBILE6 model, the
    non-compliant vehicles will be represented as a fraction of the identified high emitters that
    did not pass or receive a cost waiver. A default value of 15 percent will be built into the
    model  for the non-compliance rate.  It is based on studies where  large samples of I/M
    vehicles were tracked as they passed through I/M programs. Optional user inputs will also
    be available that permit any number from 0 percent to 100 percent to be used if supported
    by data.

           The other type of non-compliant vehicle is one which does not show up for its
    initial test (owner ignores I/M). If these vehicles are normal emitting vehicles (passing the
    I/M test) they have no effect on the result; however,  if they are high emitters then they
    should have the same effect  as the initial  failures  which never pass or get waived.
    Unfortunately, because they do not show up for I/M  it is impossible to determine these
    statistics.  As an approximation, it is assumed that the 15 percent non-compliance rate
    (from above) includes the effect of high emitters which did not show up for their first test.
    Similarly a user defined input for non-compliance should take these  vehicles into account.
    3.11   Average Emissions After I/M

           An important step in calculating the I/M credits is to calculate the average emissions
    of the fleet after a cycle of I/M testing and repair.  The average is calculated for each
    vehicle type / pollutant / model year group/ technology group at many odometer points
    during the life of the group. Conceptually, the average emissions of the fleet after I/M is
    a weighted sum of (1) the normal emitters which were unaffected by I/M, (2) the high
    emitters which were not identified by I/M and which keep the same high emissions, (3) the
    high emitters which were non-compliant and which keep the same high emissions, (4) the
    high emitters which were identified and cost waived, and (5) the high emitters which were
    identified and successfully repaired by the I/M process. The last type drops down to the
    after repair levels (FIX in Equation 12) calculated in Section 3.6  (Line D).
M6IM001.WPD DRAFT                 27                        Mar 24, 1999

-------
                                                                  DRAFT

           Equation 12 is used to calculate the average emissions of the fleet after a cycle of
    I/M.
           EIM  =      N*(1-X)+  H*X*(1-IDR)+  X*IDR*W*H*RW  +
                              N*R*X*IDR*FIX +  H*X*IDR*NC             Bpl2a
    Where:

          N*(1-X) = Normal Emitters emission effect

          H*X*(1-IDR) = High Emitters not identified emission effect

          X*IDR*W*H*RW = High Emitters identified and Waived emission effect

          N*R*X*IDR*FIX = High Emitters identified and Repaired emission effect

          H*X*IDR*NC =  High Emitters identified and Non-Compliant vehicles emission
                              effect

    Variables:

          N     Emission Level of Normal Emitters (g/mi).
          H     Emission Level of High Emitters (g/mi).
          X     Fraction of High Emitters in the fleet before the cycle of I/M.
          DDR   Fraction of Total Fleet High Emitters Identified by an I/M test.
          W    Fraction of Identified High Emitters which get a repair cost waiver.
          NC   Fraction of Identified High Emitters which are in non-compliance of the I/M
                 program.
          FIX   Fraction of Identified High Emitters which get repaired to pass the test.

          R     Fraction of the normal emitter level that high emitters are repaired after I/M
                 (value is >  1.0).
          RW   Fraction of the  high emitter level that waived high emitters are repaired
                 after  I/M.

          The fractions W, NC, and FIX are all applied to the DDR fraction.  An identified
    vehicle is either waived, in non-compliance, or is properly repaired.  Thus,

                 W + NC + FIX   =    1.0                                    BpEb


M6IM001.WPD DRAFT                 28                         Mar 24, 1999

-------
                                                                 DRAFT
    4.0   Periodic I/M Credit Algorithm (Sawtooth Methodology)

          This section describes the methodology for the periodic I/M credit algorithm over
    time or the "sawtooth" methodology.  The first few sections describe the algorithm and
    equations for the "sawtooth" when neither a remote sensing program that identifies high
    emitters, nor a change of ownership testing requirement is present. The remaining sections
    describe the "sawtooth" algorithm for a combined program of periodic I/M (I/M), change
    of ownership I/M (COIM), and remote sensing. Both algorithms are essentially the same,
    except the I/M + RSD + COIM algorithm works on a bi-annual  (every six months) basis,
    and the I/M only algorithm works on an annual or biennial basis. MOBILE6  will also
    compute benefits for other variations of I/M, RSD, and COIM. These include: I/M+RSD,
    I/M+COEVI, and COIM only. The algorithm for these variations is essentially the same as
    for the base I/M+RSD+COEVI case.
    4.1    Discrete Points

          The MOBILE6 program will not use "continuous" regression lines of emissions
    versus mileage to represent the before and after I/M emission rates, but instead will use
    discrete points on these lines. Each point on the line will represent a particular vehicle age
    that ranges from 1 to 26 years.  Table 3 shows the correspondence of age and cumulative
    mileage for cars. Each particular age and mileage corresponds to a January 1st reference.
    The six month mileage points needed for RSD and COIM will be generated from the
    mileages on this table by averaging the two surrounding points.  For example, age =1.5
    years (18 months) is obtained by averaging the points at age = 1 and age = 2.

          The text describing the I/M credits with NO RSD or COIM will use the index
    variable 'ii' to  represent yearly intervals. The text describing the I/M credits  with RSD
    and/or COIM will use the index variable 'i' to represent six month intervals.
M6IM001.WPD DRAFT                 29                        Mar 24, 1999

-------
                                                                  DRAFT
Table 3
January 1st Mileage and Age Correspondence for Cars in MOBILE6
Age
1
2
3
4
5
6
7
8
9
10
11
12
13
Cumulative Mileage
(in 1000's)
2.142
12.823
29.335
45.050
60.006
74.239
87.786
100.678
112.948
124.625
135.738
146.315
156.380
Age
14
15
16
17
18
19
20
21
22
23
24
25
26
Cumulative Mileage
(in 1000's)
165.960
175.077
183.753
192.010
199.869
207.349
214.466
221.241
227.688
233.823
239.663
245.220
250.509
    4.2    Effect of Exemptions on I/M Credits

           I/M exemptions are a provision granted to some vehicles which would ordinarily
    be subject to an I/M inspection that excuses them from all  of the testing and repair
    requirements of I/M. In practice, this means that the motorist does not have to bring the
    vehicle in for an I/M test; however, it may require the motorist to have received a roadside
    remote sensing device (RSD) "clean screening" test(s), or to have paid a fee in-lieu of the
    test.

           MOBILE6 will be able to account for consecutive age / model year exemptions
    starting with age = 1. For example, most programs exempt the youngest fleet age or the
    newest model year in the fleet. This means that vehicles which are one year old are not
    tested.  Because it is so common, a one year exemption is the default pattern shown for the
M6IM001.WPD DRAFT
30
Mar 24, 1999

-------
                                                                   DRAFT

    annual and 1-3-5 biennial I/M algorithms. A two year exemption is the default pattern for
    the 2-4-6 biennial algorithm. MOBILE6 will also have the ability to exempt the first of any
    consecutive new model years (i.e., exempt vehicles 2, 3, 4, 5, ... n years old). The effect
    of this is to shift the 'sawtooth' curve to the right.  Exempting older model year / age
    vehicles (i.e., all of the vehicle 15 years or older) will also be available in MOBILE6.
    Mechanically, this will be done by computing the credits using the standard algorithm, and
    then setting them to zero for the exempted model years.
    4.3    Annual I/M Credits with NO Remote Sensing

           The MOBILE6 model will generate separate I/M credits for each combination of
    vehicle type / pollutant / model year group / technology class. In concept, these credits
    could be generated by comparing the basic emission rate line - (see Section 3.3) with the
    average emissions after I/M line  - (see Section 3.9). However, because of a number of
    complications these lines cannot be used directly. Instead EPA developed the 'sawtooth'
    algorithm shown conceptually in Figure 2a.

           One  of the problems with a linear approach is the distribution of ages within a
    model year group. For purposes of modeling, all vehicles are assumed to be inspected on
    the first anniversary of their purchase and periodically thereafter, always on that same date.
    It is also assumed that sales occur exactly in the 12 month period from October of the
    calendar year previous to the model year through September of the next calendar year. For
    example, in January, 1999, the age distribution of the 1997 model year vehicles will range
    from 2.25 years to 1.25 years. With an annual inspection program, vehicles between one-
    and-two-years-old have only been inspected once. Any vehicles two years and older should
    have received their second inspection.  In this example, 25 percent of the emissions on the
    evaluation date come from vehicles recently completing their second inspection and 75
    percent of the emissions come from vehicles which have been inspected only once.

           Another factor which must be taken into account is the deterioration of the vehicles
    in between their yearly inspections and repairs. Existing evidence suggests that the type
    of problems which cause I/M failures can re-occur as often in the repaired vehicles as they
    do in the unrepaired fleet. Thus, it is assumed that the fleet, after repair, will have the same
    emission deterioration as before repairs.

           Figure 2a graphically shows the I/M credit methodology. The top set of 26 (ages)
    points (only 6 are shown) is the basic emission rate for  a given group (vehicle type /
    pollutant / model year group / technology). For instance, Points B, C, and D show the non
    I/M line for vehicle ages of 1, 2, and 3 years. The lower 'sawtooth' figure is the I/M line.
    The 'sawtooth' illustrates the effect of I/M inspection and repair and the  subsequent
M6IMOO1.WPD DRAFT                  31                         Mar 24, 1999

-------
                                                                  DRAFT
                                       FIGURE 2A
                         ANNUAL I/M CREDITS SAWTOOTH
          LA4
      EMISSIONS
B
                                         MILEAGE
    deterioration of the fleet. All deterioration slopes are parallel (i.e., segment B-C is parallel
    to segment E-F).  The repair effect is represented by the sudden drop in emission level at
    each inspection interval (i.e, from Point F to Point G).  The heavy shaded portions of the
    lines illustrate how an I/M credit for the given group at age X is produced. MOBILE6
    always chooses January 1 st as the evaluation date. The vehicles sold from October through
    December are represented by the short line segment to the right of the two year anniversary
    point. These are vehicles which are older than X years.  The longer line segment to the left
    of the anniversary point represents the vehicles sold from January through September,
    which are still less than X years old at the January evaluation date. The weighted average
    of each segment is calculated and the percent difference between the two weighted averages
    is computed. This percent difference is the I/M credit for a given age.
           Mathematically, this process is shown for the Non I/M (top line) and the I/M
    (sawtooth line) as:
M6IM001.WPD DRAFT
              32
Mar 24, 1999

-------
                                                                  DRAFT

    Nonl/M

           NOIM(ii)  =   0.75 *  (E(ii) -  0.375*(E(ii)  - E(ii-l))  )  +
                        0.25 *  (  E(ii) +  0.125*(E(ii)  - E(ii+l))  )        Egn  14
           Where NOIM is the average Non I/M value at age = ii in the equations above. It is
    the average of the two segments.   Note "ii" means the "ii th" point.  It should not be
    confused with Point I on the Figure 2B.

    E(ii) is the basic emission rate at point ii. E(0) is the value at Point A.

    The values of E(ii), E(ii-l), and E(ii+l) take into account the slope of the line between age
    = ii and age = ii + 1 and age = ii -1. Figure 2a is an idealized drawing using a straight line.
    In the MOBILE6 model, the lines have some slight curvature due to the high emitter
    correction factor; thus, the slope is generally not the same between all segments.

    Also, the weighting factor values of 0.375 and 0.125  shown throughout these equations
    represent the average of the heavy shaded segments in Figure 2A. For example, the 0.375
    represents the average weighting of the highlighted  segment between points B and C, and
    the 0.125 represents the average weighting of the highlighted segment between points C
    andD.

    I/M          Special Case age = ii = 1

           IM(1)  =     0.75  *  (  E(ii)    -  0.375*(E(ii)  -  E(ii-l)  ) +
                         0.25  *  (  EIM(ii)  +  0.125*(E(ii + 1)  -  E(ii))  )    Egn 15
                 General Case age > 1

           IM(ii)   =    0.75  *  [  EIM(ii)  -  0.375*(E(ii)  - E(ii-l)  ]  +
                        0.25  *  [  EIM(ii)  +  0.125* (E(ii + 1) -  E(ii)) ]      Egn  15

           Where EVI(ii) is the I/M line. Where EIM(i) is the average I/M emission line after
    repair, waiver, and non-compliance factors from Section 3.9.

           The I/M credits are computed by dividing the difference between the NOIM
    emission value and the EVI emission value by the NOIM emission value. An I/M credit is
    obtained for the ages 1 through 25.

           IMCRED(ii)   =      [NOIM(ii)  - IM(ii)]  /   NOIM(ii)            Egn  15
M6IMOO1.WPD DRAFT                 3 3                        Mar 24, 1999

-------
                                                                   DRAFT

    4.4    Biennial I/M Credits with NO Remote Sensing

           The values of E(ii) and EIM(ii) used in Equations 13 through 16 are also used to
    compute biennial I/M credits using a 'sawtooth' algorithm.  The only difference between
    the annual and the biennial I/M credits is that the biennial values are applied every other
    year and that there is consequently a longer period of deterioration between I/M inspections
    and repairs. Figures 2b and 2c are analogous to Figure 2a.  Figure 2b is an example of a
    1-3-5 biennial program in which a vehicle is first inspected when it is one year old and then
    every two years thereafter.  Figure 2c illustrates a 2-4-6  biennial program which first
    inspects a vehicle when it is two years old and then does an inspection every other year.
    The differences are small for a fleet that has a full complement of vehicle ages.  The
    "Mixed" Biennial credits (Mix Bien EVICRED) are an average of these two program types.
    "Mixed Biennial" was the default for MOBILES.  This is an average of the 1-3-5 and 2-4-6
    plans, or any mixed biennial program in which half of each model year or half of the fleet
    is inspected during each calendar year.
M6IM001.WPD DRAFT                 34                         Mar 24, 1999

-------
                            FIGURE 2B
              1-3-5 BIENNIAL I/M CREDITS SAWTOOTH
       LA4
    EMISSIONS
                               MILEAGE
                            FIGURE 2C
              2-4-6 BIENNIAL I/M CREDITS SAWTOOTH
      LA4
    EMISSIONS
                                MILEAGE
M6IM001.WPD DRAFT
35
Mar 24, 1999

-------
                                                                 DRAFT
          Years of exemption = X. The Biennial 1-3-5 pattern and the Biennial 2-4-6 pattern
    have the same general equations except i starts at i=l (X=l) for the first and i=2 for the
    second (X=2).
    Nonl/M
          NOIM(ii) =   0.75 *  [E(ii)  - 0.375*(E(ii)  - E(ii-l))  ]  +
                        0.25 *  [E(ii)  + 0.125*(E(ii)  - E(ii+l))  ]         Eqn  14
    I/M          Special Case ii = 1 for 1-3-5 Case or ii = 2 for 2-4-6 Case

           IM(1)   =      0.75 *  [ E(ii)    - 0.375*(E(ii)   -  E(ii-l) ]   +
                         0.25 *  [ EIM(ii)  + 0.125*(E(ii + 1)  -  E(ii)) ]    Eqn  15
    General Case ii > 2*X

    IM(ii) =     0.75*[EIM(ii-2)+E(ii-l)-E(ii-2)  + 0.625*(E(ii)-E(ii-1))]
                 0.25* [EIM(ii) +  0.125*(E(ii + 1)-E(ii))  ]                 Eqn20
    IMCRED(ii)   =      [NOIM(ii)  -  IM(ii)]  /   NOIM(ii)                  Eqn21


    MixBien  IMCRED  =   [(1-3-5 Bien Credit+ 2-4-6 Bien Credit]  / 2     Eqn22



    4.5   I/M Credits with Remote  Sensing and Change of Ownership

          The MOBILE6 program will be able to calculate I/M credits for programs which
    conduct periodic I/M tests, RSD testing to identify high emitters, and change of ownership
    (COIM) I/M credits.  The methodology in Sections 4.3 and 4.4 was for periodic I/M only
    type programs. Since RSD and COIM are non-periodic in nature, it is assumed that they
    identify and repair vehicles in a continuous distribution throughout a calendar year (i.e., all
    of the RSD testing or change of ownership testing is not conducted in June). Based on this
    assumption, this testing pattern is equivalent to a periodic test every six months. The RSD
    factor used in Equations in Sections 4.5.1 through 4.5.3 is computed as a product of the


M6IMOO1.WPD DRAFT                 3 6                        Mar 24, 1999

-------
                                                                  DRAFT

    RSD effectiveness parameter (Effectiveness), and the RSD Fleet coverage (Coverage).
    These two basic parameters account for the RSD's ability to individually identify high
    emitters, and its ability to test the entire fleet.

           Also, the RSD and COEVI credits are computed at each six month interval between
    the regular periodic inspections.  Thus, there is one RSD / COEVI interval (sawtooth
    pattern) between each periodic inspection in an annual program, and three RSD / COEVI
    intervals between each biennial inspection.  This assumes that on average the entire I/M
    fleet will not be inspected by RSD more than once during a year.
    4.5.1   RSD Effectiveness Values

           Table 4 shows the RSD effectiveness for HC, CO and NOx pollutants versus RSD
    percent CO readings.  The individual values in the table represent the emissions identified
    by RSD at particular  cutpoints ranging from 0.5% CO to 7.5% CO.  The effectiveness
    values are based on studies done by EPA in Arizona, and by CARB in Sacramento.  These
    values will  be used as the RSD Effectiveness parameters in the MOBILE6 model.  For
    more details on how these values were derived see "EPA420-R-96-004 - "User Guide and
    Description for Interim Remote Sensing Program Credit Utility".
M6IMOO1.WPD DRAFT                  3 7                        Mar 24, 1999

-------
                                                                 DRAFT
Table 4
Remote Sensing Effectiveness Versus CO Cutpoint for 1981-93 Model Year Vehicles

RSD CO Cutpoint
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
4.0%
4.5%
5.0%
5.5%
6.0%
6.5%
7.0%
7.5%
HC Effectiveness
0.570
0.433
0.387
0.348
0.319
0.262
0.217
0.182
0.150
0.109
0.071
0.060
0.046
0.039
0.028
CO Effectiveness
0.596
0.499
0.442
0.396
0.352
0.278
0.213
0.178
0.133
0.107
0.072
0.053
0.044
0.034
0.017
NOx Effectiveness
0.283
0.178
0.122
0.091
0.059
0.054
0.042
0.018
0.015
0.009
0.006
0.003
0.003
0.003
0.003
    4.5.2  RSD Coverage Options

          Three basic RSD program options will be available to the MOBILE6 user. Option
    1 is the "Level of Effort Commitment", Option 2 is the "Specific Level of Fleet Coverage
    Commitment", and Option 3 is the "Number of Failures Commitment". All three of these
    options utilize the coverage and effectiveness parameters. However, different methods are
    used to calculate the two parameters, and different user inputs are required.
    (See RSD references for more details).
M6IM001.WPD DRAFT
38
Mar 24, 1999

-------
                                                                  DRAFT
    Option 1 - Level of Effort Commitment
           In this option, the user enters the number of vehicle RSD tests which are to be done
    on an annual basis, and the total size of the fleet (in number of vehicles) which is subject
    to inspection. A modified Poisson algorithm is used to estimate the number of vehicles
    seen by remote sensing in order to calculate the fraction of the fleet tested by RSD
    (Coverage).  This is necessary, since the fraction of all vehicles in the fleet which are
    measured by remote sensing is a function of the total number of RSD readings since some
    vehicles are seen multiple times by RSD. A Poisson algorithm is a standard method to
    model such a situation.  The coverage fraction is also a function of the annual average
    VMT of a vehicle model year at a given age compared to its VMT when new.   The
    equations which are used are:
           P(X)
Lambda* *X * exp(-Lambda) / X!
           Where,

           X
           P(X)

           Lambda



           Lambda =
Is an integer number starting at zero that represents the number of
RSD tests which a vehicle receives before it is called in for a
confirmatory test.  Most programs will require at least two failing
tests (X=2) prior to the confirmatory test.

The vehicle age under evaluation.

The Poisson distribution function

Is the mean number of RSD tests during a given year. For example,
if half of the fleet is inspected on average then Lambda would be
0.50. Mathematically, it is represented as:

# RSD tests/yr * VMT(i) / ((# Veh in Fleet)*VMT(l))
           Coverage(X)  =      1.0  - SUM(P(X-1))
           "No RSD" is the default option, However, if RSD is to modeled then Option 1 is
    recommended.
M6IM001.WPD DRAFT
                39
Mar 24, 1999

-------
                                                                  DRAFT

    Option 2 - Specific Level of Fleet Coverage Commitment

           In this option, the user specifies the fraction of the fleet in each model year that is
    seen using remote sensing. This fraction should only be the fraction of the fleet which has
    had sufficient valid remote sensing measurements to be identified as remote sensing
    failures for purposes of further I/M inspection.  For example, if the RSD  program is
    designed so that three RSD failures are needed before a vehicle is  sent for off-cycle I/M
    testing (confirmatory testing), the fraction of the fleet used as the coverage commitment
    should represent the portion of the fleet that has received three RSD readings.

           Thus,

           Coverage = User Input

    Option 3 - Number of Failures Commitment

           In this option, the user specifies the fleet size by model year (in units of number of
    vehicles),  and number of confirmed I/M failures by model year  which the RSD will
    identify. Only vehicles identified for inspection by remote sensing and which fail the I/M
    inspection are to be included. In the MOBILE6 model, this option effectively combines the
    Effectiveness and the Coverage into one parameter - the fraction of high emitters identified
    by RSD. This value is then normalized to the entire fleet fraction of high emitters. The
    variable "Highs"  from Equation 7 in Section 3.8 is used to normalize the RSD failure rate.
    The normalized value becomes the RSD parameter used in the equations in Sections 4.5.2
    and 4.5.3.

           RSD Fail Rate        =      # RSD Failures / Fleet Size

           RSD                =      RSD Fail Rate / Highs

    4.5.3   RSD Effective and Coverage Together

           The final RSD value used in the I/M + RSD credit equations in Equations 4.5.2 and
    4.5.3 is the product of the RSD Effectiveness and the RSD Coverage.  For Coverage
    Options 1 and 2, the values are calculated separately and multiplied together.  For Coverage
    Option 3, the RSD effectiveness and coverage are implied in the RSD Fail Rate. To assure
    consistency between all three coverage methods, MOBILE6 will prompt  the user if
    unreasonable values for "RSD" or RSD Coverage parameters are used.

           RSD = RSD Coverage * RSD Effectiveness
M6IM001.WPD DRAFT                 40                        Mar 24, 1999

-------
                                                                   DRAFT
    4.5.4   Change of Ownership Parameters

           The change of ownership testing frequency is assumed to be quite random, and will
    be modeled for simplicity by the same six month interval as RSD. This effectively assumes
    that the change of ownership vehicles are representative of the overall fleet, and that this
    fraction of the fleet receives an additional I/M test every six months. Based on IM240 data
    from Wisconsin, this fraction was determined to be  16 percent annually. A default value
    of 8 percent will  be applied for each six month interval.  The 16 percent value was
    estimated from the Wisconsin change of ownership versus periodic test volume data shown
    in Table 5.  The  numbers are test vehicle counts,  and are based on a quasi-random sample
    of full IM240 initial tests conducted either at Station #12 in  Wisconsin or at  other
    Wisconsin test stations on Saturdays. The Wisconsin I/M program is biennial and the
    periodic tests for the 1996 calendar year are on the odd model years only whereas the
    change of ownership testing is on all model years.

           In addition, anecdotal  evidence from Wisconsin suggests that this model may be a
    simplistic, and under predict the benefits of change of ownership. For example, analysis
    of change of ownership vehicles suggests that they (1) often contain a higher percentage of
    high emitters than the overall fleet, (2) that the high  emitters change ownership  more
    frequently than  the  more normal emitters, and  (3) that a percentage of the change of
    ownership vehicles  change owners more than once during a year.  Therefore, to help
    balance these possible effects, the  effect of waivers and non-compliance  will not be
    assessed on change of ownership I/M testing.

           Because of the uncertainty in estimating change of ownership parameters, the user
    will be allowed to input their own value into the MOBILE6 model. The COIM factor used
    in Equations in  Sections 4.5.1 through 4.5.3 is computed as a  product of the change of
    ownership fraction  and the high emitter identification rate (IDR).   These two factors
    account for how many change of ownership tests are done, and the effectiveness of each
    test.
M6IMOO1.WPD DRAFT                 41                         Mar 24, 1999

-------
                                                                  DRAFT
Table 5
Wisconsin Change-of-Owner Test Volumes
in a 1996 Calendar Year Data Sample
Model Year
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
ALL
Periodic Testing
100

275

679

943

1231

1238
4,466
Change Owner
Testing
31
48
62
201
147
250
212
280
209
250
181
842
% Change of
Owner
24.0%

18.4%

17.8%

18.4%

14.5%

12.8%
15.9%
           The RSD / COIM credits are computed as a fraction of the maximum possible
    periodic I/M credit at that given age. This has the effect of producing smaller "sawteeth"
    for six month intervals which do not coincide with a periodic inspection (smaller means
    that the bottom of the "sawtooth" is higher than the bottom of the periodic inspection
    "sawtooth"). There is no additional credit given, when the RSD / COIM intervals coincide
    with the periodic inspection interval.  See Figures 3a and 3b.
    4.5.1   Special Case for Year 1 of Program

    Exemption = 1 Year  (Annual I/M or 1-3-5 Biennial) then i = 2
    Exemption = 2 Years  (2-4-6 Biennial) then i = 4
M6IM001.WPD DRAFT
42
Mar 24, 1999

-------
                                                              DRAFT
    Exemption = X Years then i = 2*X
    Index variable 'i' represents six month intervals; thus, 2*i = ii
    Nonl/M

    NOIM(i)     =     0.75 * ( E(i) -  0.375*(E(i) - E(i-l)) ) +
                             0.25 * ( E(i) + 0.125*(E(i) - E(i+l)) )      Eqn 14
    I/M

    (a) Points E(i) and EIM(i) are all known.
    (b) I/M and non-I/M lines are parallel

    EO/2) = (E(i-2HE(i-lV)
                 2

    E(i-2),  E(i-l), E(i), and E(i+l) are Known Values
    EEVI(i),  EEVI(i+1) are Known Values

    TOP(i+l) = EIM(i) + (E(i+l) - E(i) )

    SEGMENT 1A = (Ef1/^ + Ei
    SEGMENT2A=   (EIM(i) + [EIM(i) + (E(i+1) - E(i))/2])

                                    2


    IM(i) = 0.75 * SEGMENT1A + 0.25 * SEGMENT 2A


    IMCRED(i)  =    fNOIM(i) - I
                          JNUlM(l)
    4.5.2  Annual I/M Credits


M6IM001.WPD DRAFT                43                       Mar 24, 1999

-------
                                                                 DRAFT
    General Case#l: Annual I/M for Year 2 through 25;

    General algorithm is i + n

    RSD = RSD Coverage * RSD Effectiveness
    COEVI = Change of Ownership Fraction * IDR

    RSD + COIM<=1.0

    E (i - 2)....E (i + 1) are derived from the basic emission factors
    EEVI (i - 2)....EEVI(i  + 1) are calculated in Section 3.11.

    TOP (i - 1) = EIM(i - 2) + (E (i-1) - E(i - 2))
    MID (i - 1) = TOP(i-l) - [TOP(i - 1) - EIM(i - 1)] * RSD
    TOP(i)     =MID(i- 1) + (E(i) - E (i - 1) )
    MID(i)     = EIM(i)
    TOP(i + 1)  = MID(i) + (E (i + 1) - E(i))
    SEGMENT IB =
             I  \EIM (i - 2) + (E(i -1) - E(i - 2)) + TOP (i -1)  I/ 2
                =   (jVIID (i -1} + TOP (i))
SEGMENT 2B =   \ MID (i - 1) + TOP (i)
                        2
    SEGMENT 3B =  I  MID(i) + [ MID(i) + ECi +2D -E(D J   I / 2

    IM(i) = 0.25 *  SEGMENT IB + 0.50 * SEGMENT 2B + 0.25 * SEGMENT 3B
    IMCRED(i)    =     (NOIM(n - IM(D
                            NOIM( i)
M6IM001.WPD DRAFT                 44                        Mar 24, 1999

-------
                                                                 DRAFT

    4.5.3  Biennial I/M Credits

    General Case #1:  Biennial I/M for Year i through 25;

    General algorithm is  i + n
    i = 2 for the 1-3-5 Biennial
    i = 4 for the 2-4-6 Biennial

             CHART
General
i-4
i-3
i-2
i- 1
i
i+ 1
Example
2
3
4
5
6
7
    RSD = RSD Coverage * RSD Effectiveness

    COEVI = Change of Ownership Fraction * DDR


    RSD + COIM<=1.0

    E (i - 2)....E (i + 1) are derived from the basic emission factors

    EEVI (i - 2)....EEVI(i + 1) are calculated in Section 3.11.

    EIM(i-4)      =     MID(i-4)
    TOP(i-3)      =     MID(i-4) + (E(i-3) - E(i-4))
    MID(i-3)      =     TOP(i-3) - [TOP(i-3) - EIM(i-3)] * RSD
    TOP(i-2)      =     MID(i-3) + (E(i-2) - E (i-3))
    MID(i-2)      =     TOP(i-2) - [TOP(i-2) - EIM(i-2)] * RSD
    TOP(i-l)      =     MID(i-2) + (E(i-l) - E(i-2))
    MID(i-l)      =     TOP(i-l) - [TOP(i-l) - EIM(i-l)] * RSD
    TOP(i)       =
    EIM(i) =      MID(i)
    TOP(i+l)     =
M6IM001.WPD DRAFT                 45                        Mar 24, 1999

-------
                                                                    DRAFT
SEGMENT IB =
                  I  (EIM (i - 4) + ( E(i - 3) - E(i - 4))/2 + TOP (i - 3) I/ 2
                = (j
SEGMENT 2B = \ MID (i - 3) + TOP (1-2)


                       2
    SEGMENT 3B =
              I MID(i-2) + [ MID(i-2) + (E(i-l) - E(i-2))/2 J   I / 2
    IM(i) = 0.25 * SEGMENT IB + 0.50 * SEGMENT 2B + 0.25 * SEGMENT 3B
    IMCRED(i)     =      (NOIM (i 1 - IMCD
                            NOIM( i)
                  I  (jVHD(i-2) + ( E(i-l) - E(i-2))/2 + TOP (i - 1)  I/ 2
SEGMENT 1C =
    SEGMENT 2C = \MSD (i -1) + TOP (i)J


                          2
                   I  MID(i) + |_ MID(i) + (E(i+l) - E(i))/2J   I / 2
SEGMENT 3C = I MID(i) +    MID(i) + (E(i+l) - E(i
    IM(i+l) = 0.25 * SEGMENT 1C + 0.50 * SEGMENT 2C + 0.25 * SEGMENT 3C
    IMCRED(i+l) =        (NOIM (i 1 - IM(i1
                            NOIM( i)
M6IM001.WPD DRAFT                  46                         Mar 24, 1999

-------
                                                                   DRAFT


    4.6    RSD / Vehicle Profiling Exemptions


    4.6.1   RSD Exemptions

           An important use of the RSD test in the context of I/M may be to use it as a method
    of screening out 'normal' emitting vehicles, and exempting them from regular I/M. The
    motivation  for such a program might be to reduce the inspection cost by exempting a
    fraction of the fleet which is very likely to pass anyway.  Also, since the test is largely
    unknown to the vehicle owner, and rather automatic, it might help build public support for
    an I/M program by inconveniencing fewer motorists.

           The  RSD  clean screening  logic is  similar to  that used  in the  high  emitter
    identification  algorithm.   Both involve the  terms RSD fleet  coverage  and RSD
    effectiveness.  However, clean screening is attempting to properly identify low emitting
    vehicles for exemption from further program  requirements while RSD  high  emitter
    identification is concerned with identifying high emitters for further testing and repair.


    Clean Screening Coverage Options

           Options 1  and 2 presented in Section 4.5.2 will be used for the clean screening
    coverage.  These two options are the "Level of Effort Commitment", and the "Specific
    Level of Fleet Coverage Commitment". The same equations and algorithms will be used
    to model clean screening coverage as were used to model high emitter identification
    coverage (i.e., Poisson distribution equations). Option 3 will not be used because it only
    applies to high emitters.


    Clean Screening Effectiveness Values

           The RSD clean screening effectiveness values are shown in Tables 6a and 6b. They
    are shown as a percentage of the I/M credit which is lost through clean screening. They are
    a function of the RSD cutpoint combination, and the stringency of the underlying I/M test
    from which a clean screened vehicle is exempted. Also shown in the table is the percentage
    of the fleet which is exempted (clean screened).  For example, if the RSD cutpoints of 200
    ppm HC and 0.5% CO are used for clean screening, and the less stringent (phase-in) EVI240
    cutpoints are used, then 51 percent of the RSD tested fleet is exempted, 2 percent of the HC
    IDR is lost, 7 percent of the CO IDR is lost and 23 percent of the NOx IDR is  lost.
M6IM001.WPD DRAFT                 47                         Mar 24, 1999

-------
                                                                 DRAFT
    RSD Clean Screening Effectiveness and Coverage Together
           The RSD clean screening effectiveness and RSD coverage are multiplied together
    to produce an overall RSD clean screening effect. Mathematically, this is:
          RSD Loss
RSD Coverage * RSD Clean Screening Effectiveness
           The resulting value for the RSD_Loss is applied by subtracting it from the I/M DDR
    (IDR) obtained in the previous I/M credit equations in Section 3.8. This produces the final
    IDR_RSD.  For example, if the original I/M IDR (IDR) is 80 percent, the RSD coverage
    is 50 percent, and the losses from falsely exempting high emitters using RSD is 2 percent,
    then the I/M credit with RSD (IDR_RSD) is 79 percent.
          A simplified mathematical equation is:
          IDR RSD
IDR -  RSD Loss
Table 6a
Remote Sensing Clean Screening Effectiveness
Interim (Less Stringent) I/M Standards
Clean
Screening
Outpoints
HC 200 ppm
CO 0.5%
NOx - None
HC 200 ppm
CO 0.5%
NOx - 2000 ppm
HC 200 ppm
CO 0.5%
NOx- 1500 ppm
HC 200 ppm
CO 0.5%
NOx- 1000 ppm
Vehicles
Tested
594
594
594
594
% Vehicles
Passing Clean
Screening
51%
40%
37%
29%
% HC IDR
Credit LOST
2%
2%
1%
1%
% CO IDR
Credit LOST
7%
7%
0%
0%
% NOx IDR
Credit LOST
23%
12%
11%
7%
           The default RSD exemptions used in the MOBILE6  model are  based  on an
    extensive study of RSD data and I/M data collected in various cities. The full methodology
M6IM001.WPD DRAFT
         48
Mar 24, 1999

-------
                                                                 DRAFT

    and numbers used in the MOBILE6 model are fully documented in the EPA document
    "Draft Description and Documentation for Interim Vehicle Clean Screening Credit Utility -
    EPA420-P-98-008".
Table 6b
Remote Sensing Clean Screening Effectiveness
Final (More Stringent) I/M Standards
Clean
Screening
Outpoints
HC 200 ppm
CO 0.5%
NOx - None
HC 200 ppm
CO 0.5%
NOx - 2000 ppm
HC 200 ppm
CO 0.5%
NOx- 1500 ppm
HC 200 ppm
CO 0.5%
NOx- 1000 ppm
Vehicles
Tested
594
594
594
594
% Vehicles
Passing Clean
Screening
51%
40%
37%
29%
% HC IDR
Credit LOST
9%
6%
5%
4%
% CO IDR
Credit LOST
7%
5%
1%
1%
% NOx IDR
Credit LOST
28%
15%
12%
7%
    4.6.2  High Emitter Profiling

          High emitter profiling is similar to RSD in that it seeks to screen out low emitting
    vehicles and exempt them from the regular I/M inspection. The benefit is a saving of
    testing resources, and less inconveniencing of motorists.  Like RSD the drawback is the
    loss of I/M benefits from exempting high emitters which should not be exempted.  The
    equations which are used (shown below) are completely analogous to the RSD equations
    in terms of form and use.

          Profile_High =      Function [ %fleet Profiled, Error Rate of Profile]

          IDR_Prof    =      IDR  - Profile_High

          A user of MOBILE6 may want to model an I/M program which does both RSD and
    high emitter profile exemptions. In that case, both the RSD_High and the Profile_High
    losses are subtracted from the based IDR to produce a new IDR.
          IDR RSD Prof
IDR - RSD_High - Profile_High
M6IM001.WPD DRAFT
   49
Mar 24, 1999

-------
                                                                   DRAFT

    5.0    I/M ALGORITHM FOR START EMISSIONS


    5.1    General I/M Algorithm

           The MOBILE6 model will also compute I/M credit reductions for start emissions
    in addition to the running LA4 emissions.  The start I/M credits will be small in magnitude
    since the typical I/M test (i.e., EVI240, idle, etc) does not intentionally involve testing a
    vehicle during start or warm-up. The I/M credits for start emissions will reflect this fact
    by assuming that vehicles with high  start emissions are identified in conjunction with a
    running emission failure.

           The generalized structure of the start I/M credit algorithm is the same structure as
    used for the running LA4 emission credits (See Figure 1).  However, the Y-axis represents
    start emissions in grams and the X-axis represents mileage. Line A shows the basic start
    emission factor line before an I/M reduction. Line B shows the average start emissions of
    the normal emitting vehicles. Line C shows the average start emissions of the high emitting
    vehicles.
    5.2    I/M Start Emission Rates

           The basic emission rates  for start emissions (Line A of Figure 1) and the
    methodology used to develop them can be found in the EPA document "Determination of
    Start Emissions as a Function of Mileage and Soak Time for 1981-1993 Model Year Light-
    Duty Vehicles" - Report Number M6.STE.003.

           Table 4 contains the start emission regression coefficients for the normal emitting
    vehicles for all eight technology and model year groups. Table 5 contains the average start
    emissions from the high emitting vehicles (high  emitters are defined based on twice or
    thrice FTP standards - see Section 3.2). Table 6 shows the average after repair level of the
    high emitting vehicles.  The values shown in Table 6 are based on after repair emission
    testing. In these cases high emitting vehicles (high FTP emissions or EVI240 failures) were
    tested, repaired and retested. The analysis of the start emissions before and  after repair is
    discussed in detail in EPA document M6.IM.002  "Determining Repair Effects of EVI240
    Cold Start Emissions for 1981 and Later Light-duty Vehicles".
M6IM001.WPD DRAFT                 50                         Mar 24, 1999

-------
                                                      DRAFT
Table 4a
Regression Coefficients for START Emissions from Normal Emitter CARS
MY
Group

1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group

PFI
TBI
FI
Carb
Carb
FI
Carb
HC Coefficients
ZML
1.9987
1.9019
2.3589
1.4934
1.5892
2.3543
2.1213
DET
0.006830
0.002679
0.001388
0.018238
0.009408
0.008533
0.013610
CO Coefficients
ZML
18.972
19.233
19.949
24.698
24.442
20.038
28.637
DET
0.00703
0.00000
0.00000
0.10947
0.10577
0.22673
0.22673
NOx Coefficients
ZML
1.444
2.300
1.461
1.405
0.748
1.530
1.601
DET
0.00220
0.00000
0.00141
0.00000
0.00524
0.00059
0.00000
Table 4b
Mean START Emissions of Hish Emitter CARS
MY Group
1988-93
1988-93
1983-87
1986-89
1983-85
1981-82
1981-82
Tech
Group
PFI
TBI
FI
Carb
Carb
FI
Carb
HC Mean
4.829
3.293
5.313
10.520
10.520
5.313
10.520
CO Mean
38.06
27.16
65.31
92.82
92.82
92.82
92.82
NOx Mean
Same as Normals
Same as Normals
Same as Normals
Same as Normals
Same as Normals
Same as Normals
Same as Normals
M6IM001.WPD DRAFT
51
Mar 24, 1999

-------
                                                      DRAFT

MY
Group

1988-93
1988-93
1981-87
1984-93
1981-83
Table 5a
Regression Coefficients for START Emissions from
Normal Emitter Light Trucks
Tech
Group

PFI
TBI
FI
Carb
Carb
HC Coefficients
ZML
2.873
4.073
2.599
3.916
6.817
DET
0.00000
0.01309
0.00964
0.00854
0.00154
CO Coefficients
ZML
32.178
42.456
23.497
78.286
98.432
DET
0.0168
0.1411
0.0613
0.2564
0.3240

NOx Coefficients
ZML
1.597
4.294
1.384
0.143
1.082
DET
0.00000
0.00324
0.00000
0.00436
0.00000
Table 5b
Mean START Emissions of Hish Emitter Trucks
MY Group
1988-93
1988-93
1981-87
1984-93
1981-83
Tech
Group
PFI
TBI
FI
Carb
Carb
HC Mean
5.212
5.212
5.826
9.406
17.865
CO Mean
83.862
83.862
60.319
162.115
179.549
NOx Mean
Same as Normals
Same as Normals
Same as Normals
Same as Normals
Same as Normals
M6IM001.WPD DRAFT
52
Mar 24, 1999

-------
                                                                  DRAFT
Table 6
START Emission Regression Coefficients for High Emitters After Repair
Cars and Trucks
MY
Group

1990-93
1990-93
1986-89
1986-89
1983-85
1983-85
1981-82
1981-82
Tech
Group

PFI
TBI
FI
Carb
FI
Carb
FI
Carb
HC Coefficients
ZML
2.60
2.60
3.11
3.11
2.70
2.70
2.70
2.70
DET
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
CO Coefficients
ZML
18.90
18.90
30.05
30.05
28.33
28.33
28.33
28.33
DET
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.0000
NOx Coefficients
ZML
1.48
1.48
1.49
1.49
1.84
1.84
1.84
1.84
DET
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
0.00000
    5.3    Fraction of High and Normal Emitters in the Fleet

           The basic start emission factor is computed from a weighted average of the highs
    and normals. The fraction of high emitters (fraction of normal emitters = 1 - fraction of high
    emitters) in the fleet is the weighting factor. The fraction of high start emitters is the same
    fraction as the one used for the running emissions calculations.  Tables 3a and 3b and
    Appendix  A in EPA document M6.STE.003 "Determination  of Start Emissions as a
    Function of Mileage and Soak Time for 1981-1993 Model Year Light-duty Vehicles" show
    and explain the fraction of HC and CO high emitters in the fleet at selected mileages / ages
    for each pollutant. The fraction of NOx high emitters is not shown because for NOx the
    Normals and Highs are assumed to have the same emission rate (no start NOx highs are
    assumed to exist).
M6IM001.WPD DRAFT
53
Mar 24, 1999

-------
                                                                   DRAFT

    5.4    I/M Start Identification Rates

           The algorithm for start emissions is based on test data that indicates that a portion
    of the vehicles with high running emissions that are identified by the I/M process will also
    have high start emissions, and that these will be identified and corrected in conjunction
    with the repairs to pass the I/M test. Also, because significant NOx emissions usually form
    only after the vehicle is warm, it was assumed that an I/M program could only reduce HC
    and CO start emissions.

           A mathematical function that relates HC / CO cutpoint with the start emissions
    identification rate (IDR) was developed from the 910 vehicle sample used to develop the
    running emissions DDR.  The same methodology was used to develop the Start emission
    IDR as was used to develop the running emission IDR (See Section 3.9 for a more detailed
    explanation). This function also has the same range of HC and CO cutpoints (HC ranges
    from 0.50 g/mi to 5.0 g/mi and CO ranges from 5.0 g/mi to 100 g/mi) used in the running
    emission analysis. It predicts the percentage of start emissions from high emitters which
    are identified at a specific HC/CO cutpoint level. This is the percentage of the emissions
    from high emitters at Line C in Figure 1 that are reduced down to average fleet emission
    levels (Line A in Figure 1).  The statistical results are shown in Appendix D.  The functions
    are:

           StartHC IDR = 0.9814 - 0.1590*ln(HCCUT) - 0.1409*ln(COCUT)      Eqn32

           StartCO IDR=  1.1460 - 0.1593*ln(HCCUT) - 0.1707*ln(COCUT)      Eqn33

    5.5    Average Start Emissions After I/M

           The equation used to calculate the average start emissions after I/M is very similar
    in form to Equation 12a used to  calculate the average running emissions after I/M. Several
    of the parameters are the same such as the fraction of high emitters in the fleet, the waiver
    rate, the waiver repair percentage, and the non-compliance rate. The principal differences
    are the different IDR rates (the start IDRs are calculated in Equations 32 and 33), and the
    different after repair emission levels. Equation 34 is used to calculate the After I/M start
    emissions (S_EEVI).  S_IDR is the start emission IDR from Equations 32 and 33, and
    S_RLEV is the  after successful repair emission level (in units of grams). The variable
    S_RLEV is used in place of the variables N*R (normal emission level times the after repair
    emission level percentage) used in the running  emissions calculation.
           Equation 34 is used to calculate the average emissions of the fleet after I/M, and is
    used in the "sawtooth" methodology for I/M start emissions.
M6IM001.WPD DRAFT                  54                        Mar 24, 1999

-------
                                                                   DRAFT
           S_EIM  =      N*(1-X)  +    H*X*(1-S_IDR)  +    X*S_IDR*W*H*RW +
                        S_RLEV*X*S_IDR*FIX   +    H*x*s_iDR*NC             Eqn 34
    5.6    I/M "Sawtooth"

           The I/M credits for start emissions will also utilize the 'sawtooth' algorithm in the
    final calculation steps.  This algorithm  is virtually identical in structure to the ones
    presented in Section 4 for the running emissions. The structures used to model change of
    ownership and RSD are the same. Because the structure is the same, the methodology will
    not be repeated in this  section.  The only difference between the start and running
    algorithms are the actual emission rate parameters and values which are described in the
    previous sections. These include the normal and high emission levels, the IDRs, and the
    repair effects.

    5.7    Remote  Sensing and High Emitter Profile Start Emissions Parameters

           Currently, the same remote sensing and high emitter profile parameters will be used
    for the start emissions as were used for the running emissions. In the case of RSD this may
    introduce some error since RSD is defined to be a warm emission test, and is not designed
    to identify high  start emitters or screen out low start emitters. Presumably a high emitter
    profile which correctly profiles high and low start emissions can also be  developed.
    However, it is likely to differ from the one used for running emissions.
M6IMOO1.WPD DRAFT                 5 5                         Mar 24, 1999

-------
                                                                  DRAFT

    6.0    I/M Credits for Non-IM240 Tests
           The previous sections discussed the general algorithm and methodology used to
    develop the I/M credits for MOBILE6. The EVI240 test was used as the basis for the credits
    because of the large  amount of EVI240 data which are available to develop the IDR
    estimates and the after repair levels. I/M credits for other tests are also needed such as the
    Idle test, the 2500 RPM / Idle test,  and the  ASM tests.   The algorithm  used to
    mathematically implement these test types in MOBILE6 is analogous to the EVI240
    algorithm. The difference between the various I/M test types in MOBILE6 will be based
    on the differences in the IDRs for each test.
    6.1     Other I/M Tests

           The MOBILE6 model will also compute I/M credits for tests other than the EVI240
    test.  The test options which will be built into the model are (1) Idle test, (2) 2500 RPM
    / Idle test, (3) ASM tests, and (4) On-board Diagnostic (OBD) I/M tests. In addition,
    MOBILE6 will have the flexibility to model user defined test(s), or future test(s) which are
    currently unspecified.

           The default I/M tests in addition to the EVI240 test which MOBILE6 will able to
    model are:
    1.      Annual Two-Mode ASM 2525/5015 with Phase-in Outpoints
    2.      Annual Two-Mode ASM 2525/5015 with Final Outpoints
    3.      Annual Single-Mode ASM 5015 with Phase-in Cutpoints
    4.      Annual Single-Mode ASM 5015 with Final Cutpoints
    5.      Annual Single-Mode ASM 2525 with Phase-in Cutpoints
    6.      Annual Single-Mode ASM 2525 with Final Cutpoints
    7.      Annual Idle Test
    8.      Annual 2500 RPM / Idle Test
    9.      Biennial Two-Mode ASM 2525/5015 with Phase-in Cutpoints
    10.     Biennial Two-Mode ASM 2525/5015 with Final Cutpoints
    11.     Biennial Single-Mode ASM 5015 with Phase-in Cutpoints
    12.     Biennial Single-Mode ASM 5015 with Final Cutpoints
    13.     Biennial Single-Mode ASM 2525 with Phase-in Cutpoints
    14.     Biennial Single-Mode ASM 2525 with Final Cutpoints
    15.     Biennial Idle Test
    16.     Biennial 2500 RPM / Idle Test
M6IM001.WPD DRAFT                  56                        Mar 24, 1999

-------
                                                                   DRAFT
    6.2    ASM Tests

           Unfortunately, new paired ASM and FTP test data are not available on any ASM
    I/M tests in-order to compute new and specific IDR rates or repair effectiveness rates. As
    a result, the relative size of the I/M credits of these tests versus the EVI240 will remain the
    same between MOBILES and MOBILE6. This was accomplished by first computing the
    ratio of the MOBILES I/M credit value for an alternative ASM test over the MOBILES I/M
    credit value for the EVI240 at final cutpoints of 0.8 HC / 15 CO / 2.0 NOx. When done for
    each combination of model year, age and pollutant, this produces a large array of ratios (25
    ages x  18 model year x 3 pollutants). Rather than store all those ratios in the MOBILE6
    program, the ratio data were reduced by fitting it to a linear-quadratic equations using least
    squares regression. The independent variables in the regression were age and model year.
    The age range is from 1 to 25 and the model year range is from 81 through 98.  The 98
    model year credits will be used to represent all subsequent model years. The equation form
    is:

           ASM =      A * age + B * ageA2 + C * model year + D               Eqn 35

           Separate equation coefficients (A, B, C and D) were developed for each ASM test,
    cutpoint group, and pollutant. They are shown in  Tables 7a and 7b below.  Table 7a
    provides the coefficients for the Final ASM cutpoints and Table 7b shows the Phase-in
    ASM coefficients.  Within each of these tables different coefficients were also developed
    for vehicle ages which are less than or equal to 10 years, and greater than 10 years. These
    ratios are then multiplied by the MOBILE6 EVI240 IDR at the 0.80 HC / 15 CO / 2.0 NOx
    cutpoints to compute the MOBILE6 ASM DDR. This is done for both running and start
    IDRs.  After computation, the ASM IDR is used in Equation 12a to compute the ASM
    After I/M line, and the I/M credits.  Typically, the ASM ratioes which are applied to the
    EVI240 credits are in the range of 0.60 to 1.30. This may lower or boost the EVI240 credits
    by 0.30 times or raise by 0.40 times.  The lower ratios prevail for HC and CO emissions,
    and the higher ratios are occasionally seen for NOx emissions at the lower ages. Also, the
    ratios are typically very similar to each other within a given ASM test type and pollutant
    - generally ranging from 0 to 10 percent different within a model year group.

           The advantage of this approach is that it enables the ASM I/M test procedure credits
    to be easily  assimilated into the MOBILE6 I/M approach.  It also preserves a similar
    relative effectiveness of ASM versus EVI240 as was present in the MOBILES model. This
    is reasonable since no new ASM data are available in conjunction with FTP data to update
    the ASM credits.  One drawback of this approach is that it does not update the effect of
    different after repair levels, and assumes that the ASM after repair levels are the same as
    those for the EVI240. This means that the after repair levels for the 0.8/15/2.0 HC, CO and
M6IM001.WPD DRAFT                 57                         Mar 24, 1999

-------
                                                                 DRAFT

    NOx IM240 outpoints will be used for the final ASM outpoint after repair levels. Similarly,
    the 1.2/20/3.0 HC,  CO and NOx IM240 outpoints will be used for the phase-in ASM
    outpoint after repair levels.  Also, it assumes that the ratio between the ASM and EVI240
    credits in MOBILES based on FTP emissions can be equally applied for both running and
    start ASM credits in MOBILE6.
M6IMOO1.WPD DRAFT                 5 8                        Mar 24, 1999

-------
                                                      DRAFT

Table 7a
ASM I/M IDR Coefficients for Final ASM Cutnoints

Description
Test
ASM 2525
ASM 2525
ASM 2525
ASM 5015
ASM 5015
ASM 5015
ASM 2mod
ASM 2mod
ASM 2mod
Cutpoint
Phase -in
Phase -in
Phase-in
Phase-in
Phase-in
Phase -in
Phase-in
Phase-in
Phase -in
Pollutant
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx

Test
ASM 2525
ASM 2525
ASM 2525
ASM 5015
ASM 5015
ASM 5015
ASM 2mod
ASM 2mod
ASM 2mod
Cutpoint
Phase-in
Phase-in
Phase -in
Phase -in
Phase-in
Phase-in
Phase -in
Phase-in
Phase-in
Pollutant
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx

For AGE <= 10
CoeffA
0.001759
0.007035
-0.05733
-0.00494
0.003682
-0.12004
-0.006005
9.809e-04
-0.1461
CoeffB
1.588e-04
-1.826e-04
0.002875
5.766e-04
5.1103e-05
0.006165
6.0291e-04
1.5345e-04
0.007589
CoeffC
-0.001383
0.001893
-0.03234
-0.002577
0.001625
-0.02997
-0.001904
0.002478
-0.036311
CoeffD
0.9655
0.6641
4.1179
1.0894
0.6787
4.1908
1.0791
0.6573
4.9515
For AGE > 10
CoeffA
-0.005209
-0.002926
-0.001853
-0.005936
-0.003783
-0.004784
-0.004063
-0.002706
-0.005176
CoeffB
1.161e-04
6.517e-05
3.9017e-05
1.3247e-04
8.5342e-05
1.1196e-04
9.1144e-05
6.0543e-05
1.1762e-04
CoeffC
-0.001458
0.001498
-0.006412
-0.002217
0.001490
5.749e-04
-0.001374
0.002150
-0.002785
CoeffD
1.0459
0.7751
1.5271
1.1102
0.7627
0.9120
1.0614
0.7326
1.2899
M6IM001.WPD DRAFT
59
Mar 24, 1999

-------
                                                      DRAFT
Table 7b
ASM I/M IDR Coefficients for Phase-in ASM Cutnoints
Description
Test
ASM 2525
ASM 2525
ASM 2525
ASM 5015
ASM 5015
ASM 5015
ASM 2mod
ASM 2mod
ASM 2mod
Cutpoint
Phase-in
Phase-in
Phase -in
Phase -in
Phase-in
Phase-in
Phase -in
Phase -in
Phase-in
Pollutant
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx

Test
ASM 2525
ASM 2525
ASM 2525
ASM 5015
ASM 5015
ASM 5015
ASM 2mod
ASM 2mod
ASM 2mod
Cutpoint
Phase -in
Phase -in
Phase-in
Phase-in
Phase-in
Phase -in
Phase-in
Phase-in
Phase -in
Pollutant
HC
CO
NOx
HC
CO
NOx
HC
CO
NOx
For AGE <= 10
CoeffA
-0.0301
0.00171
0.0289
-0.03507
1.775e-05
-0.07537
-0.03397
-3.874e-04
-0.3024
CoeffB
0.002811
6.151e-04
-0.001775
0.003147
7.131e-04
0.003535
0.003077
7.258e-04
0.01462
CoeffC
-9.764e-04
0.00390
0.015844
0.002789
0.006458
5.805e-04
-2.039e-04
0.004660
-0.10688
CoeffD
0.7324
0.2676
-1.0368
0.4213
0.05215
0.9142
0.6986
0.2228
11.890
For AGE > 10
CoeffA
-0.01390
-0.00747
-0.00118
-0.01281
-0.007068
-0.005603
-0.01242
-0.00714
-0.01342
CoeffB
3.1742e-04
1.698e-04
7.0546e-05
2.945e-04
1.6312e-04
1.7994e-04
2.8523e-04
1.6475e-04
3.5207e-04
CoeffC
-0.001254
0.00331
0.00833
0.003960
0.005987
0.01033
6.932e-04
0.004418
-0.02177
CoeffD
0.8387
0.4557
-0.2571
0.3707
0.2188
-0.3290
0.6740
0.3669
2.8099
M6IM001.WPD DRAFT
60
Mar 24, 1999

-------
                                                                   DRAFT

    6.3    Idle and 2500RPM/Idle Tests

           The I/M credits for the Idle and 2500RPM/Idle tests were not developed like the
    ASM credits by ratioing the MOBILES Idle test results with the MOBILES IM240 results
    and applying the ratio to the MOBILE6IM240 results to get the MOBILE6 Idle test credits.
    Although, they could have been developed this way. Instead, the Idle and Idle/2500 RPM
    test credits were developed from a new analysis of the available paired Idle / 2500RPM/Idle
    and FTP data sources collected by EPA from 1981 through 1998.

    6.3.1   Available Data

           Two primary EPA datasets were available. The first dataset is called the "4MID"
    dataset.  The abbreviation "4MID" stands  for "Four Mode Idle dataset".  It contains
    virtually all of EPA's paired Idle and FTP data collected at EPA's various labs from 1981
    through 1998.  The four mode test is a special EPA Idle I/M test procedure developed for
    research work that simulates in-use Idle tests. The first mode is an unpreconditioned idle,
    the second mode is a 2500 RPM segment used to precondition the third  Idle mode, and
    used to pass or fail vehicles for the 2500RPM/Idle test. The third mode is a preconditioned
    Idle, and the fourth mode is an idle in drive mode.  Only the 2500 RPM mode and the third
    mode (pre-conditioned Idle) were used to develop the credits. Only the HC  emissions from
    the 2500 RPR mode were used in the development of the 2500RPM/Idle credits. The
    analogous CO 2500 RPM mode readings were not used because  of their tendency to
    produce false failures due to evaporative canister purge during the 2500 RPM mode. The
    preconditioned Idle test was used in both the Idle test and the 2500RPM/Idle test credits.
    The unpreconditioned Idle mode and the Idle in Drive modes were  not used for the I/M
    credit development.

           Test results from the Restart /Idle test used to test some early 1980's Ford vehicles
    were not used in this analysis due to their inconsistent availability in the dataset. The effect
    of this is thought to be very negligible. However, since the basis of the IDR consists only
    of High emitting vehicles, use of the Four mode test instead  of the Restart / Idle test for
    Ford vehicles could potentially overstate the Idle test credits slightly if the higher readings
    from the Four Mode test identify more high emitters that the Restart / Idle test would
    identify.

           The second primary dataset was the "IMLane" dataset.  It consisted of I/M lane Idle
    and 2500RPM/Idle test results from EPA's pilot I/M lane test program conducted in both
    Hammond, IN and Phoenix, AR by ATL. These data were paired with vehicle FTP data
    collected at ATL's laboratory. The test procedure consisted of a 2500RPM mode, and a
    subsequent preconditioned Idle mode. The unpreconditioned Idle and the Idle in Drive
    modes were not performed. The advantage of these data over the 4MID sample is that they
    were collected in an actual I/M lane rather than  in the  EPA laboratory  like the 4MID

M6IMOO1.WPD DRAFT                 61                         Mar 24,  1999

-------
                                                                 DRAFT

    sample.

          For the final results, both databases were combined together to produce overall IDR
    rates for the Idle test and the 2500RPM/Idle test.  Despite the slight differences in the I/M
    test procedures, the combination of the data makes sense for several reasons.   First, it
    produces a larger sample of vehicles. This is important because for this analysis only the
    High emitters are used to compute the IDRs, and the number of High emitters can get small
    in some model year groups. Also, both databases seem to complement each other in terms
    of model year coverage. For example, the "4MID" sample has a large preponderance of its
    data in the 1981 and 1982 model years; however, it does have some newer mid 1990's
    vehicles and trucks. The ATL sample on the other hand contains only cars, and is mostly
    represented by late 1980's to early 1990's cars.  Tables 8a and 8b show the model year and
    technology breakdown for both databases.
M6IM001.WPD DRAFT                 62                        Mar 24, 1999

-------
                                                      DRAFT


MY
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
Table 8a
Four Mode Idle / 2500RPM Idle and FTP Test Pairs

Cars
CARS
962
125
87
32
90
41
16
15
22






TBI
15
66
122
44
52
52
64
60
35
46
4
2
4


PFI
29
5
59
34
61
86
92
103
82
85
59
37
16
27
2

Trucks
CARS
120
45
10
48
63
17









TBI




13
23







1

PFI
4


1
6
41




2

2
1

M6IM001.WPD DRAFT
63
Mar 24, 1999

-------
                                                                 DRAFT


MY
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
Table 8b
IM Lane Idle / 2500RPM Idle and FTP Test Pairs

Idle Test
CARS
39
37
22
21
14
11
9
4
1
1



TBI
1
3
18
56
65
61
39
41
34
25
6
2

PFI
2
1
11
29
48
47
48
61
53
33
17
18
6

2500 RPM / Idle Test
CARS
39
37
22
21
14
11
9
4
1
1



TBI
1
3
18
56
63
61
39
40
34
25
5
2

PFI
2
1
10
29
47
47
48
60
53
33
17
18
6
    6.3.2  Idle and 2500RPM/Idle Test IDRs

          The calculation of the IDRs for the Idle and 2500RPM/Idle tests is very similar to
    the calculation done for IM240 IDRs in Section 3.9. One difference is that IDRs for a range
    of cutpoints was not performed. Instead only one set of Idle and 2500RPM/Idle cutpoints
    were developed. These were at the CO/HC cutpoints of 1.2%CO and 220ppm HC.  Also,
    IDRs for only HC and CO emissions for  running and start were  developed. Idle and
    2500RPM/Idle IDRs for NOx emissions were not developed. Neither the Idle Test or the
    2500RPM/Idle test will produce NOx benefits or NOx "Dis-benefits" for MOBILE6. In
    comparison, MOBILES contained NOx "Dis-benefits" if an Idle or 2500RPM Idle test were
    performed.
M6IM001.WPD DRAFT
64
Mar 24, 1999

-------
                                                                 DRAFT



Test
Idle
2500/Idle

Test
Idle
2500/Idle
Table 9a
Idle and 2500RPM / Idle Test IDRs for Each Sample


IDRs Based on I/M Lane Sample
Hot Running LA4 HC
Carb
63.3
76.5
PFI
58.7
59.3
TBI
53.2
53.9
Cold Start HC
Carb
41.9
48.6
PFI
39.1
40.2
TBI
33.9
34.8
Hot Running LA4 CO
Carb
54.9
68.8
PFI
57.5
57.5
TBI
60.6
60.6
Cold Start CO
Carb
29.1
29.1
PFI
23.6
23.6
TBI
20.9
20.9



Test
Idle
2500/Idle

Test
Idle
2500/Idle
IDRs Based on Four Mode Sample
Hot Running LA4 HC
Carb
48.8
66.1
PFI
74.3
74.3
TBI
52.2
61.6
Cold Start HC
Carb
20.2
24.4
PFI
42.6
42.6
TBI
17.7
25.4
Hot Running LA4 CO
Carb
53.4
63.8
PFI
81.1
81.1
TBI
40.7
55.7
Cold Start CO
Carb
21.4
27.1
PFI
57.8
57.8
TBI
30.1
33.9
          Table 9a shows the Hot Running LA4 and Cold Start DDR rates for the Idle and
    2500RPM/Idle tests for each of the two datasets.  It is further broken down into three
    technology groups. These are Carbureted, Throttle Body Injection (TBI), and Ported Fuel
    Injection (PFI). The IDRs were not made a function of model year because of the small
    sample sizes in many individual model years.  Table 9b shows the DDR results for the
    combined dataset. The two datasets were combined together based on total emissions from
M6IM001.WPD DRAFT
65
Mar 24, 1999

-------
                                                                   DRAFT

    the high emitters rather than on the number of vehicles in the sample. The IDRs are shown
    as a percentage in both tables, but will be programmed into MOBILE6 as fractions.  They
    represent the fraction of emissions from  high  emitters which  are identified by the
    prospective I/M test. Separate IDRs for each pollutant and technology were developed for
    Hot Running LA4 emissions and Start emissions based on Bagged FTP data.
Table 9b
Idle and 2500RPM / Idle Test IDRs Based on the COMBINED Sample



Test
Idle
2500/Idle

Test
Idle
2500/Idle
IDRs Based on I/M Lane Sample
Hot Running LA4 HC
Carb
54.6
70.2
PFI
63.5
63.9
TBI
52.8
56.8
Cold Start HC
Carb
25.5
30.3
PFI
40.8
41.3
TBI
29.5
32.3
Hot Running LA4 CO
Carb
54.0
65.9
PFI
63.0
62.9
TBI
53.5
58.8
Cold Start CO
Carb
23.3
27.6
PFI
37.8
37.8
TBI
25.1
26.8
    6.3.3   After Repair Emission Level for Idle and Idle/2500 Tests

           The Idle Test after repair emission levels for MOBILE6 were calculated from a
    dataset which was used for MOBILES development.  It consisted of 36, 1981 and later
    vehicles which initially failed the idle test, were repaired, and passed the final idle test at
    standard cutpoints.  These data were collected as part of an EPA test program conducted
    to evaluate the effect of repair on idle test failures. The repairs were conducted by qualified
    technicians. The vehicle sample mean FTP emission values after Idle test I/M repair were
    found to be 1.89 g/mi HC and 19.49 g/mi CO.  These compare with means of 1.26 g/mi
    HC and 13.46 g/mi CO for the EVI240 at the 1.2/20 HC and CO cutpoint. Idle test repair
    effects for NOx emissions  are not computed because MOBILE6 will not give NOx benefits
    or disbenefits to an idle test program.

           The ratio of the idle test after repair FTP emission level to the EVI240 after repair
    FTP emission level at 1.2/20/3.0 cutpoints is computed from the data and used to generate
M6IM001.WPD DRAFT
66
Mar 24, 1999

-------
                                                                   DRAFT

    the after repair idle test emission level for running LA4 emissions. A consistent ratio based
    on the FTP will be used for all mileages, vehicle types, and model years. The ratios which
    are used for HC and CO are:

           HC Ratio:    1.89 g/mi / 1.26 g/mi        =      1.50
           CO Ratio:    19.49 g/mi / 13.46 g/mi      =      1.45

           They are used in MOBILE6 to generate the idle test after repair running LA4
    emission level by  multiplying the ratio  by the EVI240 after repair emission level at
    1.2/20/3.0 cutpoints. The same after repair emission levels will be used for the Idle test and
    the Idle/2500 RPM test.
    6.4    OBD I/M Tests

           This document does not explicitly cover vehicles which are equipped with an OBD
    system. However, most OBD equipped vehicles will continue to receive exhaust based I/M
    tests such as the EVI240 or the Idle test for much of their early lives.  Thus, the topic is
    mentioned briefly in this document as an introduction. For more complete details on EPA's
    modeling of OBD  equipped vehicles (1996+ model years) please read EPA document
    M6.EXH.007 "Determination of Emissions, OBD, and I/M Effects for Tierl, TLEV, LEV,
    and ULEV Vehicles".

           The OBD system is an electronic diagnostic system built into most 1996 and later
    and some 1994 and 1995 model year vehicles. It is designed to (1) continuously monitor
    the performance of the car's emission control system, and detect serious problem(s) which
    cause the vehicle's FTP emissions to exceed 1.5 times its applicable certification standards,
    (2) register a code in the vehicle's computer and turn on a dashboard warning light to notify
    the owner. The  system will also have the capability to be electronically accessed in an I/M
    lane. The vehicle will be required to pass the OBD test (no trouble codes are present)  in-
    order to pass the state I/M program requirements.

           In MOBILE6 an I/M program conducting an OBD check on properly equipped OBD
    vehicles will be assigned an DDR of 90 percent (fraction 0.90). This value will be given
    regardless of whether an exhaust I/M test such as the EVI240 or the ASM test is performed
    or not performed.  Also, the with and without technician training  levels in an OBD I/M
    program  will be equivalent.   It is assumed that  the technicians specializing in OBD
    diagnosis and repair will either be fully qualified, or not involved in the industry.
M6IM001.WPD DRAFT                 67                         Mar 24, 1999

-------
                                                    DRAFT
                            APPENDIX A
               Running LA4 Emissions from 1990-93 MY PFI Normal Emitters
             Figure A-1
             HC Emissions from  Normals
                  20
             MILEAGE
40
60
                     80
100
120
140
160
180
200
            Figure A-2
            CO Emissions from Normals
         20.
         18.
         16.
         14.
         12.
         10.
          8.
          6.
          4.
0    20    40    60    80    100   120   140
            MILEAGE
                        160
                                  180
                                                              200
M6IM001.WPD DRAFT
         68
                         Mar 24, 1999

-------
                                               DRAFT
             Figure A-3
             NOx Emissions from Normals
20
             MILEAGE
40
60
80
100
         120
                                          140
160  180
200
M6IM001.WPD DRAFT
69
             Mar 24, 1999

-------
                                                                              DRAFT

                                         APPENDIX B
                                          Sample Calculations

     Sample calculation for determining percentage of HC running emission Highs in the fleet at age = 5 for 1990-
     1993 PFI technology.

     Calculating Line A in Figure 1 (basic emission rate)

     X< 21.27       A = 0.0508 +0.0013 * mile

     X > 21.27       A = 0.0508 + 0.0013 * mile + (mile - 21.27)  * 0.0023

     from Table 3,  mile = 49.835

     X is the inflection point of the basic emission rate (thousand mile units).
     See the document "Determination of Running Emissions as a Function of Mileage for 1981-1993 Model
     Year Light-Duty Vehicles."

                    A = 0.181g/miHC

     Calculating Line B in Figure 1 (normal emitter rate)

                    B = 0.0249+ 0.00113* mile

                    B = 0.081g/miHC

     Calculating Line C in Figure 1 (high emitter rate)

                    C= 1.367 g/miHC

     Calculating Line D in Figure 1 (After I/M repair emission rate)

                    D = ( 2.24 - 0.07595 * Age ) * B

                    0=1.86*0.081=0.151

     Calculating percentage of Highs from equation 7.

                    Highs = (A - B) / © - B)

                    Highs = 0.078 = (0.181 - 0.081) / (1.367 - 0.081)

                    % Highs = 7.8 percent
M6IM001.WPD DRAFT                     70                             Mar 24, 1999

-------
                                                       DRAFT
                             APPENDIX C
         Periodic I/M and RSD / Change of Ownership Sawtooth Illustrations
                   Figure 1 - Annual I/M with RSD
       SSI
       Uj
          O
          0
               ^CMf/J^lA
c«Mf>/r^      5e<5M£(JrZ8
     •ffSMo/Tie /      /-'-f.sMEAjr^ 8
      '/Z
M6IM001.WPD DRAFT
        71
Mar 24, 1999

-------
                                                       DRAFT
                       Figure 2 - Biennial I/M with RSD
                  2  i  4  5  &  1  9  9  10  il  II  13  14-
        YEAfiS
   4
M6IM001.WPD DRAFT
72
Mar 24, 1999

-------
                                                                  DRAFT
                                     Appendix D

         Description of the FORTRAN Algorithm Adopted from EPA Document
                                      M6.IM.001
                    Used to Code the I/M Methodology in MOBILE6
      TASK:2-663                                                     DRAFT:
                                       12/01/99
                                   DynTel Report:
       Mobile 6 IM Benefits Methodology for 1981 through 1993 Model Year Light
                                       Vehicles

                              Employee: Robert Ducharme

    1. Introduction
       In most inspection and maintenance programs vehicles  are inspected annually or
    biennially. However, some inspections are prompted by special events such as change of
    ownership (COEVI) or identification of high emitting vehicles using a remote sensing device
    (RSD). The objective of this report is to derive an equation for the emissions from a fleet
    of vehicles that is subject to both periodic and selective (RSD+COEVI) EVI programs. The
    most general form of this equation allows for an arbitrary period N between inspections and
    an arbitrary grace period GPRD before each vehicle receives its first test. However, both
    N and GPRD must be an integer number of years.

          For the purposes of modeling, light duty gasoline vehicles and trucks are classed
    either as normal or high emitters. High emitters are the vehicles with broken emission
    control systems. The influence of inspection and maintenance programs is to reduce the
    basic exhaust emission levels from high emitting vehicles compared to what they would be
    if no EVI program were in force. No EVI correction is required for normal emitting vehicles.

          This report is based on the inspection and maintenance methodology described in
    the US EPA draft report M6.EVI.001 though there are some differences. One such point of
    departure is that the EVI  sawtooth methodology from Mobile 5 is replaced  using  an
    algebraic approach for calculating the benefits of EVI programs that does not require the use
    of sawtooth diagrams. This has led to two refinements of the EPA model. Firstly, the basic
    emission factor lines drawn straight in the sawtooth diagrams are slightly curved in reality.

M6EVI001.WPD DRAFT                  73                        Mar 24, 1999

-------
                                                                   DRAFT

    This curvature has now been taken into account. Secondly the sawtooth method applies the
    benefits of periodic IM testing on a single day (October  1) during each relevant twelve
    month test period preceding the emissions evaluation date. Here, a continuous model is
    used that assumes vehicles are always tested on the anniversaries of their sales. It is further
    assumed that new  vehicle sales are uniform throughout each  model year so that the
    distribution of the anniversaries of those sales is also uniform in future model years. The
    notation used in this report is not month specific so that the mathematical formulation is
    equally applicable for both January 1 and July 1 calculations.

    2. IM240 tests

      EPA have worked out an explicit equation for the quantity that must be subtracted from
    the basic emission factor of any high emitting 1981-1993 model year light duty gasoline
    vehicles and trucks  in order to take into account the benefits of having an JM240 program
    in force. There is no IM correction for normal emitting vehicles. The form of this correction
    factor is readily deduced from  eqn 12a and eqn 34 (ref M6.IM.001) to be:

           IMCF(JDX,AIM) = XIM(JDX,AIM)*  IDR

           *[HIM(JDX, AIM) * (W * R W + NC - 1) + A * FIX ]            (1)
    where the symbols have the following meaning

           AIM:  Integer age of a vehicle in years on the date of its IM test previous to the
                 emissions evaluation date. It is assumed that vehicles are always tested
                 on their anniversaries of their sales.
           JDX:  Integer model year index of a vehicle referred to the year ICY and month
                 MEVAL of the emission factor calculations.
           XIM:  Fraction of the fleet composed of high emitters on the date of the IM test.
           HIM:  High average emission factor on the date of the IM test.
           DDR:  Fraction of all the high emitters in a target group identified by an IM test.
           W:    Fraction of all the identified high emitters that get a repair cost waiver.
           RW:  Fraction of the high emitter level that waived vehicles are repaired after
                 IM.
           NC:   Fraction of identified high emitters which are in non-compliance of the
                 IM program.
           FIX:  Fraction of identified high emitters which get repaired to pass the test.
           A:    Average emission level from vehicles after they have been repaired and
                 passed an IM test.

M6IM001.WPD DRAFT                 74                         Mar 24, 1999

-------
                                                                   DRAFT

    Eqn (1) normally has a negative value and applies for both running (ISR=1) and start
    (ISR=2)  emissions. However, the calculation of the after  repair level  A  and the
    identification rate IDR is different depending on the value of ISR. The details of these
    calculations can be found in M6.EVI.001 together with default values for R, RW and NC.
    The value of FIX is 1-W-NC. The calculation of XEVI and HIM are discussed in section 4.

    3. Other IM Tests Types

      Eq. (1) is valid for the following additional EVI test types.
           1. Two-Mode ASM 2525/5015 with phase-in cutpoints
           2. Two-Mode ASM 2525/5015 with final cutpoints
           3. Single-Mode ASM 5015  with phase-in cutpoints
           4. Single-Mode ASM 5015  with final cutpoints
           5. Single-Mode ASM 2525  with phase-in cutpoints
           6. Single-Mode ASM 2525  with final cutpoints
           7. Idle test
           8. 2500 RPM/Idle test

    The test type affects the benefit through the high emitter identification rate IDR and the
    after repairs emission level A. IDR is also corrected for RSD  clean screening and high
    emitter profiling. The maximum allowable value of IDR is 0.9. The after repairs emission
    level A includes a  correction for technician  training.  It cannot be higher than  the high
    emission factor or lower than the normal emission factor.
    4. IM Emission Factors

       The EPA EVI methodology for periodic (annual, biennial, triennial etc.) EVI programs
    makes two simplifying assumptions.
           17.    All vehicles are tested on the anniversaries of their sales.
           18.    Vehicle sales are uniform throughout each model year.
    The assumption that EVI tests always tested on the anniversaries of their sales will have to
    be revised when COEVI and RSD prompted testing is considered later.

           The normal (INH=1) and high (INH=2) basic emission factors for a vehicle of
    model year MY on the date of an EVI test is
           BIM(INH,MY,AM) =ZML(INH,MY) +KIM(AIM) *DR(INH,MY)          (2)

    where ZML is the zero mile level, DR is the deterioration rate, KEVI is the vehicle miles


M6EVI001.WPD DRAFT                 75                         Mar 24, 1999

-------
                                                                 DRAFT

    traveled in units of thousands of miles for a vehicle of age AIM. Note, MY=ICY-JDX+1
    where JDX has previously been defined as the model year index referred to the evaluation
    year ICY.  Default values for ZML and DR are stored in the mobile model. The value of
    KIM for AEVI>0 is readily calculated from the expression
                                                                            (3)
    where AMAR(I) is the annual mileage accrual rate also stored in Mobile. It is a reasonable
    approximation to assume the vehicle has traveled zero miles when its first owner aquires
    it so KIM(0)=0.0.

           The HIM variable defined in section 2 is:

           HIM =BIM(INH=2,MY,AM)

    The default value of the deterioration rate for highs emitters in Mobile is zero but the user
    can override this default.

           The probability that a vehicle of age AIM will be a high emitter is XIM(MY, AGE).
    This quantity  like  BEVI can also be expressed  exclusively as  a function of MOBILE 6
    regression coefficients and KIM.


    5. The NO IM case

      Mobile 6 calculates emissions on January 1 or July 1. The existing method of
    evaluating the uncorrected (FTP) basic emission factors on these dates is through the
    equation

           Normals: NOIM(INH=1,MY,JDX)=BEV(INH=1,MY,JDX) *(1-XEV(MY,JDX))

           Highs:   NOIM(INH=2,MY,JDX) =BEV(INH=2,MY,JDX) *XEV(MY,JDX)
                                                                            (4)
    where
           BEV(INH,MY,JDX) =ZML(INH,MY)+KMILES(JDX)*DR(INH,MY)      (5)

    and XEV is the fraction of high emitters in the fleet on the evaluation date. Expression (5)
    is identical to  (2) except the model year index is now calculated from the evaluation date
    and KMTLES  has replaced KIM where KMTLES is the  average vehicle mileage on the
    evaluation date. Similar arguments apply to XEV and its XEVI counterpart. The method of

M6IM001.WPD DRAFT                 76                        Mar 24, 1999

-------
                                                              DRAFT

    adjusting vehicle mileage for the month of evaluation is described in AP42 for the special
    case of January 1 emissions. This method has since been extended to treat the July 1 case

FRN
JDX=4
FRC
FRN
JDX=3
FRC
FRN
JDX=2
FRC 1
JDX=lJ
                                      Evaluation
                                      date.
    Figure 10 Shows the partitioning of each model year into segments.

    as well. The general formula for KMTLES expressed in terms of KJJVI is
          KMILES(l) = -FRC * KIM (I)

      KMILES(JDX) = KIM(JDX)+-FRC2 * [KIM(JDX + 1) - KIM(JDX)]

                      --FRN2 * [KIM(JDX)-KIM(JDX -1)]  	(6)
    where FRC is the elapsed fraction of a year since the model year changed on October 1 and
    FRN=1-FRC. The fraction FRC is readily calculated in terms of the month of evaluation
    MEVAL using the algorithm

            DIFF = MEVAL -10
            FRC(DIFF >0) = DIFF 112
            FRC(DIFF < 0) = 1 - DIFF 112
M6JM001.WPD DRAFT
77
Mar 24, 1999

-------
                                                                 DRAFT

    Thus, FRC=0.25 on January 1 and FRC=0.75 on July 1.

    6.0   Annual I/M Inspection
    In annual JM programs vehicles are inspected every year on the anniversary of the sale to
    their first owner. For generality, it will be assumed that vehicles do not receive their first
    inspection until they have been in operation for GRPD years. Consequently, all vehicles
    with model year index greater than GRPD+1 should receive one inspection in the twelve
    month period preceding the date when the emissions are to be evaluated.
    An essential concept in analyzing periodic JJVI programs is the fact that the age of a vehicle
    AIM on the date of its previous EVI test is not a unique function of its model year index.
    However, it is possible to partition each model year into two segments in such a manner
    that a unique value of AIM can be assigned to each segment. This breakdown is done next
    for the case of an annual EVI inspection program with a grace period of 1 year. The value
    of AIM can be determined for each model year and model year segment with the help of
    figure 1. For example, if emissions are to be evaluated in 1990 from a  1988 model year
    vehicle then JDX=3 will be the model year index of the vehicle. The JDX=3 model year
    can be divided, as can any other year, into FRC and FRN segments. Therefore, suppose that
    the vehicle was purchased new in the FRC model year segment. The choice of JDX=3 and
    the FRC segment give the starting point in the diagram. It is then simply a question of
    counting forward an integer number of years (AIM) until the date of the previous EVI test
    before the evaluation date is found. The result is AIM=2. Table 1 shows the value of AEVI
    for other values of JDX.
          JDX          SEGMENT         AIM
           1              FRC               0
           2              FRN               0
           2              FRC               1
           3              FRN               1
           3              FRC               2
           4              FRN               2
           4              FRC               3
          JDX            FRN              JDX-2
          JDX            FRN              JDX-1

          Table 1. Vehicle ages in annual EVI programs.


M6EVT001.WPD DRAFT                 78                        Mar 24, 1999

-------
                                                                 DRAFT

    The above table gives sufficient information to calculate periodic EVI corrections for any
    model year. For example, for JDX=3 it can be seen that the correction is

          PIMCF(JDX=3)=FRN*IMCF(JDX=3,AIM=1)+FRC*IMCF(JDX=3,AIM=2)
          (7)

          In completion of the annual EVI program problem it is necessary to treat arbitrary
    values of the grace period. The JDX
-------
                                                                   DRAFT

           2. Move forward an integer number of years equal to the grace period.
           3. If you have moved beyond the emissions evaluation date set AIM1=0.
           4. Else move move forward in steps of N years until the date of the
             previous EVI test. In this case, AEVI1 is equal to the total number of years
             between the purchase date and previous EVI test date for the vehicle.

    The algorithm for calculating AEVI2 is identical except that step  1 begins in the FRC
    segment of the JDXth model year. The complication of vehicles that are too young to have
    received their first test is readily handled using the convention EVICF(JDX, AEVI=0)=0. One
    simple test of this algorithm is to reproduce the results in section 6 for an annual EVI
    program. It is also of interest to calculate the vehicle ages for biennial programs with grace
    period of 1 and 2 years. These results are given in table 2.

           JDX         SEGMENT                AIM

           1               FRC
           2               FRN
           2               FRC
           3               FRN
           3               FRC
           4               FRN
           4               FRC
           5               FRN
           5               FRC
           6               FRN
           6               FRC

           Table 2. Vehicle ages in biennial EVI programs.
    8. Selective IM programs

      In periodic EVI programs all vehicles are tested every N years following an initial grace
    period GPRD years after they were first bought into the fleet. In selective EVI programs,
    vehicles are only tested if they meet certain criteria such as a recent change of ownership
    (COEVI) or detection as a high emitter using a remote sensing device (RSD). Selective EVI
    testing is  usually  done in areas where a periodic program is also  in operation. One
    important  difference between selective  and periodic programs is that a vehicle can be
    tested at any time  during the year rather than just on the anniversary of its  purchase.
    However,  for modeling purposes it will be assumed that selective EVI tests only affect
    vehicles of integer and half-integer ages.
M6EVI001.WPD DRAFT                 80                         Mar 24, 1999
GPRD=1
0
0
1
1
1
1
O
3
O
3
5
GPRD=2
0
0
0
0
2
2
2
2
4
4
4

-------
                                                                 DRAFT

          Let PSTE(JDX, AGE) and PSTL(JDX, AGE) denote the respective probabilities that
    a FRC  and FRN  segment vehicle  with   model  year  index JDX and  age AGE
    (=0.5,1.0,1.5....) will have an EVI test as a result of identification in the previous six months
    by either a COEVI or RSD program. These quantities can be defined more precisely as

          PSTE(JDX,AGE)=FSTE(JDX,AGE) *STR(JDX,AGE)                   (10)

          PSTL(JDX,AGE)=FSTL(JDX,AGE) *STR(JDX,AGE)                   (11)

    where F STL and FSTE are the FRC and FRN segment fractions of all the JDX model year
    vehicles eligible by virtue of having reached the age AGE for a selective EVI test and

          STR(JDX,AGE)=RSD(JDX,AGE)+COIM (JDX,AGE)                  (12)

    is the normalized probability that an eligible vehicle will be selected for a test as a result
    of change of ownership or detection by a remote sensing device. For example, if
    PSTE(JDX=3,AGE=1.5)=0.01
    then 1% of all the JDX=3 vehicles will receive an JJVI benefit EVICF(JDX=3,AGE=1.5) as
    a result of the fact they were all purchased new in the same FRC model year segment and
    tested at the same age of 1.5 years.

          Selective EVI tests only benefit vehicle emissions if they take place after the vehicles
    previous periodic EVI test. All FRC segment vehicles

          FSTE(JDX,AGE) = FRC                                           (13)

    will be eligible for selective EVI tests for integer and half integer values of AGE in the range

            AIM +0.5< AGE  < JDX-I                                  (14)

    providing AEVKJDX-1. This result can be deduced from figure 1.  If FRC>0.5 then a
    fraction

          FSTE(JDX,AGE) = FRC-0.5                                        (15)

    of the JDX model year vehicles will also be eligible for one additional test at age JDX-0.5.
    Eqns (13) and (15) give the only nonzero values of FSTE. The arguments pertaining to the
    FRN segment vehicles are similar. In particular, all FRN segment vehicles

          FSTL(JDX,AGE) = FRN                                           (16)
M6EVIOO1.WPD DRAFT                 81                        Mar 24, 1999

-------
                                                                 DRAFT

    will be eligible for selective IM tests for integer and half integer values of AGE in the range

           AIM +0.5< AGE < JDX-2                                   (17)

    providing AEVKJDX-2. If FRN<0.5 then all FRN segment vehicles will also be eligible for
    an additional test at age JDX-1.5. Otherwise if FRN>0.5 then only a fraction

          FSTL(JDX,AGE)=0.5                                       (18)

    of the JDX model year vehicles will be eligible.

          The probability CPSIME(JDX,AIM) that a vehicle purchased new in the FRC
    segment of the JDX model year receiving a selective EVI  test between its previous periodic
    EVI test date at age AIM and the emissions evaluation date at the end of the FRC segment
    of the JDX=1 model year is equal to the sum over PSTE(JDX,AGE) for all the values of
    AGE satisfying equations (14) and (15). This is given by

                                 M ,
          CPSIME(JDI,AIM)= ^[PSTE(JDX,AGE = AIM +0.5*M)]       (19)
    where ME=2*(JDX-AEVI-0.5) is the maximum number of possible selective EVI test dates.
    The arguments for vehicles in the FRN model year segment are similar with all the possible
    values  of the AGE variable  calculable from  eqns. (17) and (18). This leads to the
    probability
           CPSIML(JDX,AIM) = ^[PSTL(JDX,AGE = AIM +0.5* M)]       (20)
    with ML=2*(JDX-AIM-1.5). It is a further requirement that vehicles cannot receive the
    benefits from more than one EVI test. Consequently, the values of CPSEVIE and CPSIML
    cannot exceed FRC and FRN respectively.

          The arguments  in this section up to here have been quite formal. It is therefore
    instructive to once again consider an example. Consider the case of a biennial EVI program
    with selective EVI testing and a grace period of one year. Let us set JDX=4 and select a
    vehicle that was bought new in the FRN model year segment.  Table 2 indicates that this
    vehicle will have received its previous
    periodic EVI test at age AIM=1 year. From eqn (20) the probability that this vehicle will be
    tested as a result of the  selective EVI program is

M6EVT001.WPD DRAFT                 82                        Mar 24, 1999

-------
                                                               DRAFT

          CPSIML(JDX=4,AIM=1) =PSTL(JDX=4,AGE=1.5) + PSTL(JDX=4,AGE=2.0)
                                 + PSTL(JDX=4, AGE=2.5)

    By contrast, the vehicles belonging to the FRC segment of the JDX=4 model year segment
    are all old enough to have received their second periodic EVI test at age AEVI=3. In this case
    eqn (19) gives

          CPSIME(JDX=4,AM=3) = PSTE(JDX=4,AGE=3.5)

    where this  expression contains only one term because a relatively short period elapses
    between the date when the vehicles receive their  periodic EVI test and the emissions
    evaluation date.
          Each PSTE(JDX,AGE) term in eqn (19) gives the probability that a vehicle bought
    new in the FRC segment of the JDX model year will receive a selective EVI test at age AGE.
    The benefit that arises from such a test is therefore PSTE(JDX, AGE)*EVICF(JDX, AGE).
    These benefits can therefore be summed over all the possible selective EVI test dates to give
    the total benefit from all the selective EVI tests carried out on the FRC segment vehicles to
    be
    SIMCFE(JDX,AIM) =
    Mg
    ^[PSTE(JDX,AGE = AIM + 0.5 * M) * IMCF(JDX,AGE = AIM +0.5*M)]
    M=l

                                                                         (21)

    The total benefit from all the selective EVI tests carried out on the FRN segment vehicles
    is then similarly




    SIMCFL(JDX,AIM) =
    M L
    ^[PSTL(JDX,AGE = AIM + 0.5 * M) * IMCF(JDX,AGE = AIM +0.5*M)]
                                                                         (22)
    Here, EVICF is evaluated using the vehicle mileage equation

M6IM001.WPD DRAFT                 83                       Mar 24, 1999

-------
                                                                DRAFT

                                                    j
             KIM (J + 0.5) = 0.5* AMAR(J + 1) + ^AMAR(I)           (23)
                                                   7=1

    for all half-integer ages.


    9. Calculation of IM Benefits

        Suppose a selective IM program is operating alongside the annual program.  The
    generalization of eqn. (9) to  include the effect of the selective testing is then

          PMCF(JDX)=(FRN-CPSML(JDX,AM1)) *IMCF(JDX,AM1)
                        + SIMCFL(JDX,AM1)
                        + (FRC-CPSIME(JDX,AM2)) *IMCF(JDX,AM2)
                        + SIMCFE(JDX,AIM)                                (24)

    There are two points to note. Firstly, the earlier and later SEVICF terms are included to
    account for the benefits of the selective EVI tests that take place after the periodic EVI tests.
    Secondly, the CPSEVI terms are subtracted from the FRC and FRN fractions so that the
    selectively tested vehicles do not also receive a benefit for their earlier periodic test.

          It is instructive to evaluate eqn. (24) for the correction to July 1 emissions arising
    from an annual EVI program  with COEVI. Let JDX=3. Annual I/M programs are treated in
    section 6 where the values  AEVI1=1 and AIM2=2 can be read from table 1. With this
    information, the arguments in section 8 then give the cumulative probabilities and selective
    I/M correction factors for this problem to be

          CPSIME(JDX=3,AM=2)=0.25 *STR(JDX=3,AGE=2.5)

          CPSIML(JDX=3,AM=1)=0.25*STR(JDX=3,AGE=1.5)

          SIMCFE(JDX=3,AM=2)=CPSIME(JDX=3,AM=2)* MCF(JDX=3,AGE=2.5)

          SIMCFL(JDX=3,AM=2)=CPSML(JDX=3,AM=1)* MCF(JDX=3,AGE=1.5)

    having set FRC=0.75  and FRN=0.25. Here, the  value of STR depends on the rate of
    selective testing. For example, if a COEVI program  is in operation in an area with 16% per
    annum change of ownership then STR=0.08. In this case, eqn (24) simplifies to

          PMCF(JDX=3)=0.23*IMCF(JDX=3,AGE=1)


M6EVI001.WPD DRAFT                 84                        Mar 24, 1999

-------
                                                            DRAFT

                      + 0.02* IMCF(JDX=3,AGE=1.5)
                      + 0.73*IMCF(JDX=3,AGE=2)
                      + 0.02* IMCF(JDX=3,AGE=2.5)

    where IMCF can be calculated directly from eqn (1) for each of the 4 vehicle ages.
M6IM001.WPD DRAFT               85                      Mar 24, 1999

-------
                                                                                DRAFT



                                           APPENDIX E

           Statistical Diagnostics for Running Emissions IDR Determination


     -> REGRESSION
     ->   /DESCRIPTIVES MEAN STDDEV CORR  SIG N
     ->   /MISSING LISTWISE
     ->   /STATISTICS COEFF OUTS CI R ANOVA
     ->   /CRITERIA=PIN(.05) POUT(.IO)
     ->   /NOORIGIN
     ->   /DEPENDENT hcrun_id
     ->   /METHOD=ENTER ln_hccut ln_cocut


               ****  MULTIPLE   REGRESSION   ****

     Equation Number 1   Dependent Variable..   HCRUN_ID   HCRun ID

       Descriptive Statistics are printed  on Page    2

     Block Number  1.  Method:  Enter     LN_HCCUT LN_COCUT


     Variable(s) Entered on Step Number
        1..    LN_COCUT
        2 . .    LN_HCCUT


     Multiple R           .90947
     R  Square             .82713
     Adjusted R Square    .82246
     Standard Error       .06411

     Analysis of Variance
                        DF      Sum of Squares      Mean Square
     Regression          2            1.45516           .72758
     Residual            74              .30413           .00411

     F  =     177.03226      Signif F =   .0000


     	 Variables in  the Equation 	

     Variable             B        SE B     95% Confdnce Intrvl B      Beta

     LN_HCCUT       -.136503     .010483     -.157390    -.115615   -.629362
     LN_COCUT       -.106888     .007869     -.122568    -.091209   -.656531
     (Constant)     1.145095     .026063     1.093164    1.197027

     	 in	

     Variable           T Sig T

     LN_HCCUT     -13.021  .0000
     LN_COCUT     -13.583  .0000
     (Constant)    43.936  .0000


     -> REGRESSION
     ->   /DESCRIPTIVES MEAN STDDEV CORR  SIG N
     ->   /MISSING LISTWISE
     ->   /STATISTICS COEFF OUTS CI R ANOVA
     ->   /CRITERIA=PIN(.05) POUT(.IO)
     ->   /NOORIGIN
     ->   /DEPENDENT corun_id
     ->   /METHOD=ENTER ln_hccut ln_cocut


               ****  MULTIPLE   REGRESSION   ****

     Equation Number 1   Dependent Variable..   CORUN_ID   CORun ID


     Block Number  1.  Method:  Enter     LN HCCUT LN  COCUT
M6IM001.WPD DRAFT                     86                             Mar 24, 1999

-------
     Variable(s)  Entered on Step  Number
        1..     LN_COCUT
        2..     LN HCCUT
                                                                                    DRAFT
     Multiple R           .90658
     R Square             .82188
     Adjusted R Square    .81707
     Standard Error       .06736

     Analysis of Variance
     Regression
     Residual
             170.72789
       DF      Sum of Squares
        2            1.54920
       74              .33574

           Signif F =   .0000
         Mean Square
              .77460
              .00454
     Variable
                            Variables  in  the Equation 	

                           B        SE B     95% Confdnce Intrvl B
LN HCCUT
LN COCUT
(Constant)
- .107306
- .129819
1.188020
.011014
.008268
.027384
- .129253
- .146293
1.133456
- .085360
- .113344
1.242584
- .477976
- .770339

     Variable

     LN_HCCUT
     LN_COCUT
     (Constant)
      T  Sig  T

 -9.742  .0000
-15.702  .0000
 43.384  .0000
     -> * Curve Estimation.
     -> TSET NEWVAR=NONE .
     -> CURVEFIT /VARIABLES=noid  WITH nocut
     ->   /CONSTANT
     ->   /MODEL=CUBIC
     ->   /PRINT ANOVA
     ->   /PLOT FIT.

     Dependent variable..  NOID

     Listwise Deletion of  Missing  Data

     Multiple R           .99902
     R Square             .99805
     Adjusted R Square    .99658
     Standard Error       .01860
                 Analysis of  Variance:

                   DF   Sum of  Squares
     Regression
     Residuals
           .70707598
           .00138343
                         Method.. CUBIC
Mean Square

  .23569199
  .00034586
     F =     681.46957       Signif  F  =   .0000

     	  Variables  in the Equation 	

     Variable                  B        SE B       Beta         T  Sig T
     NOCUT
     NOCUT**2
     NOCUT**3
     (Constant)
       .756842      .102036   3.175112
      -.368671      .037175  -9.352562
       .040631      .004083   5.358327
       .545291      .082060
                7.417   .0018
               -9.917   .0006
                9.951   .0006
                6.645   .0027
M6IM001.WPD DRAFT
                                 87
                                         Mar 24,  1999

-------
                                                                                DRAFT

                                          APPENDIX F
             Statistical Diagnostics for Start Emissions IDR Determination
     -> REGRESSION
     ->   /MISSING LISTWISE
     ->   /STATISTICS COEFF OUTS CI  R ANOVA
     ->   /CRITERIA=PIN(.05) POUT(.IO)
     ->   /NOORIGIN
     ->   /DEPENDENT hc_strt_
     ->   /METHOD=ENTER  In hccut In  cocut
               ****   MULTIPLE   REGRESSION  ***


     Listwise Deletion of Missing Data

     Equation Number 1   Dependent  Variable..    HC_STRT_   HC Strt ID

     Block Number  1.  Method:  Enter      LN HCCUT LN COCUT
     Variable(s) Entered on Step Number
        1..    LN_COCUT
        2..    LN HCCUT
     Multiple R           .85506
     R Square             .73113
     Adjusted R Square     .70669
     Standard Error       .11633

     Analysis of Variance
                       DF      Sum  of Squares     Mean Square
     Regression           2              .80951          .40476
     Residual           22              .29769          .01353

     F =      29.91216       Signif F =   .0000
                        - Variables  in the Equation 	

                         B        SE B     95% Confdnce Intrvl B       Beta

                                                      -.099126   -.609838
                                                      -.089645   -.630732
                                                      1.155752
LN HCCUT
LN COCUT
(Constant )
- .158962
- .140941
.981406
.028853
.024734
.084067
- .218799
- .192237
.807061
     Variable          T  Sig T

     LN_HCCUT      -5.509  .0000
     LN_COCUT      -5.698  .0000
     (Constant)    11.674  .0000

     -> REGRESSION
     ->   /MISSING LISTWISE
     ->   /STATISTICS COEFF OUTS CI  R ANOVA
     ->   /CRITERIA=PIN(.05)  POUT(.IO)
     ->   /NOORIGIN
     ->   /DEPENDENT co_strt_
     ->   /METHOD=ENTER In hccut In  cocut
               ****   MULTIPLE   REGRESSION  ***


     Listwise Deletion of Missing Data

     Equation Number 1   Dependent  Variable..    CO_STRT_   CO Strt ID

     Block Number  1.  Method:  Enter      LN HCCUT LN COCUT
M6IM001.WPD DRAFT                     88                              Mar 24, 1999

-------
                                                                                  DRAFT
     Variable(s)  Entered on Step Number
        1..     LN_COCUT
        2..     LN HCCUT
     Multiple R           .84999
     R Square             .72249
     Adjusted R Square     .69726
     Standard Error        .13266

     Analysis of Variance
                        DF      Sum of  Squares      Mean Square
     Regression          2             1.00799           .50399
     Residual           22              .38718           .01760

     F =      28.63762       Signif F =   .0000


     	 Variables in the Equation 	

     Variable             B        SE B    95% Confdnce Intrvl B       Beta

     LN_HCCUT       -.159301     .032905    -.227541     -.091061    -.544428
     LN_COCUT       -.170728     .028208    -.229228     -.112229    -.680635
     (Constant)      1.145947     .095873      .947118     1.344777
     	  ln	

     Variable          T  Sig T

     LN_HCCUT     -4.841  .0001
     LN_COCUT     -6.053  .0000
     (Constant)    11.953  .0000
M6IM001.WPD DRAFT                      89                              Mar 24, 1999

-------
                                                                         DRAFT

                                       APPENDIX G
          Statistical Diagnostics for Running and Start High Emitter Levels
     -> USE ALL.
     -> COMPUTE filter_$=(vehicle =  1  & hc_2x = 2 &  (grp88 = 3 |  grp88  =  6  )).
     -> VARIABLE LABEL filter_$ 'vehicle  = 1 & hc_2x = 2 & (grp88 = 3  | grp88  = 6  )
     ->   '  (FILTER)'.
     -> VALUE LABELS filter_$  0  'Not  Selected' 1 'Selected'.
     -> FORMAT filter_$ (fl.O).
     -> FILTER BY filter_$.

     -> EXECUTE .

     -> EXAMINE
     ->   VARIABLES=hc_cs hc_la4ho BY  filter_$
     ->   /PLOT NONE
     ->   /STATISTICS DESCRIPTIVES
     ->   /CINTERVAL 95
     ->   /MISSING LISTWISE
     ->   /NOTOTAL.
         HC_CS
     By  FILTER_$  1

     Valid cases:
                   Selected

               118.0   Missing cases:
                                                          Percent missing:
     Mean        5.3127  Std Err      .9562  Min
     Median      3.8660  Variance   107.8864  Max
     5% Trim     4.3032  Std Dev    10.3868  Range
     95% CI for Mean (3.4190,  7.2064)        IQR
                                             -23.3000  Skewness    6.7474
                                             100.5300  S E Skew     .2227
                                             123.8300  Kurtosis   61.6924
                                               2.8755  S E Kurt     .4419
         HC_LA4HO  HC_LA4HOT
     By  FILTER_$  1         Selected
     Valid cases:
                        118.0   Missing cases:
                                                 Percent missing:
     Mean        2.3725  Std Err      .4448  Min
     Median      1.0085  Variance    23.3486  Max
     5% Trim     1.4788  Std Dev     4.8320  Range
     95% CI for Mean (1.4916,  3.2535)        IQR
                                                .2690  Skewness    5.1217
                                              34.8100  S E Skew     .2227
                                              34.5410  Kurtosis   28.7006
                                               1.3385  S E Kurt     .4419
     -> USE ALL.
     -> COMPUTE filter_$=(vehicle =  1  &  co_3x = 2 &  (grp88 = 3 |  grp88  =  6  )).
     -> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x = 2 & (grp88 = 3  | grp88  = 6  )
     ->   '  (FILTER)'.
     -> VALUE LABELS filter_$  0  'Not  Selected' 1 'Selected'.
     -> FORMAT filter_$ (fl.O).
     -> FILTER BY filter_$.

     -> EXAMINE
     ->   VARIABLES=co_cs co_la4ho BY  filter_$
     ->   /PLOT NONE
     ->   /STATISTICS DESCRIPTIVES
     ->   /CINTERVAL 95
     ->   /MISSING LISTWISE
     ->   /NOTOTAL.
     By
CO_CS
FILTER $
                             Selected
M6IM001.WPD DRAFT
                                   90
Mar 24, 1999

-------
                                                                           DRAFT
     Valid  cases:
                          97.0
                                Missing cases:
                          Percent missing:
     Mean        65.3116  Std Err     9.4172  Min
     Median      41.1230  Variance  8602.238  Max
     5% Trim     63.1510  Std Dev    92.7483  Range
     95% CI for  Mean  (46.6187, 84.0045)      IQR
                       -181.100  Skewness      .7955
                       441.8000  S E Skew      .2450
                       622.9000  Kurtosis    2.4172
                        95.2160  S E Kurt      .4853
          CO_LA4HO  CO_LA4HOT
     By   FILTER_$  1         Selected
     Valid  cases:
                          97.0
                                Missing cases:
                          Percent missing:
     Mean        37.9327  Std Err     5.2679  Min
     Median      14.1360  Variance  2691.801  Max
     5% Trim     30.7305  Std Dev    51.8826  Range
     95% CI for  Mean  (27.4761, 48.3894)      IQR
                          .2920  Skewness    2.4569
                      288.6300  S E Skew      .2450
                      288.3380  Kurtosis    6.7550
                        33.8540  S E Kurt      .4853
     -> USE ALL.
     -> COMPUTE  filter_$=(vehicle = 1 & no_2x = 2 &  (grp88 = 3  | grp88 = 6 )).
     -> VARIABLE LABEL  filter_$  'vehicle = 1 & no_2x = 2 & (grp88 = 3 |  grp88 = 6 )
     ->   '  (FILTER)'.
     -> VALUE LABELS  filter_$  0  'Not Selected' 1 'Selected'.
     -> FORMAT filter_$  (fl.O).
     -> FILTER BY  filter_$.

     -> EXECUTE  .

     -> EXAMINE
     ->   VARIABLES=no_la4ho BY  filter_$
     ->   /PLOT NONE
     ->   /STATISTICS DESCRIPTIVES
     ->   /CINTERVAL  95
     ->   /MISSING LISTWISE
     ->   /NOTOTAL.
         NO_LA4HO  NO_LA4HOT
     By  FILTER_$  1         Selected
     Valid  cases:
                         44.0   Missing cases:
                          Percent missing:
                                                                                 .0
     Mean        2.9513  Std Err       .1349  Min
     Median      2.5785  Variance      .8006  Max
     5% Trim     2.8761  Std Dev       .8948  Range
     95% CI for Mean  (2.6793, 3.2233)        IQR
                         1.9530  Skewness    1.2149
                         5.6660  S E Skew      .3575
                         3.7130  Kurtosis      .9399
                         1.2920  S E Kurt      .7017
       USE ALL.
       COMPUTE  filter_$=(vehicle = 1 & hc_2x = 2 &
          7 ) ) .
       VARIABLE LABEL  filter_$  'vehicle = 1 & hc_2x = 2
         ' grp88 =  7  )  (FILTER)'.
       VALUE  LABELS  filter_$   0
       FORMAT filter_$  (fl.O).
       FILTER BY filter_$.
                   (grp88  = 4  | grp88 = 5  | grp88 =

                          (grp88 = 4  | grp88 = 5  |
'Not  Selected'  1  'Selected'
     -> EXECUTE  .

     -> EXAMINE
     ->   VARIABLES=hc_cs hc_la4ho BY filter_$
     ->   /PLOT NONE
     ->   /STATISTICS DESCRIPTIVES
     ->   /CINTERVAL 95
     ->   /MISSING LISTWISE
M6IM001.WPD DRAFT
            91
Mar 24, 1999

-------
          /NOTOTAL.
                                                                          DRAFT
         HC_CS
     By  FILTER_$  1

     Valid cases:
     Selected

212.0   Missing cases:
                                                           Percent missing:
                                                                                 .0
     Mean        10.5195  Std Err     1.6407  Min
     Median      5.8390  Variance  570.6977  Max
     5% Trim     7.8954  Std Dev    23.8893  Range
     95% CI for  Mean  (7.2852, 13.7538)       IQR
                                -5.3850   Skewness    11.1465
                               326.0100   S  E  Skew      .1671
                               331.3950   Kurtosis   145.4885
                                 6.8020   S  E  Kurt      .3326
         HC_LA4HO  HC_LA4HOT
     By  FILTER_$  1         Selected
     Valid cases:
                        212.0   Missing cases:
                                   Percent  missing:
                                                                                 .0
     Mean        1.8447  Std Err       .3111  Min
     Median        .7975  Variance   20.5202  Max
     5% Trim     1.2606  Std Dev     4.5299  Range
     95% CI for Mean  (1.2314, 2.4580)        IQR
                                  .1390   Skewness    10.4292
                                59.8590   S  E  Skew      .1671
                                59.7200   Kurtosis   129.3009
                                 1.3962   S  E  Kurt      .3326
     -> USE ALL.
     -> COMPUTE  filter_$=(vehicle = 1 & co_3x = 2 &  (grp88 = 4 |  grp88 = 5 |  grp88 =
     ->    7 )) .
     -> VARIABLE LABEL  filter_$  'vehicle = 1 & co_3x = 2 & (grp88 = 4 |  grp88 = 5 |'+
     ->   ' grp88 = 7  )  (FILTER)'.
     -> VALUE  LABELS  filter_$   0  'Not Selected' 1 'Selected'.
     -> FORMAT filter_$  (fl.O).
     -> FILTER BY filter_$.

     -> EXECUTE  .

     -> EXAMINE
     ->    VARIABLES=co_cs co_la4ho BY filter_$
     ->    /PLOT NONE
     ->    /STATISTICS DESCRIPTIVES
     ->    /CINTERVAL  95
     ->    /MISSING LISTWISE
     ->    /NOTOTAL.
          CO_CS
     By   FILTER_$  1

     Valid cases:
     Selected

233.0   Missing cases:
                                                           Percent missing:
                                                                                 .0
     Mean        92.8206  Std Err     5.4515  Min
     Median      78.5740  Variance  6924.600  Max
     5% Trim     88.8831  Std Dev    83.2142  Range
     95% CI for  Mean  (82.0797, 103.5614)     IQR
                               -145.000   Skewness      .8815
                               401.0900   S  E  Skew      .1595
                               546.0900   Kurtosis     1.8693
                                88.6325   S  E  Kurt      .3176
          CO_LA4HO  CO_LA4HOT
     By   FILTER_$  1         Selected
     Valid cases:
                        233.0   Missing cases:
                                   Percent  missing:
                                                                                 .0
     Mean       27.6531  Std Err     2.7249  Min
     Median     11.4820  Variance  1729.998  Max
     5% Trim    21.1470  Std Dev    41.5932  Range
     95% CI for Mean  (22.2845, 33.0217)      IQR
                                  .1330   Skewness     3.2284
                               298.0400   S  E  Skew      .1595
                               297.9070   Kurtosis    12.9400
                                21.1570   S  E  Kurt      .3176
M6IM001.WPD DRAFT
                    92
Mar 24, 1999

-------
                                                                           DRAFT
     -> USE ALL.
     -> COMPUTE  filter_$=(vehicle = 1 & no_2x = 2 &  (grp88 = 4  | grp88 = 5 |  grp88 =
     ->    7 ) ) .
     -> VARIABLE LABEL  filter_$  'vehicle = 1 & no_2x = 2 & (grp88 = 4 |  grp88 = 5 |'+
     ->   ' grp88 =  7  )  (FILTER)'.
     -> VALUE  LABELS  filter_$   0  'Not Selected' 1 'Selected'.
     -> FORMAT filter_$  (fl.O).
     -> FILTER BY filter_$.

     -> EXECUTE  .

     -> EXAMINE
     ->    VARIABLES=no_la4ho BY  filter_$
     ->    /PLOT NONE
     ->    /STATISTICS DESCRIPTIVES
     ->    /CINTERVAL  95
     ->    /MISSING  LISTWISE
     ->    /NOTOTAL.
         NO_LA4HO  NO_LA4HOT
     By  FILTER_$  1         Selected
     Valid  cases:
                         60.0
                                Missing cases:
                                                  Percent  missing:
     Mean        2.8719  Std Err       .0991  Min
     Median      2.6320  Variance      .5898  Max
     5% Trim     2.8139  Std Dev       .7680  Range
     95% CI for Mean  (2.6735, 3.0703)        IQR
                                                1.8730   Skewness     1.2166
                                                5.8210   S  E  Skew      .3087
                                                3.9480   Kurtosis     2.2562
                                                1.1900   S  E  Kurt      .6085
     -> USE ALL.
     -> COMPUTE  filter_$=(vehicle = 1 & hc_2x = 2 &  (n_group = 1 )).
     -> VARIABLE LABEL  filter_$  'vehicle = 1 & hc_2x = 2 &  (n_group = 1 )   (FILTER)
     -> VALUE LABELS  filter_$  0  'Not Selected' 1 'Selected'.
     -> FORMAT filter_$  (fl.O).
     -> FILTER BY  filter_$.

     -> EXECUTE  .

     -> EXAMINE
     ->   VARIABLES=hc_cs hc_la4ho BY filter_$
     ->   /PLOT NONE
     ->   /STATISTICS DESCRIPTIVES
     ->   /CINTERVAL  95
     ->   /MISSING LISTWISE
     ->   /NOTOTAL.
         HC_CS
     By  FILTER_$   1

     Valid  cases:
                    Selected

                58.0   Missing cases:
                                                           Percent missing:
     Mean        4.8290  Std Err       .7673  Min
     Median      3.9220  Variance   34.1484  Max
     5% Trim     4.6639  Std Dev     5.8437  Range
     95% CI for Mean  (3.2925, 6.3655)        IQR
                                              -23.3000   Skewness     -.7800
                                               24.2470   S  E  Skew      .3137
                                               47.5470   Kurtosis    11.2352
                                                3.3150   S  E  Kurt      .6181
     By
HC_LA4HO
FILTER_$
                   HC_LA4HOT
                             Selected
M6IM001.WPD DRAFT
                                   93
Mar 24, 1999

-------
                                                                           DRAFT
     Valid  cases:
                         58.0
                                Missing cases:
                                  Percent  missing:
     Mean         1.7400  Std Err       .5316  Min
     Median        .8790  Variance   16.3917  Max
     5% Trim      1.1879  Std Dev     4.0487  Range
     95% CI for Mean  (.6754, 2.8045)         IQR
                                 .1450   Skewness     6.9800
                               31.1790   S  E  Skew      .3137
                               31.0340   Kurtosis    51.3330
                                1.0667   S  E  Kurt      .6181
     -> USE ALL.
     -> COMPUTE  filter_$=(vehicle = 1 & hc_2x = 2 &  (n_group = 2 )).
     -> VARIABLE LABEL  filter_$  'vehicle = 1 & hc_2x = 2 &  (n_group = 2 )   (FILTER)
     -> VALUE LABELS  filter_$  0  'Not Selected' 1 'Selected'.
     -> FORMAT filter_$  (fl.O).
     -> FILTER BY  filter_$.

     -> EXECUTE  .

     -> EXAMINE
     ->   VARIABLES=hc_cs hc_la4ho BY filter_$
     ->   /PLOT NONE
     ->   /STATISTICS DESCRIPTIVES
     ->   /CINTERVAL  95
     ->   /MISSING LISTWISE
     ->   /NOTOTAL.
         HC_CS
     By  FILTER_$   1

     Valid  cases:
    Selected

38.0   Missing cases:
                                                           Percent missing:
                                                                                 .0
     Mean         3.2927  Std Err       .4646  Min
     Median       3.1575  Variance    8.2028  Max
     5% Trim      3.2990  Std Dev     2.8640  Range
     95% CI for Mean  (2.3513, 4.2341)        IQR
                               -3.2350   Skewness     -.0333
                               10.4150   S  E  Skew      .3828
                               13.6500   Kurtosis      .7728
                                3.0717   S  E  Kurt      .7497
         HC_LA4HO  HC_LA4HOT
     By  FILTER_$  1         Selected
     Valid  cases:
                         38.0
                                Missing cases:
                                  Percent  missing:
                                                                                 .0
     Mean         3.3937  Std Err     1.0523  Min
     Median       1.5370  Variance   42.0754  Max
     5% Trim      2.2123  Std Dev     6.4866  Range
     95% CI for Mean  (1.2616, 5.5257)        IQR
                                 .5030   Skewness     3.8432
                               34.8100   S  E  Skew      .3828
                               34.3070   Kurtosis    15.8588
                                1.5170   S  E  Kurt      .7497
     -> USE ALL.
     -> COMPUTE  filter_$=(vehicle = 1 & co_3x = 2 &  (n_group = 1 )).
     -> VARIABLE LABEL  filter_$  'vehicle = 1 & co_3x = 2 &  (n_group = 1 )   (FILTER)
     -> VALUE LABELS  filter_$  0  'Not Selected' 1 'Selected'.
     -> FORMAT filter_$  (fl.O).
     -> FILTER BY  filter_$.

     -> EXECUTE  .

     -> EXAMINE
     ->   VARIABLES=co_cs co_la4ho BY filter_$
     ->   /PLOT NONE
     ->   /STATISTICS DESCRIPTIVES
     ->   /CINTERVAL  95
     ->   /MISSING LISTWISE
     ->   /NOTOTAL.
M6IM001.WPD DRAFT
                   94
Mar 24, 1999

-------
                                                                           DRAFT
          co_cs
     By   FILTER_$  1

     Valid  cases:
    Selected

44.0   Missing cases:
                                                           Percent missing:
                                                                                 .0
     Mean        38.0579  Std Err    11.1153  Min
     Median      33.6220  Variance  5436.158  Max
     5% Trim     36.5122  Std Dev    73.7303  Range
     95% CI for  Mean  (15.6419, 60.4740)      IQR
                              -134.000   Skewness      .5754
                              286.3600   S E  Skew      .3575
                              420.3600   Kurtosis     2.4815
                               73.3610   S E  Kurt      .7017
          CO_LA4HO  CO_LA4HOT
     By   FILTER_$  1         Selected
     Valid  cases:
                         44.0   Missing cases:
                                  Percent missing:
                                                                                 .0
     Mean        36.1057  Std Err     7.1383  Min
     Median      19.5880  Variance  2242.051  Max
     5% Trim     29.1347  Std Dev    47.3503  Range
     95% CI for  Mean  (21.7099, 50.5015)      IQR
                                5.0870   Skewness     3.8473
                              288.6300   S E  Skew      .3575
                              283.5430   Kurtosis    18.8008
                               30.8875   S E  Kurt      .7017
     -> USE ALL.
     -> COMPUTE  filter_$=(vehicle = 1 & co_3x = 2 &  (n_group = 2 )).
     -> VARIABLE LABEL  filter_$  'vehicle = 1 & co_3x = 2 &  (n_group = 2 )  (FILTER)
     -> VALUE LABELS  filter_$  0  'Not Selected' 1 'Selected'.
     -> FORMAT filter_$  (fl.O).
     -> FILTER BY  filter_$.

     -> EXECUTE  .

     -> EXAMINE
     ->   VARIABLES=co_cs co_la4ho BY filter_$
     ->   /PLOT NONE
     ->   /STATISTICS DESCRIPTIVES
     ->   /CINTERVAL  95
     ->   /MISSING LISTWISE
     ->   /NOTOTAL.
          CO_CS
     By   FILTER_$  1

     Valid  cases:
    Selected

43.0   Missing cases:
                                                           Percent missing:
     Mean       27.1649  Std Err    14.4273  Min
     Median     35.0280  Variance  8950.311  Max
     5% Trim    33.9697  Std Dev    94.6061  Range
     95% CI for Mean  (-1.9505, 56.2804)      IQR
                              -280.000   Skewness    -1.3909
                              218.1000   S E  Skew      .3614
                              498.1000   Kurtosis     4.0345
                               74.8400   S E  Kurt      .7090
          CO_LA4HO  CO_LA4HOT
     By   FILTER_$  1         Selected
     Valid  cases:
                         43.0   Missing cases:
                                  Percent missing:
     Mean       46.5270  Std Err     8.1257  Min
     Median     21.1950  Variance  2839.142  Max
     5% Trim    40.2982  Std Dev    53.2836  Range
     95% CI for Mean  (30.1287, 62.9252)      IQR
                                3.9840   Skewness     1.7022
                              216.8700   S E  Skew      .3614
                              212.8860   Kurtosis     2.4073
                               55.1990   S E  Kurt      .7090
M6IM001.WPD DRAFT
                   95
Mar 24, 1999

-------
                                                                          DRAFT
     -> USE ALL.
     -> COMPUTE  filter_$=(vehicle = 1 & no_2x = 2 &  (n_group = 1 )).
     -> VARIABLE LABEL filter_$  'vehicle = 1 & no_2x = 2 &  (n_group = 1 )  (FILTER)
     -> VALUE LABELS  filter_$  0  'Not Selected' 1 'Selected'.
     -> FORMAT filter_$  (fl.O).
     -> FILTER BY  filter_$.

     -> EXECUTE  .

     -> EXAMINE
     ->   VARIABLES=no_la4ho BY  filter_$
     ->   /PLOT NONE
     ->   /STATISTICS DESCRIPTIVES
     ->   /CINTERVAL  95
     ->   /MISSING LISTWISE
     ->   /NOTOTAL.
         NO_LA4HO  NO_LA4HOT
     By  FILTER_$  1         Selected
     Valid cases:
                         11.0   Missing cases:
              Percent missing:
                                                                                 .0
     Mean        2.8455  Std Err       .3223  Min
     Median      2.3870  Variance    1.1423  Max
     5% Trim     2.7867  Std Dev     1.0688  Range
     95% CI for Mean  (2.1274, 3.5635)        IQR
            1.7130  Skewness      .9851
            5.0350  S E Skew      .6607
            3.3220  Kurtosis      .0612
            1.6230  S E Kurt    1.2794
       USE ALL.
       COMPUTE  filter_$=(vehicle = 1 & no_2x = 2 &  (n_group = 2 )).
       VARIABLE LABEL filter_$  'vehicle = 1 & no_2x = 2 &  (n_group = 2 )  (FILTER)
       VALUE LABELS  filter_$  0  'Not Selected' 1 'Selected'.
       FORMAT filter_$  (fl.O).
       FILTER BY  filter_$.

       EXECUTE  .
       EXAMINE
         VARIABLES=no_la4ho BY filter_$
         /PLOT NONE
         /STATISTICS DESCRIPTIVES
         /CINTERVAL 95
         /MISSING LISTWISE
         /NOTOTAL.
         NO_LA4HO  NO_LA4HOT
     By  FILTER_$  1         Selected
     Valid cases:
                         15.0   Missing cases:
              Percent missing:
                                                                                 .0
     Mean        2.8723  Std Err       .2612  Min
     Median      2.4130  Variance    1.0235  Max
     5% Trim     2.7682  Std Dev     1.0117  Range
     95% CI for Mean  (2.3121, 3.4326)        IQR
            1.9530  Skewness    1.9401
            5.6660  S E Skew      .5801
            3.7130  Kurtosis    3.5993
              .9110  S E Kurt    1.1209
M6IM001.WPD DRAFT
96
Mar 24, 1999

-------
                                                                                DRAFT



                                          APPENDIX H

         Statistical Diagnostics for Running and Start Normal Emitter Levels


     -> GET
     -> FILE='D:\MOBILE6\IM\IM_CRED\NEW_CRED\EF5_DAT.SAV .

     -> COMPUTE filter_$=(vehicle = 1  & hc_2x = 1 & grp88 = 1).

     -> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x = 1 & grp88  =  1  (FILTER)'.

     -> REGRESSION
     ->   /MISSING LISTWISE
     ->   /STATISTICS COEFF OUTS CI R ANOVA
     ->   /CRITERIA=PIN(.05) POUT(.IO)
     ->   /NOORIGIN
     ->   /DEPENDENT hc_la4ho
     ->   /METHOD=ENTER mileage


               ****   MULTIPLE   REGRESSION  ****


     Listwise Deletion of Missing Data

     Equation Number 1   Dependent Variable..    HC_LA4HO   HC_LA4HOT

     Block Number  1. Method:  Enter      MILEAGE


     Variable(s) Entered on Step Number
        1..    MILEAGE


     Multiple R           .44218
     R  Square             .19552
     Adjusted R Square    .19501
     Standard Error       .07163

     Analysis of Variance
                       DF      Sum of Squares      Mean Square
     Regression          1            1.97161         1.97161
     Residual         1581            8.11227          .00513

     F  =    384.24703      Signif F  =  .0000
     	 Variables  in the Equation  	

     Variable             B        SE B     95% Confdnce Intrvl B       Beta

     MILEAGE         .001385  7.0661E-05      .001247      .001524     .442178
     (Constant)      .021397     .003347      .014831      .027963


     	 in	

     Variable           T  Sig T

     MILEAGE       19.602  .0000
     (Constant)     6.392  .0000


     End Block Number   1   All requested variables  entered.


     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS  COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05) POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT hc_cs
     ->    /METHOD=ENTER mileage


               ****  MULTIPLE   REGRESSION  ****



M6IM001.WPD DRAFT                     97                             Mar 24, 1999

-------
     Listwise Deletion of Missing Data

     Equation Number 1    Dependent Variable..   HC_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
                                                                                    DRAFT
     Multiple R           .18470
     R Square             .03411
     Adjusted R Square    .03350
     Standard Error       .92659
     Analysis of Variance
                         DF
     Regression           1
     Residual          1581
     F =
              55.83968
      Sum of Squares
           47.94271
         1357.41150

  Signif F =   .0000
                                 Mean Square
                                    47.94271
                                      .85858

Variable
MILEAGE
(Constant )

B SE B
.006830 9.1404E-04
1.998720 .043300
< n
H
95% Confdnce
.005037
1.913788

Intrvl B
.008623
2 .083652

Beta
.184701

     Variable
     MILEAGE
     (Constant)
                        T  Sig  T
 7.473
46.159
.0000
.0000
     End Block Number
                            All  requested variables entered.
     -> COMPUTE filter_$=( vehicle  =  1 & hc_2x = 1 & grp88 =2).

     -> VARIABLE LABEL filter_$  'vehicle = 1 & hc_2x = 1 & grp88 = 1 (FILTER)

     -> VALUE LABELS filter_$  0  'Not Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS  CI R ANOVA
     ->    /CRITERIA=PIN( .05)  POUT(.IO)
     - >    /NOORIGIN
     ->    /DEPENDENT hc_la4ho
     ->    /METHOD=ENTER mileage


                ****   MULTIPLE   REGRESSION   ****


     Listwise Deletion of  Missing  Data

     Equation Number 1    Dependent  Variable. .   HC_LA4HO   HC_LA4HOT

     Block Number  1.  Method:  Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
M6IM001.WPD DRAFT
                       98
                                                                 Mar 24, 1999

-------
     Multiple R           .54551
     R Square             .29758
     Adjusted R Square    .29596
     Standard Error       .07463
                                                                                    DRAFT
     Analysis of Variance
                         DF
     Regression           1
     Residual           434
     F =
             183.86531
             Sum of Squares
                   1.02400
                   2.41708

         Signif F  =   .0000
                   Mean Square
                       1.02400
                        .00557
H
Variable
MILEAGE
(Constant)
H i
B
.001701
.004198
SE B
1.2544E-04
.007088
95% Confdnce Intrvl B
.001454 .001947
-.009733 .018128
Beta
.545510

     Variable
                        T  Sig  T
     MILEAGE       13.560  .0000
     (Constant)       .592  .5540
     End Block Number   1   All  requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF  OUTS  CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT hc_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent Variable..   HC_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
     Multiple R           .10853
     R Square             .01178
     Adjusted R Square    .00950
     Standard Error       .70086
     Analysis of Variance
     Regression
     Residual
     DF      Sum of Squares
      1            2.54088
    434           213.18120

         Signif F  =   .0234
                   Mean Square
                       2.54088
                        .49120
     Variable
     MILEAGE
     (Constant)
                            Variables in the Equation 	

                           B       SE B     95% Confdnce Intrvl B
 .002679
1.901893
.001178  3.63926E-04
.066564      1.771064
 .004995
2 .032721
                                                                        Beta

                                                                     .108529
                 in
M6IM001.WPD DRAFT
                               99
                                                   Mar 24, 1999

-------
                                                                               DRAFT
Variable
     MILEAGE
     (Constant)
                   T  Sig T
 2.274
28.572
                      .0234
                      .0000
End Block Number
                       All  requested variables entered.
-> COMPUTE filter_$=( vehicle  =  1  & hc_2x = 1 & grp88 =3).

-> VARIABLE LABEL filter_$  'vehicle  =  1 & hc_2x = 1 & grp88 = 1 (FILTER) '

-> VALUE LABELS filter_$   0  'Not  Selected' 1 'Selected'.

-> FORMAT filter_$ (fl.O).

-> FILTER BY filter_$.

-> EXECUTE .

-> REGRESSION
->    /MISSING LISTWISE
->    /STATISTICS COEFF OUTS  CI R ANOVA
->    /CRITERIA=PIN( .05)  POUT(.IO)
- >    /NOORIGIN
->    /DEPENDENT hc_la4ho
->    /METHOD=ENTER mileage


           ****   MULTIPLE  REGRESSION   ****


Listwise Deletion of  Missing  Data

Equation Number 1    Dependent  Variable. .   HC_LA4HO   HC_LA4HOT

Block Number  1.  Method:  Enter       MILEAGE
Variable (s)  Entered on Step  Number
   1 . .     MILEAGE
Multiple R           .34643
R Square             .12001
Adjusted R Square    .11859
Standard Error       .12378
     Analysis of Variance
                         DF
     Regression           1
     Residual           621
F =
         84.69051
              Sum of  Squares
                     1.29756
                     9.51443

          Signif  F =   .0000
                                               Mean Square
                                                   1.29756
                                                    .01532
Variable
     MILEAGE
     (Constant)
                       Variables  in  the Equation

                      B        SE B     95% Confdnce Intrvl B
  .001439  1.5635E-04
  .094216     .009189
                                         .001132
                                         .076171
                                                                   Beta

                                                                .346426
            in
Variable
                   T  Sig T
MILEAGE        9.203   .0000
(Constant)     10.254   .0000
End Block Number   1   All  requested variables entered.
-> REGRESSION
M6IMOO 1 . WPD DRAFT
                                1 00
                                                                               Mar 24, 1 999

-------
                                                                                     DRAFT
           /MISSING LISTWISE
           /STATISTICS COEFF OUTS CI  R ANOVA
           /CRITERIA=PIN(.05)  POUT(.IO)
           /NOORIGIN
           /DEPENDENT hc_cs
           /METHOD=ENTER mileage
                ****   MULTIPLE    REGRESSION


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent  Variable..   HC_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .03793
     R Square             .00144
     Adjusted R Square   -.00017
     Standard Error      1.16208

     Analysis of Variance
                         DF      Sum of  Squares      Mean Square
     Regression           1             1.20815          1.20815
     Residual           621           838.61274          1.35042

     F =        .89465       Signif  F =   .3446
     	 Variables  in the  Equation  	

     Variable              B        SE B     95%  Confdnce Intrvl B       Beta

     MILEAGE         .001388     .001468     -.001494       .004271     .037929
     (Constant)     2.358932     .086266     2.189523     2.528341
     Variable           T  Sig T

     MILEAGE         .946  .3446
     (Constant)    27.345  .0000


     End Block Number   1   All requested  variables entered.


     - >
     -> COMPUTE filter_$=(vehicle =  1  &  hc_2x  =  1  & grp88 = 4).

     -> VARIABLE LABEL filter_$ 'vehicle = 1 & hc_2x =  1 & grp88 = 1  (FILTER)'.

     -> VALUE LABELS filter_$  0 'Not  Selected'  1  'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05) POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT hc_la4ho
     ->    /METHOD=ENTER mileage


                ****   MULTIPLE   REGRESSION   ****




M6IMOO1.WPD DRAFT                      101                              Mar 24, 1999

-------
                                                                                    DRAFT
     Listwise Deletion of Missing Data

     Equation Number 1    Dependent Variable..   HC_LA4HO   HC_LA4HOT

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
     Multiple R           .25073
     R Square             .06286
     Adjusted R Square    .05245
     Standard Error       .09393
     Analysis of Variance
     Regression
     Residual
         DF      Sum of Squares
          1               .05327
         90               .79414

             Signif F =   .0159
Mean Square
     .05327
     .00882
     MILEAGE
     (Constant)
	  Variables in the Equation 	

          B       SE B     95% Confdnce Intrvl B

 8.12421E-04 3.3064E-04  1.55544E-04      .001469
     .077383     .020471      .036713      .118053
                                                                        Beta

                                                                     .250729
     Variable           T  Sig T

     MILEAGE        2.457  .0159
     (Constant)      3.780  .0003
     End Block Number   1   All  requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF  OUTS  CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT hc_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent Variable..   HC_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
     Multiple R           .48064
     R Square             .23102
     Adjusted R Square    .22247
     Standard Error       .99649
     Analysis of Variance
     Regression
     Residual
         DF      Sum of Squares
          1            26.84774
         90            89.36862
Mean Square
   26.84774
     .99298
     F =      27.03742        Signif F =   .0000
M6IM001.WPD DRAFT
                                   102
                                Mar 24, 1999

-------
                                                                                    DRAFT

Variable
MILEAGE
(Constant )
B
.018238
1.493421
SE B
.003508
.217166
95% Confdnce
.011270
1.061982
Intrvl B
.025207
1.924860
Beta
.480640

     MILEAGE
     (Constant)
    T  Sig T

5.200  .0000
6.877  .0000
     End Block Number
                           All requested variables entered.
     -> COMPUTE filter_$=(vehicle = 1 & hc_2x = 1 & grp88 = 5).

     -> VARIABLE LABEL filter_$  'vehicle = 1 & hc_2x = 1 & grp88  =  1  (FILTER)

     -> VALUE LABELS filter_$  0  'Not Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT hc_la4ho
     ->    /METHOD=ENTER mileage
                          MULTIPLE
                                           REGRESSION
     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..    HC_LA4HO   HC_LA4HOT

     Block Number  1.   Method:  Enter      MILEAGE
     Variable(s)  Entered on  Step Number
        1..     MILEAGE
     Multiple R           .20394
     R Square             .04159
     Adjusted R Square    .03746
     Standard Error       .12495

     Analysis of Variance
     Regression
     Residual
     DF      Sum of Squares
      1              .15719
    232            3.62209

         Signif F =   .0017
Mean Square
     .15719
     .01561
     MILEAGE
     (Constant)
	  Variables  in the Equation 	

       B        SE B     95% Confdnce Intrvl B

 .001214  3.8251E-04  4.60068E-04      .001967
 .126577     .014947       .097128      .156026
                                                                       Beta

                                                                      203940
M6IM001.WPD DRAFT
                               103
                                Mar 24,  1999

-------
                                                                                    DRAFT
     Variable           T  Sig T

     MILEAGE        3.173  .0017
     (Constant)      8.469  .0000
     End Block Number   1   All  requested  variables entered.


     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS  CI  R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT hc_cs
     ->    /METHOD=ENTER mileage


                ****   MULTIPLE   REGRESSION


     Listwise Deletion of Missing  Data

     Equation Number 1    Dependent Variable..   HC_CS

     Block Number  1.  Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .16213
     R Square             .02628
     Adjusted R Square    .02209
     Standard Error      1.22801

     Analysis of Variance
                         DF      Sum of  Squares      Mean Square
     Regression           1             9.44422          9.44422
     Residual           232           349.86033          1.50802

     F =       6.26267       Signif  F =   .0130
     	  Variables  in  the  Equation 	

     Variable              B        SE B      95% Confdnce Intrvl B       Beta

     MILEAGE         .009408     .003759       .002001       .016815    .162126
     (Constant)      1.589214     .146898      1.299790     1.878638


     	 in	

     Variable           T  Sig T

     MILEAGE        2.503  .0130
     (Constant)     10.818  .0000


     End Block Number   1   All requested variables entered.


     - >
     -> COMPUTE filter_$=(vehicle  =  1  & hc_2x = 1 & grp88 = 6).

     -> VARIABLE LABEL filter_$ 'vehicle  = 1  & hc_2x = 1 & grp88 = 1  (FILTER)'.

     -> VALUE LABELS filter_$  0  'Not  Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION



M6IMOO1.WPD DRAFT                      104                              Mar 24, 1999

-------
                                                                                    DRAFT
           /MISSING LISTWISE
           /STATISTICS COEFF OUTS  CI  R  ANOVA
           /CRITERIA=PIN(.05)  POUT(.IO)
           /NOORIGIN
           /DEPENDENT hc_la4ho
           /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION   **


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent  Variable..   HC_LA4HO   HC_LA4HOT

     Block Number  1.   Method:   Enter       MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .50527
     R Square             .25529
     Adjusted R Square    .24806
     Standard Error       .11052

     Analysis of Variance
                         DF      Sum of  Squares      Mean Square
     Regression           1              .43130           .43130
     Residual           103             1.25812           .01221

     F =      35.30979       Signif  F =   .0000
     ----------------------  Variables  in  the  Equation -----------------------

     Variable              B        SE B      95% Confdnce Intrvl B       Beta

     MILEAGE         .002250  3.7865E-04       .001499       .003001    .505267
     (Constant)       .097024     .018871       .059598       .134450
     Variable           T  Sig T

     MILEAGE        5.942  .0000
     (Constant)      5.141  .0000
     End Block Number   1   All  requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS  CI  R  ANOVA
     ->    /CRITERIA=PIN( .05)  POUT(.IO)
     - >    /NOORIGIN
     ->    /DEPENDENT hc_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent  Variable. .   HC_CS

     Block Number  1.   Method:   Enter       MILEAGE
     Variable (s)  Entered on Step Number
        1 . .     MILEAGE
     Multiple R           .21032


M6IMOO 1 . WPD DRAFT                      1 05                             Mar 24, 1 999

-------
                                                                                    DRAFT
     R Square             .04423
     Adjusted R Square    .03495
     Standard Error      1.14075
     Analysis of Variance
                         DF
     Regression           1
     Residual           103
     F =
               4.76698
              Sum  of  Squares
                     6.20337
                   134.03607

          Signif F =   .0313
                   Mean Square
                       6.20337
                       1.30132
     MILEAGE
     (Constant)
                          -  Variables  in the Equation 	

                          B        SE B     95% Confdnce Intrvl B
  .008533
 2.354343
.003908   7.81969E-04
.194779      1.968044
 .016284
2.740641
                                                                        Beta

                                                                      210319
     MILEAGE
     (Constant)
     T  Sig T

 2.183  .0313
12.087  .0000
     End Block Number
                            All  requested variables entered.
     -> COMPUTE filter_$=(vehicle  =  1  & hc_2x = 1 & grp88 = 7).

     -> VARIABLE LABEL filter_$  'vehicle =  1 & hc_2x = 1 & grp88 = 1 (FILTER)

     -> VALUE LABELS filter_$   0  'Not  Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS  CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT hc_la4ho
     ->    /METHOD=ENTER mileage
                          MULTIPLE
                                           REGRESSION
     Listwise Deletion of Missing  Data

     Equation Number 1    Dependent Variable..   HC_LA4HO   HC_LA4HOT

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
     Multiple R           .28245
     R Square             .07978
     Adjusted R Square    .07868
     Standard Error       .11606
     Analysis of Variance
                         DF
     Regression           1
     Residual           837
     F =
              72 .56441
              Sum  of  Squares
                      .97738
                   11.27372

          Signif F =   .0000
                   Mean Square
                        .97738
                        .01347
M6IM001.WPD DRAFT
                                106
                                                   Mar 24, 1999

-------
                                                                                    DRAFT
     Variable
     MILEAGE
     (Constant)
 	  Variables  in  the Equation 	

        B        SE B     95% Confdnce Intrvl B

  .001271  1.4921E-04   9.78196E-04      .001564
  .153943     .006278       .141621      .166266
                                                                        Beta

                                                                     .282452
                 in
     Variable
     MILEAGE
     (Constant)
     T  Sig T

 8.518  .0000
24.522  .0000
     End Block Number   1    All  requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF  OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT hc_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION


     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..   HC_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .24760
     R Square             .06131
     Adjusted R Square    .06019
     Standard Error      1.43178
     Analysis of Variance
                         DF
     Regression           1
     Residual           837
     F =
              54.66519
              Sum  of Squares
                   112.06309
                  1715.84167

          Signif F =   .0000
                   Mean Square
                     112.06309
                       2.04999
                          -  Variables in the Equation 	

                          B       SE B     95% Confdnce Intrvl B
     MILEAGE
     (Constant)
  .013610
 2.121260
.001841
.077449
 .009997
1.969242
 .017224
2.273278
                                                                        Beta

                                                                      247602
     Variable           T  Sig  T

     MILEAGE        7.394  .0000
     (Constant)     27.389  .0000
     End Block Number   1    All  requested variables entered.
M6IM001.WPD DRAFT
                                107
                                                   Mar 24, 1999

-------
                                                                                    DRAFT
     -> COMPUTE filter_$=(vehicle  =  1  &  co_3x = 1 & grp88 = 1).

     -> VARIABLE LABEL filter_$  'vehicle =  1 & co_3x = 1 & grp88 = 1 (FILTER)

     -> VALUE LABELS filter_$  0  'Not  Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS  CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT co_la4ho
     ->    /METHOD=ENTER mileage


                ****   MULTIPLE  REGRESSION   ****


     Listwise Deletion of Missing  Data

     Equation Number 1    Dependent  Variable..   CO_LA4HO   CO_LA4HOT

     Block Number  1.  Method:  Enter       MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
     Multiple R           .43497
     R Square             .18920
     Adjusted R Square    .18869
     Standard Error      1.21705

     Analysis of Variance
                         DF      Sum  of  Squares      Mean Square
     Regression           1           549.21332        549.21332
     Residual          1589          2353.66001          1.48122

     F =     370.78421       Signif F =   .0000
H
Variable
MILEAGE
(Constant )
H i
B
.022927
.458769
SE B
.001191
.056622
95% Confdnce Intrvl B
.020592 .025262
.347707 .569832
Beta
.434967

     MILEAGE       19.256  .0000
     (Constant)      8.102  .0000
     End Block Number   1   All  requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS  CI  R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT co_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION   ****
M6IMOO1.WPD DRAFT                      108                             Mar 24, 1999

-------
     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..    CO_CS

     Block Number  1.   Method:  Enter      MILEAGE
     Variable(s)  Entered on  Step Number
        1..     MILEAGE
                                                                                    DRAFT
Multiple R           .01494
R Square             .00022
Adjusted R Square   -.00041
Standard Error     12.05858

Analysis of Variance
                    DF      Sum of Squares
Regression           1            51.59410
Residual          1589        231055.54059

F =        .35482       Signif F =   .5515
                                                    Mean Square
                                                       51.59410
                                                      145.40940
H
Variable B
MILEAGE .007027
(Constant) 18.972536
-i n
SE B
.011797
.561015
95% Confdnce Intrvl B
-.016112 .030166
17.872129 20.072942
Beta
.014941

     Variable
     MILEAGE
     (Constant)
                        T  Sig T
                .596
              33.818
.5515
.0000
     End Block Number
                            All requested variables entered.
     -> COMPUTE filter_$=( vehicle = 1 & co_3x = 1 & grp88 =2).

     -> VARIABLE LABEL filter_$  'vehicle = 1 & co_3x = 1 & grp88 =  1  (FILTER)

     -> VALUE LABELS filter_$  0  'Not Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN( .05)  POUT(.IO)
     - >    /NOORIGIN
     ->    /DEPENDENT co_la4ho
     ->    /METHOD=ENTER mileage


                ****  MULTIPLE   REGRESSION    ****


     Listwise Deletion of Missing Data

     Equation Number 1   Dependent Variable. .    CO_LA4HO   CO_LA4HOT

     Block Number  1.   Method:  Enter      MILEAGE
     Variable(s)  Entered on  Step Number
        1..     MILEAGE
M6IM001.WPD DRAFT
                                              109
                                                         Mar 24,  1999

-------
                                                                                   DRAFT
     Multiple R           .57491
     R Square             .33052
     Adjusted R Square     .32897
     Standard Error      1.39129
     Analysis of  Variance
                        DF
     Regression          1
     Residual          431
     F =
             212.78325
    Sum of  Squares
         411.88438
         834.28637

Signif  F =   .0000
Mean Square
  411.88438
    1.93570
H
Variable
MILEAGE
(Constant)
B
.033909
- .028277
i n
SE B
.002325
.131686
95% Confdnce Intrvl B
.029340 .038478
-.287103 .230549
Beta
.574909

     Variable
                       T  Sig T
     MILEAGE       14.587   .0000
     (Constant)      -.215   .8301
     End Block Number   1   All requested variables entered.
     -> REGRESSION
     ->    /MISSING  LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05) POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT co_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION


     Listwise Deletion of Missing Data

     Equation Number  1    Dependent Variable..    CO_CS

     Block Number  1.  Method:  Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .02245
     R Square             .00050
     Adjusted R Square   -.00182
     Standard Error      8.95890

     Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 17.43774 17.43774
Residual 431 34592.88334 80.26191
F = .21726 Signif F = .6414
Variable B SE B 95% Confdnce Intrvl B
MILEAGE -.006977 .014969 -.036398 .022444
(Constant) 19.232859 .847958 17.566211 20.899506
-i n


Beta
- .022446

M6IM001.WPD DRAFT
                      110
                               Mar 24, 1999

-------
                                                                                     DRAFT
     Variable           T  Sig T

     MILEAGE        -.466  .6414
     (Constant)    22.681  .0000


     End Block Number   1   All requested variables  entered.


     - >
     -> COMPUTE filter_$=(vehicle =  1  &  co_3x  =  1  &  grp88 =3).

     -> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x =  1 & grp88 = 1  (FILTER)'.

     -> VALUE LABELS filter_$  0 'Not  Selected'  1  'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05) POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT co_la4ho
     ->    /METHOD=ENTER mileage


                ****   MULTIPLE  REGRESSION   ****


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent  Variable..   CO_LA4HO   CO_LA4HOT

     Block Number  1.  Method:  Enter       MILEAGE


     Variable(s) Entered on Step Number
        1..     MILEAGE


     Multiple R           .34381
     R Square             .11821
     Adjusted R Square    .11683
     Standard Error      1.80541

     Analysis of Variance
                         DF      Sum of  Squares       Mean Square
     Regression           1           279.20868         279.20868
     Residual           639         2082.82364          3.25950

     F =      85.65984       Signif  F  =   .0000


     	  Variables  in the Equation 	

     Variable              B        SE B    95% Confdnce Intrvl B       Beta

     MILEAGE         .019588     .002116      .015432      .023744     .343812
     (Constant)     1.444769     .130120    1.189254    1.700284


     	 in	

     Variable           T  Sig T

     MILEAGE        9.255  .0000
     (Constant)    11.103  .0000


     End Block Number   1   All requested variables  entered.


     -> REGRESSION



M6IMOO1.WPD DRAFT                      111                              Mar 24,  1999

-------
                                                                                     DRAFT
           /MISSING LISTWISE
           /STATISTICS COEFF OUTS CI R ANOVA
           /CRITERIA=PIN(.05)  POUT(.IO)
           /NOORIGIN
           /DEPENDENT co_cs
           /METHOD=ENTER mileage
                ****   MULTIPLE    REGRESSION


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent Variable..    CO_CS

     Block Number  1.  Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .01070
     R Square             .00011
     Adjusted R Square   -.00145
     Standard Error     13.54016

     Analysis of Variance
                         DF      Sum of  Squares       Mean Square
     Regression           1            13.41337          13.41337
     Residual           639        117151.73251         183.33604

     F =        .07316       Signif F =   .7869
     	 Variables  in the  Equation  	

     Variable              B        SE B     95%  Confdnce  Intrvl B       Beta

     MILEAGE        -.004293     .015873     -.035462       .026875   -.010700
     (Constant)    19.949338     .975872   18.033034    21.865642
     Variable           T  Sig T

     MILEAGE        -.270  .7869
     (Constant)    20.443  .0000


     End Block Number   1   All requested variables  entered.


     - >
     -> COMPUTE filter_$=(vehicle = 1  & co_3x  =  1  &  grp88 = 4).

     -> VARIABLE LABEL filter_$ 'vehicle = 1 & co_3x =  1 & grp88 = 1  (FILTER)'.

     -> VALUE LABELS filter_$  0 'Not  Selected'  1  'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05) POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT co_la4ho
     ->    /METHOD=ENTER mileage


                ****   MULTIPLE  REGRESSION   ****




M6IMOO1.WPD DRAFT                      112                              Mar 24,  1999

-------
                                                                                    DRAFT
     Listwise Deletion of Missing Data

     Equation Number 1    Dependent Variable..   CO_LA4HO   CO_LA4HOT

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
     Multiple R           .38730
     R Square             .15000
     Adjusted R Square    .14076
     Standard Error      1.02382
     Analysis of Variance
     Regression
     Residual
     DF      Sum of Squares
      1            17.01790
     92            96.43454

        Signif F =   .0001
                    Mean Square
                       17.01790
                       1.04820
     MILEAGE
     (Constant)
	 Variables in the Equation 	

      B        SE B     95% Confdnce Intrvl B

 .013709      .003402      .006952      .020467
 .566553      .216869      .135832      .997274
                                                                        Beta

                                                                     .387299
     Variable           T  Sig T

     MILEAGE        4.029  .0001
     (Constant)      2.612  .0105
     End Block Number   1   All  requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF  OUTS  CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT co_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent Variable..   CO_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
     Multiple R           .16121
     R Square             .02599
     Adjusted R Square    .01540
     Standard Error     21.02393
     Analysis of Variance
     Regression
     Residual
     DF
      1
     92
Sum of Squares
    1085.07366
   40664.53507
Mean Square
 1085.07366
  442.00582
                             Signif  F =   .1206
M6IM001.WPD DRAFT
                               113
                                                   Mar 24, 1999

-------
                                                                                    DRAFT

Variable
MILEAGE
(Constant )
B
.109470
24.697606
SE B
.069868
4.453376
95% Confdnce
- .029294
15.852816
Intrvl B
.248234
33 .542395
Beta
.161214

     MILEAGE
     (Constant)
    T  Sig T

1.567  .1206
5.546  .0000
     End Block Number
                            All requested variables entered.
     -> COMPUTE filter_$=(vehicle = 1 & co_3x = 1 & grp88 = 5).

     -> VARIABLE LABEL filter_$  'vehicle = 1 & co_3x = 1 & grp88 =  1  (FILTER)

     -> VALUE LABELS filter_$  0  'Not Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT co_la4ho
     ->    /METHOD=ENTER mileage
                          MULTIPLE
                                           REGRESSION
     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..    CO_LA4HO   CO_LA4HOT

     Block Number  1.   Method:  Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .24144
     R Square             .05829
     Adjusted R Square    .05423
     Standard Error      1.46214
     Analysis of Variance
     Regression
     Residual
     DF      Sum  of Squares
      1           30.70190
    232           495.98217

         Signif F =   .0002
Mean Square
   30.70190
    2.13785
     MILEAGE
     (Constant)
	  Variables  in the Equation 	

       B        SE B     95% Confdnce Intrvl B

 .016908     .004462       .008118      .025699
 .727606     .175115       .382587     1.072626
                                                                        Beta

                                                                      241439
M6IM001.WPD DRAFT
                               114
                                Mar 24,  1999

-------
                                                                                     DRAFT
     Variable           T  Sig T

     MILEAGE        3.790  .0002
     (Constant)     4.155  .0000
     End Block Number   1   All requested  variables entered.


     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI  R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT co_cs
     ->    /METHOD=ENTER mileage


                ****   MULTIPLE    REGRESSION


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent Variable..    CO_CS

     Block Number  1.  Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .10909
     R Square             .01190
     Adjusted R Square    .00764
     Standard Error     20.73715

     Analysis of Variance
                         DF      Sum of  Squares      Mean Square
     Regression           1          1201.49589        1201.49589
     Residual           232         99766.81401        430.02937

     F =       2.79399       Signif  F =   .0960
     	 Variables  in the  Equation  	

     Variable              B        SE B     95%  Confdnce Intrvl B       Beta

     MILEAGE         .105775     .063281     -.018903       .230453     .109086
     (Constant)    24.442451    2.483616   19.549126    29.335775


     	 in	

     Variable           T  Sig T

     MILEAGE        1.672  .0960
     (Constant)     9.841  .0000


     End Block Number   1   All requested variables entered.


     - >
     -> COMPUTE filter_$=(vehicle =  1  & co_3x = 1  & grp88 = 6).

     -> VARIABLE LABEL filter_$ 'vehicle = 1  & co_3x =  1 & grp88 = 1  (FILTER)'.

     -> VALUE LABELS filter_$  0 'Not  Selected' 1  'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION



M6IMOO1.WPD DRAFT                      115                             Mar 24, 1999

-------
                                                                                     DRAFT
           /MISSING LISTWISE
           /STATISTICS COEFF OUTS CI  R ANOVA
           /CRITERIA=PIN(.05)  POUT(.IO)
           /NOORIGIN
           /DEPENDENT co_la4ho
           /METHOD=ENTER mileage
                ****   MULTIPLE    REGRESSION   **


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent  Variable..    CO_LA4HO   CO_LA4HOT

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .32292
     R Square             .10428
     Adjusted R Square    .09583
     Standard Error      1.75396

     Analysis of Variance
                         DF      Sum of  Squares      Mean Square
     Regression           1           37.96399         37.96399
     Residual           106           326.09694          3.07639

     F =      12.34045       Signif  F =   .0007
     ---------------------- Variables  in the  Equation  -----------------------

     Variable              B        SE B     95%  Confdnce Intrvl B       Beta

     MILEAGE         .021502     .006121      .009367       .033637    .322923
     (Constant)     1.576249     .300873      .979739     2.172759
     Variable           T  Sig T

     MILEAGE        3.513  .0007
     (Constant)     5.239  .0000
     End Block Number   1   All requested  variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI  R ANOVA
     ->    /CRITERIA=PIN( .05)  POUT(.IO)
     - >    /NOORIGIN
     ->    /DEPENDENT co_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE    REGRESSION


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent  Variable. .    CO_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable (s)  Entered on Step Number
        1 . .     MILEAGE
     Multiple R           .27648


M6IMOO 1 . WPD DRAFT                      1 1 6                             Mar 24, 1 999

-------
                                                                                    DRAFT
     R Square             .07644
     Adjusted R Square    .06773
     Standard Error     25.80196
     Analysis of Variance
                         DF
     Regression           1
     Residual           106
     F =
               S.77340
             Sum  of Squares
                 5840.81033
               70568.54286

         Signif F =   .0038
Mean Square
 5840.81033
  665.74097

Variable
MILEAGE
(Constant )
B
.266706
20.038190
SE B
.090043
4.426039
95% Confdnce
.088187
11.263137
Intrvl B
.445224
28.813243
Beta
.276480

     MILEAGE
     (Constant)
    T  Sig T

2.962  .0038
4.527  .0000
     End Block Number   1    All  requested variables entered.


     - >
     -> COMPUTE filter_$=(vehicle = 1 & co_3x = 1 & grp88 = 7).

     -> VARIABLE LABEL filter_$  'vehicle = 1 & co_3x = 1 & grp88 = 1 (FILTER)

     -> VALUE LABELS filter_$  0 'Not Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT co_la4ho
     ->    /METHOD=ENTER mileage
                          MULTIPLE
                                           REGRESSION
     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..   CO_LA4HO   CO_LA4HOT

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .20587
     R Square             .04238
     Adjusted R Square    .04121
     Standard Error      1.75821
     Analysis of Variance
                         DF
     Regression           1
     Residual           814
     F =
              36.02576
             Sum  of Squares
                  111.36702
                 2516.33162

         Signif F =   .0000
Mean Square
  111.36702
    3.09132
M6IM001.WPD DRAFT
                               117
                                Mar 24,  1999

-------
                                                                                    DRAFT
     Variable
     MILEAGE
     (Constant)
                            Variables in the Equation 	

                           B       SE B     95% Confdnce Intrvl B
  .013887
 1.393200
 .002314
 .096173
  .009346
 1.204424
  .018429
 1.581977
                                                                        Beta

                                                                     .205869
                 in
     Variable
     MILEAGE
     (Constant)
     T  Sig T

 6.002  .0000
14.486  .0000
     End Block Number   1    All  requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF  OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT co_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION


     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..   CO_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .23702
     R Square             .05618
     Adjusted R Square    .05502
     Standard Error     24.75160
     Analysis of Variance
                         DF
     Regression           1
     Residual           814
     F =
              48.45262
              Sum  of  Squares
                29684.08673
                498690.22165

          Signif F =   .0000
                    Mean Square
                    29684.08673
                       612.64155
                          -  Variables in the Equation 	

                          B       SE B     95% Confdnce Intrvl B
     MILEAGE
     (Constant)
  .226728
28.636627
 .032572
1.353894
  .162792
25.979092
  .290663
31.294161
                                                                        Beta

                                                                     .237023
     Variable           T  Sig  T

     MILEAGE        6.961  .0000
     (Constant)     21.151  .0000
     End Block Number   1    All  requested variables entered.
M6IM001.WPD DRAFT
                                118
                                                    Mar 24, 1999

-------
                                                                                     DRAFT
     -> COMPUTE filter_$=(vehicle =  1  &  no_2x  =  1  & grp88 = 1).

     -> VARIABLE LABEL filter_$ 'vehicle =  1 & no_2x  =  1 & grp88 = 1  (FILTER)'.

     -> VALUE LABELS filter_$  0 'Not  Selected'  1  'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05) POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT no_la4ho
     ->    /METHOD=ENTER mileage


                ****   MULTIPLE  REGRESSION   ****


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent  Variable..   NO_LA4HO   NO_LA4HOT

     Block Number  1.  Method:  Enter       MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .39011
     R Square             .15219
     Adjusted R Square    .15166
     Standard Error       .22872

     Analysis of Variance
                         DF      Sum of  Squares      Mean Square
     Regression           1            15.10851          15.10851
     Residual          1609            84.16805           .05231

     F =     288.82214       Signif  F =   .0000
     ---------------------- Variables  in the  Equation  -----------------------

     Variable              B        SE B     95%  Confdnce Intrvl B       Beta

     MILEAGE         .003756  2.2100E-04      .003322       .004189     .390110
     (Constant)      .200589     .010576      .179844       .221334
     Variable           T  Sig T

     MILEAGE       16.995  .0000
     (Constant)    18.966  .0000
     End Block Number   1   All requested variables  entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI  R ANOVA
     ->    /CRITERIA=PIN( .05)  POUT(.IO)
     - >    /NOORIGIN
     ->    /DEPENDENT no_cs
     ->    /METHOD=ENTER mileage
                          MULTIPLE    REGRESSION
M6IMOO1.WPD DRAFT                      119                              Mar 24,  1999

-------
     Listwise Deletion of Missing Data

     Equation Number 1    Dependent Variable..   NO_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
                                                                                    DRAFT
     Multiple R           .05934
     R Square             .00352
     Adjusted R Square    .00290
     Standard Error       .95537
     Analysis of Variance
                         DF
     Regression           1
     Residual          1609
     F =
               5.68559
              Sum of  Squares
                     5.18937
                  1468.57163

          Signif  F =   .0172
Mean Square
    5.18937
     .91272
     MILEAGE
     (Constant)
	  Variables  in  the Equation 	

        B        SE B     95% Confdnce Intrvl B

  .002201  9.2312E-04   3.90488E-04      .004012
 1.443620     .044179     1.356966     1.530274
                                                                        Beta

                                                                     .059340
     MILEAGE
     (Constant)
     T  Sig T

 2.384  .0172
32.677  .0000
     End Block Number   1   All  requested variables entered.


     - >
     -> COMPUTE filter_$=( vehicle  =  1  & no_2x = 1 & grp88 =2).

     -> VARIABLE LABEL filter_$  'vehicle = 1 & no_2x = 1 & grp88 = 1 (FILTER)

     -> VALUE LABELS filter_$   0 'Not  Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS  CI R ANOVA
     ->    /CRITERIA=PIN( .05)  POUT(.IO)
     - >    /NOORIGIN
     ->    /DEPENDENT no_la4ho
     ->    /METHOD=ENTER mileage
                          MULTIPLE
                                           REGRESSION
     Listwise Deletion of Missing Data

     Equation Number 1    Dependent Variable. .   NO_LA4HO   NO_LA4HOT

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
M6IM001.WPD DRAFT
                                120
                                Mar 24, 1999

-------
     Multiple R           .44967
     R Square             . 20220
     Adjusted R Square    .20038
     Standard Error       .21488
                                                                                    DRAFT
     Analysis of Variance
     Regression
     Residual
             111.26452
     DF
      1
    439
    Sum of  Squares
           5.13724
          20.26926

Signif F =   .0000
        Mean Square
            5.13724
             .04617
H
Variable
MILEAGE
(Constant )
H i
B
.003806
.225302
SE B
3 .6084E-04
.020306
95% Confdnce Intrvl B
.003097 .004515
.185393 .265212
Beta
.449669

     Variable
                        T  Sig T
     MILEAGE       10.548   .0000
     (Constant)     11.095   .0000
     End Block Number   1   All requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT no_cs
     ->    /METHOD=ENTER mileage
                ****    MULTIPLE   REGRESSION


     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..    NO_CS

     Block Number  1.  Method:  Enter      MILEAGE
     Variable(s)  Entered on  Step Number
        1..     MILEAGE
     Multiple R           .10744
     R Square             .01154
     Adjusted R Square    .00929
     Standard Error      1.17515
     Analysis of Variance
                        DF
     Regression           1
     Residual           439
     F =
               5.12705
             Sum of Squares
                   7.08035
                  606.25000

         Signif F  =   .0240
                       Mean Square
                           7.08035
                           1.38098
     MILEAGE
     (Constant)
                          -  Variables in the Equation 	

                          B       SE B     95% Confdnce Intrvl  B
-.004468
2.300454
    .001973
    .111055
-.008347 -5.89886E-04
2.082188     2.518720
                                                                       Beta

                                                                    - .107444
M6IM001.WPD DRAFT
                               121
                                                       Mar 24, 1999

-------
                                                                               DRAFT
Variable
     MILEAGE
     (Constant)
                   T  Sig  T
-2.264
20.715
                      .0240
                      .0000
End Block Number
                       All  requested variables entered.
-> COMPUTE filter_$=( vehicle = 1 & no_2x = 1 & grp88 =3).

-> VARIABLE LABEL filter_$  'vehicle = 1 & no_2x = 1 & grp88 = 1 (FILTER)

-> VALUE LABELS filter_$  0  'Not Selected' 1 'Selected'.

-> FORMAT filter_$ (fl.O).

-> FILTER BY filter_$.

-> EXECUTE .

-> REGRESSION
->    /MISSING LISTWISE
->    /STATISTICS COEFF OUTS CI R ANOVA
->    /CRITERIA=PIN( .05)  POUT(.IO)
- >    /NOORIGIN
->    /DEPENDENT no_la4ho
->    /METHOD=ENTER mileage


           ****  MULTIPLE   REGRESSION   ****


Listwise Deletion of Missing Data

Equation Number 1   Dependent Variable. .    NO_LA4HO   NO_LA4HOT

Block Number  1.   Method:  Enter      MILEAGE
Variable(s)  Entered on Step Number
   1..     MILEAGE
Multiple R           .18619
R Square             .03467
Adjusted R Square    .03327
Standard Error       .33374
     Analysis of Variance
                         DF
     Regression           1
     Residual           692
F =
         24.85020
              Sum  of  Squares
                     2.76791
                   77.07747

          Signif F =   .0000
                                               Mean Square
                                                   2.76791
                                                    .11138
H
Variable
MILEAGE
(Constant)
H i
B
.001883
.479830
SE B
3 .7774E-04
.023534
95% Confdnce Intrvl B
.001141 .002625
.433623 .526037
Beta
.186188

Variable
                   T  Sig  T
MILEAGE        4.985   .0000
(Constant)     20.389   .0000
End Block Number   1    All  requested variables entered.
-> REGRESSION
M6IMOO 1 . WPD DRAFT
                                1 22
                                                                               Mar 24, 1 999

-------
                                                                                     DRAFT
           /MISSING LISTWISE
           /STATISTICS COEFF OUTS CI  R ANOVA
           /CRITERIA=PIN(.05)  POUT(.IO)
           /NOORIGIN
           /DEPENDENT no_cs
           /METHOD=ENTER mileage
                ****   MULTIPLE    REGRESSION


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent  Variable..   NO_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .04406
     R Square             .00194
     Adjusted R Square    .00050
     Standard Error      1.07313

     Analysis of Variance
                         DF      Sum of  Squares      Mean Square
     Regression           1             1.54972          1.54972
     Residual           692           796.91339          1.15161

     F =       1.34570       Signif  F =   .2464
     	 Variables  in the  Equation  	

     Variable              B        SE B     95%  Confdnce Intrvl B       Beta

     MILEAGE         .001409     .001215 -9.75750E-04       .003794    .044055
     (Constant)     1.406422     .075673     1.257846     1.554999
     Variable           T  Sig T

     MILEAGE        1.160  .2464
     (Constant)    18.586  .0000


     End Block Number   1   All requested  variables entered.


     - >
     -> COMPUTE filter_$=(vehicle =  1  &  no_2x  =  1  & grp88 = 4).

     -> VARIABLE LABEL filter_$ 'vehicle = 1 & no_2x =  1 & grp88 = 1  (FILTER)'.

     -> VALUE LABELS filter_$  0 'Not  Selected'  1  'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05) POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT no_la4ho
     ->    /METHOD=ENTER mileage


                ****   MULTIPLE   REGRESSION   ****




M6IMOO1.WPD DRAFT                      123                             Mar 24, 1999

-------
                                                                                    DRAFT
     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..   NO_LA4HO   NO_LA4HOT

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .16411
     R Square             .02693
     Adjusted R Square    .01647
     Standard Error       .32441
     Analysis of Variance
     Regression
     Residual
     DF      Sum of Squares
      1               .27088
     93             9.78756

        Signif F =   .1120
                           Mean Square
                                .27088
                                .10524
     MILEAGE
     (Constant)
	 Variables in the Equation 	

      B        SE B     95% Confdnce Intrvl B

 .001702      .001061 -4.04579E-04      .003808
 .495967      .067974      .360984      .630950
                                                                        Beta

                                                                     .164107
     Variable           T  Sig  T

     MILEAGE        1.604  .1120
     (Constant)      7.296  .0000
     End Block Number   1    All  requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF  OUTS  CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT no_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION


     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..   NO_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .06506
     R Square             .00423
     Adjusted R Square   -.00647
     Standard Error      1.05703
     Analysis of Variance
     Regression
     Residual
DF
 1
93
            Sum of Squares
                     .44169
                 103.91026
Mean Square
     .44169
    1.11731
                             Signif F =   .5311
M6IM001.WPD DRAFT
                               124
                                                           Mar 24, 1999

-------
                                                                                    DRAFT

Variable
MILEAGE
(Constant )
B
- .002173
1.404902
SE B
.003456
.221481
95% Confdnce
- .009035
.965084
Intrvl B
.004690
1.844719
Beta
- .065059

     MILEAGE
     (Constant)
    T  Sig T

-.629  .5311
6.343  .0000
     End Block Number
                           All requested variables entered.
     -> COMPUTE filter_$=(vehicle = 1 & no_2x = 1 & grp88 = 5).

     -> VARIABLE LABEL filter_$  'vehicle = 1 & no_2x = 1 & grp88  =  1  (FILTER)

     -> VALUE LABELS filter_$  0  'Not Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT no_la4ho
     ->    /METHOD=ENTER mileage
                          MULTIPLE
                                           REGRESSION
     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..    NO_LA4HO   NO_LA4HOT

     Block Number  1.   Method:  Enter      MILEAGE
     Variable(s)  Entered on  Step Number
        1..     MILEAGE
     Multiple R           .15950
     R Square             .02544
     Adjusted R Square    .02148
     Standard Error       .38404
     Analysis of Variance
     Regression
     Residual
     DF      Sum of Squares
      1              .94709
    246           36.28266

         Signif F =   .0119
Mean Square
     .94709
     .14749
     MILEAGE
     (Constant)
	  Variables in the Equation 	

      B       SE B     95% Confdnce Intrvl B

 .002725     .001075  6.06968E-04      .004843
 .555546     .044174       .468539      .642552
                                                                       Beta

                                                                     .159497
M6IM001.WPD DRAFT
                               125
                                Mar 24,  1999

-------
                                                                                    DRAFT
     Variable           T  Sig T

     MILEAGE        2.534  .0119
     (Constant)     12.576  .0000
     End Block Number   1   All  requested variables entered.


     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS  CI  R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT no_cs
     ->    /METHOD=ENTER mileage


                ****   MULTIPLE   REGRESSION


     Listwise Deletion of Missing  Data

     Equation Number 1    Dependent Variable..   NO_CS

     Block Number  1.  Method:   Enter     MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .10136
     R Square             .01027
     Adjusted R Square    .00625
     Standard Error      1.17091

     Analysis of Variance
                         DF      Sum of  Squares      Mean Square
     Regression           1             3.50127          3.50127
     Residual           246           337.27551          1.37104

     F =       2.55374       Signif  F =   .1113
     	  Variables  in  the  Equation 	

     Variable              B        SE B      95% Confdnce Intrvl B       Beta

     MILEAGE         .005240     .003279      -.001218       .011698    .101363
     (Constant)       .747776     .134681       .482502     1.013050


     	 in	

     Variable           T  Sig T

     MILEAGE        1.598  .1113
     (Constant)      5.552  .0000


     End Block Number   1   All requested variables entered.


     - >
     -> COMPUTE filter_$=(vehicle  =  1  & no_2x = 1 & grp88 = 6).

     -> VARIABLE LABEL filter_$ 'vehicle  =  1  & no_2x = 1 & grp88 = 1 (FILTER)'.

     -> VALUE LABELS filter_$  0  'Not  Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION



M6IMOO1.WPD DRAFT                      126                              Mar 24, 1999

-------
                                                                                    DRAFT
           /MISSING LISTWISE
           /STATISTICS COEFF OUTS  CI  R ANOVA
           /CRITERIA=PIN(.05)  POUT(.IO)
           /NOORIGIN
           /DEPENDENT no_la4ho
           /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION   **


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent  Variable..   NO_LA4HO   NO_LA4HOT

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step Number
        1..     MILEAGE
     Multiple R           .41146
     R Square             .16930
     Adjusted R Square    .16146
     Standard Error       .38335

     Analysis of Variance
                         DF      Sum  of  Squares      Mean Square
     Regression           1             3.17478          3.17478
     Residual           106           15.57778           .14696

     F =      21.60302       Signif F =   .0000
     ----------------------  Variables  in  the Equation -----------------------

     Variable              B        SE B     95% Confdnce Intrvl B       Beta

     MILEAGE         .006326     .001361       .003628       .009024    .411459
     (Constant)       .459727     .066906       .327079       .592376
     Variable           T  Sig T

     MILEAGE        4.648  .0000
     (Constant)      6.871  .0000
     End Block Number   1   All  requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS  CI  R ANOVA
     ->    /CRITERIA=PIN( .05)  POUT(.IO)
     - >    /NOORIGIN
     ->    /DEPENDENT no_cs
     ->    /METHOD=ENTER mileage
                ****   MULTIPLE   REGRESSION


     Listwise Deletion of Missing Data

     Equation Number 1    Dependent  Variable. .   NO_CS

     Block Number  1.   Method:   Enter      MILEAGE
     Variable (s)  Entered on Step Number
        1 . .     MILEAGE
     Multiple R           .01562


M6IMOO 1 . WPD DRAFT                      1 27                             Mar 24, 1 999

-------
                                                                                    DRAFT
     R Square             .00024
     Adjusted R Square   -.00919
     Standard Error      1.03573
     Analysis of Variance
                         DF
     Regression           1
     Residual           106
     F =
                .02587
             Sum  of  Squares
                     .02775
                  113.71076

         Signif F =   .8725
Mean Square
     .02775
    1.07274
     	  Variables  in the Equation 	

     Variable             B        SE B     95% Confdnce Intrvl B

     MILEAGE     5.91396E-04     .003677     -.006699      .007882
     (Constant)      1.530162     .180765     1.171777     1.888547
                                                    Beta

                                                 .015619
     MILEAGE
     (Constant)
    T  Sig T

 .161  .8725
8.465  .0000
     End Block Number
                            All  requested variables entered.
     -> COMPUTE filter_$=(vehicle  =  1  & no_2x = 1 & grp88 = 7).

     -> VARIABLE LABEL filter_$  'vehicle =  1 & no_2x = 1 & grp88 = 1 (FILTER)

     -> VALUE LABELS filter_$   0  'Not  Selected' 1 'Selected'.

     -> FORMAT filter_$ (fl.O).

     -> FILTER BY filter_$.

     -> EXECUTE .

     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF OUTS  CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT no_la4ho
     ->    /METHOD=ENTER mileage
                          MULTIPLE
                                           REGRESSION
     Listwise Deletion of Missing  Data

     Equation Number 1    Dependent Variable..   NO_LA4HO   NO_LA4HOT

     Block Number  1.   Method:   Enter      MILEAGE
     Variable(s)  Entered on Step  Number
        1..     MILEAGE
     Multiple R           .17186
     R Square             .02954
     Adjusted R Square    .02854
     Standard Error       .37012
     Analysis of Variance
                         DF
     Regression           1
     Residual           972
     F =
              29.58196
             Sum  of  Squares
                    4.05244
                  133.15445

         Signif F =   .0000
Mean Square
    4.05244
     .13699
M6IM001.WPD DRAFT
                               128
                                Mar 24, 1999

-------
                                                                                    DRAFT
     Variable
     MILEAGE
     (Constant)
                            Variables in the Equation
                                            95% Confdnce Intrvl  B
  .002328  4.2806E-04
  .583430     .019264
             .001488
             .545626
              .003168
              .621234
                                                                       Beta

                                                                     .171858
                 in
     Variable
     MILEAGE
     (Constant)
     T  Sig T

 5.439  .0000
30.286  .0000
     End Block Number   1   All requested variables entered.
     -> REGRESSION
     ->    /MISSING LISTWISE
     ->    /STATISTICS COEFF  OUTS CI R ANOVA
     ->    /CRITERIA=PIN(.05)  POUT(.IO)
     ->    /NOORIGIN
     ->    /DEPENDENT no_cs
     ->    /METHOD=ENTER mileage
                ****    MULTIPLE   REGRESSION


     Listwise Deletion of  Missing Data

     Equation Number 1    Dependent Variable..    NO_CS

     Block Number  1.  Method:  Enter      MILEAGE
     Variable(s)  Entered on  Step Number
        1..     MILEAGE
     Multiple R           .17750
     R Square             .03151
     Adjusted R Square    .03051
     Standard Error      1.36097
     Analysis of Variance
                        DF
     Regression           1
     Residual           972
     F =
              31.61932
              Sum  of Squares
                   58.56677
                  1800.38342

          Signif F =   .0000
                   Mean Square
                      58.56677
                       1.85225
     MILEAGE
     (Constant)
                          -  Variables in the Equation 	

                          B       SE B     95% Confdnce Intrvl  B
 -.008851
 1.601358
.001574
.070836
-.011940
1.462349
-.005762
1.740366
                                                                       Beta

                                                                    - .177497
     Variable           T  Sig T

     MILEAGE       -5.623  .0000
     (Constant)     22.607  .0000
M6IM001.WPD DRAFT
                                129
                                                   Mar 24, 1999

-------
                                                         DRAFT
                         Figure2-Biennial II with BSD
                     2  i  4
  P  9  10  ii  12. i'i  14-
M6IM001.WPD DRAFT
130
Mar 24, 1999

-------