EPA-650/4-75-016
February 1975
Environmental Monitoring  Series


                                                  /.•.v.v.v.v.%v.v.v.v.v.v.v.
             I"
             53
             V
                                                               LU
                                                               O
                                                 :::;:::

-------
                                       EPA-650/4-75-016
            COLLABORATIVE  STUDY
OF REFERENCE METHOD FOR MEASUREMENT
         OF PHOTOCHEMICAL OXIDANTS
   IN THE ATMOSPHERE (OZONE-ETHYLENE
         CHEMILUMINESCENT METHOD)
                         by

               Herbert C. McKee, Ralph E. Childers,
                     and Van B . Parr

                  Southwest Research Institute   U.S. EPA-NEIC LIBRARY
                    8500 Culebra Road      nl«',/«^ c~*~-~i n *
                    P.O. Drawer 28510     °6"Ver Federal <&**
                   San Antonio, Texas 78284    Building 25, Ent. E-3
                                     P.O. Box 25227
                                     Denver, CO 80225-0227

                   Contract No. CPA 70-40
                     ROAP No. 26AAF
                  Program Element No. 1HA327
              EPA Project Officer: John H. Margeson

        Quality Assurance and Environmental Monitoring Laboratory
              National Environmental Research Center
            Research Triangle Park, North Carolina 27711
                      Prepared for

            U.S. ENVIRONMENTAL PROTECTION AGENCY
             OFFICE OF RESEARCH AND DEVELOPMENT
                  WASHINGTON, D.C. 20460

                      February 1975

-------
                        EPA REVIEW NOTICE

This report has been reviewed by the National Environmental Research
Center - Research Triangle Park, Office of Research and Development,
EPA, and approved for publication.  Approval does not signify that the
contents necessarily reflect the views and policies of the Environmental
Protection Agency, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use.
                   RESEARCH REPORTING SERIES

Research reports of the Office of Research and Development, U[.S. Environ-
mental Protection Agency, have been grouped into series.  These broad
categories were established to facilitate further development arid applica-
tion of environmental technology-  Elimination of traditional grouping was
consciously planned to foster technology transfer and maximum interface
in related fields.  These series are:

          1.  ENVIRONMENTAL HEALTH EFFECTS RESEARCH

          2.  ENVIRONMENTAL PROTECTION TECHNOLOGY

          3.  ECOLOGICAL RESEARCH

          4.  ENVIRONMENTAL MONITORING

          5.  SOCIOECONOMIC ENVIRONMENTAL STUDIES

          6.  SCIENTIFIC AND TECHNICAL  ASSESSMENT REPORTS
          9.  MISCELLANEOUS

This report has been assigned to the ENVIRONMENTAL MONITORING
series.  This series describes research conducted to develop new or
improved methods and instrumentation for the identification and quanti-
fication of environmental pollutants at the lowest conceivably significant
concentrations.  It also includes studies to determine the ambient concen-
trations of pollutants in the environment and/or the variance of pollutants
as a function  of time or meteorological factors.
This document is available to the public for sale through the National
Technical Information Service, Springfield, Virginia 22161.

                Publication No. EPA-650/4-75-016
                                 11

-------
                                            ABSTRACT

     This report contains information on collaborative tests to determine the precision and bias of the reference
method for measurement of photochemical oxidants as published by the Environmental Protection Agency in the
Federal Register, April 30, 1971.

     In the first phase test, ten collaborators were assembled at a common site to measure (from a common source)
ambient and ozone supplemented concentrations over a range from 0 to .510 ppm (0-1000Mg/m3).  The data were
analyzed to derive an estimate for the precision of the method. Estimates of the standard deviations of the process
for total (Sj), between-laboratory (5^) and within-laboratory (.%/,) were derived, as well as for the lower detectable
limit (LDL).  The results were:

                             ST   =  0.0033 + 0.0412 x         0 < x < 0.153
                             ST   =  -0.0017 + 0.068 x     0.153 < x < 0.510
                             SL   =  0.0008 + 0.0355 x         Q 
-------
                                        ACKNOWLEDGMENTS
      The authors wish to express appreciation to the Project Officer, Mr. John H. Margeson, and staff member
Mr. Michael E. Beard, of the Methods Standardization & Performance Evaluation Branch, EPA, for assistance in
the planning and the execution of both phases of the collaborative study. Also acknowledged is the assistance of
Mr. Clarence A. Boldt, Jr., of Southwest Research Institute, who not only served as a collaborator but was also
instrumental in the installation of the sampling system.

      Southwest Research Institute and the Environmental Protection Agency are indebted to Dr. A.J. Haagen-Smit,
Dr. Fred Shair, and to the California Institute of Technology for providing laboratory space and assistance for the
conduction of the first phase.  This valuable contribution is sincerely acknowledged.

      The assistance and advice of Messrs. Ernest E. Hughes and John K. Taylor, of the National Bureau of Stan-
dards, for the construction and calibration of the ozone generators is also acknowledged.

      The assistance and cooperation of the participating laboratories is acknowledged with sincere appreciation
for the voluntary efforts of the staff members who represented each organization. The names and affiliations of
the representatives completing the test for the first phase were as follows:
                       Name
                    Organization
Clarence A. Boldt, Jr.

William D. Estes
R. J. Martin, III
Dr. James J. Franz

Kenneth T. Irwin
Gerald M. Schlatter
Carl E. Johnson

J. H. Mclver

Robert K. Stevens
Kenneth Rehme
Dr. Kenneth Tenny

Karl J. Zobel
     The collaborators for the second phase were:

                       Name

Gerald M. Schlatter

William D. Estes
Rafael Ballagas
Dewayne Ehman

Kenneth Rehme
Southwest Research Institute
Houston, Texas
Georgia Department of Natural Resources,
Atlanta, Georgia
E. I. du Pont de Nemours & Co., Inc.
Wilmington, Delaware
Louisville and Jefferson County Air
Pollution Control, Louisville, Kentucky
Texas State Department of Health
Austin, Texas
Dow Chemical Company
Freeport, Texas
Environmental Protection Agency
Research Triangle Park, N. C.
Chicago Department of Environmental Control
Chicago, Illinois
Environmental Protection Agency
Research Triangle Park, N. C.
                    Organization

Louisville and Jefferson County Air
Pollution Control, Louisville, Kentucky
Georgia Dept. of Natural Resources
Atlanta, Georgia
Texas State Dept. of Health
Austin, Texas
Environmental Protection Agency
Research Triangle Park, N. C.
                                                   IV

-------
                       Name
                    Organization
Thomas Kovacik
Richard Usciloswski
John Phillips
Ben Ivey
James W. Register, Jr.

James J. Franz

George Matula

Karl J. Zobel

Richard Dell
Larry Cornet
Pollution Control Agency
Toledo, Ohio
Memphis & Shelby County Health Dept.
Memphis, Tennessee
Southwest Research Institute
San Antonio, Texas
E. I. du Pont de Nemours & Co.
Wilmington, Delaware
Dow Chemical Company
Freeport, Texas
Environmental Protection Agency
Research Triangle Park, N. C.
Air Pollution Control Department
of Public Health
Cleveland, Ohio

-------
                                  TABLE OF CONTENTS


                                                                                          Page

ABSTRACT	                 ......          -                 '"'

ACKNOWLEDGMENTS              .     .  .       	       •     •  •   •  iv

LIST OF ILLUSTRATIONS           	            ...       •  •     .   •  vii

LIST OF TABLES  .                 ....         ....       .       .   .  -          viii

I.    INTRODUCTION   .         	            .          ...    1

II.   COLLABORATIVE TESTING OF THE METHOD     .              .                             3

     A.    Selection of Location  .....         .   .                     ....       3
     B.    Selection of Collaborators .....         ...        ...       .       4
     C.    Providing Test Samples   ...                     ...                   4
     D.    Collaborative Test Procedure      ...    .    .        .    ...    .       ....    7

III.   STATISTICAL DESIGN AND ANALYSIS (PHASE I, PRECISION)      ....            13

     A.    Design of Experiment  .            .       ...             ...            13
     B.    Results and Statistical Analysis       .   .                 .          .....  13

IV.   STATISTICAL DESIGN AND ANALYSIS (PHASE 11, BIAS)  .     .       .              .22

     A.    Purpose and Scope of the Experiment    .                   ....              22
     B.    Design of the Experiment .        .       .       .        .          ...            .22
     C.    Preliminary Data Analysis          .          ......      .    .        .23
     D.    Presentation of Data   .  .                         ....                 .          24
     E.    Tests for Outlying Observations         .       .       .                 .       .        .25
     F.    Accuracy (Bias) Determination              .       .         ...               .27
     G.    Linear Model Analysis  .            .   .       .  .            .....         .30

V.   SUMMARY AND CONCLUSIONS   .   .         	                 .32

     A.    Phase I (Precision)       .   .              .         .   .              	32
     B.    Phase II (Bias) ...               .                 .          	32

LIST OF REFERENCES    ...        .            ...                 .  .          ...  33

APPENDIX-Reference Method for the  Measurement of Photochemical Oxidants Corrected for
     Interferences Due to Nitrogen Oxides and Sulfur Dioxide                 .                        34

-------
                                   LIST OF ILLUSTRATIONS

Figure                                                                                           Page
   1       Schematic Diagram of Sampling System (Phase I)

   2       Simplified Equipment Connection Diagram (Phase II)
  3       Partial View of the Test Area Showing Some Typical Instruments and the
          Sampling System (Phase I)                .....          ...        .     .    9

  4       Plot of Individual Observations Versus Overall Average for Each Sample (Phase I)   .     .   .        15

  5       Individual Estimates of Total Standard Deviation (S) Versus Concentration (Phase I)  .     .   .   17
  6       Comparison of Total Standard Deviation Estimates Versus Concentration Level From
          Linear Model and Data Points   ...                       .   .          ....   20

  7       Between-Laboratories and Within-Laboratory Standard Deviation Estimates
          Versus Concentration Level (Derived from Linear Model)  .....               .        .21

  8       Design of Ozone Method Test (Phase II)       ......        .        .     ...   24

  9       Expected Versus Average of Observed Values (With 95% Confidence Interval)
          For Five Concentration Levels (Phase II) (Laboratory 11 Censored)  ...             .           30
                                                 vn

-------
                                         LIST OF TABLES
Table
I
II
III
IV
V
VI
VII
VIII
IX
X
XI
XII
XIII
XIV
XV

Calibration Data (Before Test) Ozone by Laboratory by Concentration Level (Phase II) .
Comparison of Results of Initial and Final Calibration at Shutter Setting of 90 (Phase II)
Laboratory Instrument Type and Method of Calculating Hourly Averages (Phase I) ...
Laboratory Instrument Type and Method of Calculating Hourly Averages (Phase II)
Ambient and Supplemented Ozone Atmospheres with Observations for Each Sample
Ambient and Supplemented Ozone Atmospheres with Observations for Each Sample
for Each Laboratory, Omitting Laboratory 9 and Samples 17 and 19 (Phase I)
Means, Slopes, and Standard Errors of Estimate for Linear Model Analysis with
Data Transformation (Phase I) . 	
Summary of Variance Components for Linear Model Analysis with Data
Transformation (Phase I) ... . . 	 	
Sources of Variability and Their Relative Importance for the Linear Model
Analysis with Data Transformation (Phase I) ... 	 . .
Analysis of Variance Table for Linear Model Analysis with Data Transformation (Phase I)
Listing of Total, Between-Laboratory and Within-Laboratory Standard Deviation Derived
From Linear Model Analysis (Phase I) . . . ... ...
Observed Values by Collaborating Laboratories (Phase II) ... . . . .
Observed Values (Adjusted) (Phase II) . . . . . . ...
Transformed Data (Phase II) . ... .....
Within-Laboratory and Cell Means and Standard Deviation for Five Concentration
Page
. 7
. . . 7
. . . 8
. . . 12
. . . 14
. . 16
. . . 18
. . . 18
. . . 19
. . 19
19
23
25
26

           Levels (Phase II)       ...                       ....             ....        27

 XVI       T' Statistic for Identification of Potential Outliers (Phase II)        ....         .  .      28

XVII      Adjusted Data, Laboratory and Concentration Level Averages, and Standard Deviation
           Average (Phase II)     ....       .....                                   ->0
                                                                                                      Zo
XVIII      Ratios of Standard Deviations to Average Standard Deviations of Concentration
           Levels (Phase II)          	
                                                                                                .   .  29
 XIX       Means, Slopes and Standard Errors of Estimate for Linear Model Analysis. Data in
           Transformed Scale (Phase II)       .     .     ....    .   .

 XX       95% Confidence Intervals on Means and Bias (Phase II)  ...
                                                  vin

-------
                                         I.  INTRODUCTION


      Ozone and related chemical contaminants in the atmosphere have been measured for many years to obtain an
 estimate of the nature and extent of a unique type of air pollution commonly termed "photochemical smog"  This
 unique type of air pollution was first identified in the Los Angeles area during the early 1950's, and has since become
 more prevalent in many urban areas throughout the  country. It is formed when organic matter (largely hydrocarbons)
 reacts with nitrogen dioxide in the presence of sunlight to produce a complex mixture of reaction products which in
 turn are responsible for eye irritation, a unique  type of vegetation damage, the formation of haze which restricts
 visibility, and other manifestations  of photochemical smog. Ozone and other highly oxidizing substances are also
 produced in this reaction mechanism.

      For many years, a measurement designated by the catchall term "oxidant" was used as a measure of photo-
 chemical smog. This measurement  was obtained by  measuring the iodine released from potassium iodide solution
 when the air sample was bubbled through, and represented the net total of all of the oxidizing substances minus
 any reducing substances (such as sulfur dioxide) which might also be present.  It was generally recognized that
 the primary component of the oxidant was ozone although oxides of nitrogen, peroxides, and  other constituents
 also contributed to the total oxidant measurement.

      In setting ambient air quality standards under  the mandates of the 1970 Amendments to the Clean Air Act,
 the Environmental Protection Agency selected a chemiluminescent method of measurement rather than the potas-
 sium iodide method since the  former is specific for ozone.  In this method, the air stream is mixed with a stream of
 ethylene, and any ozone  present reacts with the ethylene by a chemiluminescent reaction mechanism. The light
 produced during this reaction is then measured  by a photo-electric device in order to determine the amount of
 ozone present in the original air stream. The instrument is calibrated by synthetic atmospheres using an ozone
 generator which in turn has been calibrated by the potassium iodide method. Commercial instruments are
 available which provide a continuous recording  of the ozone levels being measured.

      In order to obtain reliable data in measuring ozone and other atmospheric contaminants  to determine
 compliance with the ambient air quality standards, the Environmental Protection Agency (EPA) Methods
 Standardization and Performance Evaluation Branch (MSPEB) has been working for some time to develop
 standard methods which could be used by all persons making air quality measurements.  Following the development
 of a tentative standard method, the final step in the  standardization process is a collaborative test, or interlaboratory
 comparison, of the proposed standard method.  This procedure, also called "round-robin testing", has been used to
 evaluate many different methods of measurement in such diverse  fields as water chemistry, metallurgy, paint and
 surface coatings, food and related products, and many others.  A test of this nature by a representative group of
 laboratories is the only way that the statistical limits of error inherent in any method  can be determined with
 sufficient confidence.  With this goal in mind, the two collaborative tests described in this report were conducted
 by the MSPEB and Southwest Research Institute.

      There were two primary objectives of the collaborative tests. The first objective was to derive an estimate
 of the  precision of the Standard Method, and the second was to derive an estimate of the bias of the Standard
 Method.  Throughout the body of this report, the  first test (to determine the precision of the method) will be
 referred to as Phase I, and the  second test (to determine the bias of the method) will be referred to as Phase II.

      In the Phase I collaborative test, a group of participating laboratories gathered at a single location in order
 to run simultaneous tests  on a "real life" atmosphere typical of photochemical smog.  The results of this collab-
 orative test were then analyzed statistically to determine the precision of the proposed method.

     In the Phase II collaborative test, each of the ten participants worked in his own laboratory and a standard
ozone source for five concentration levels was provided to each of the  test collaborators.  Each collaborator pre-
pared  an IT versus absorbance  graph for various  standard iodine solutions and calibration curves for the monitoring

-------
instruments, according to the method (see Appendix A). The collaborators then used the method to analyze the
ozone levels generated by the standard.

     This report describes the design of the experiments, the statistical analysis, and the results of the two collab-
orative tests.

-------
                        II. COLLABORATIVE TESTING OF THE METHOD


      An important step in the standardization of any method of measurement is collaborative testing to determine,
 on a statistical basis, the limits of error which can be expected when the method is used by a typical group of inves-
 tigators. The collaborative, or inter-laboratory, test of a method is an indispensable part^ of the development and
 standardization of an analytical procedure to insure that 1) the procedure is clear and complete and that, 2) the pro-
 cedure does give results with precision and accuracy in accord with those claimed for the method. Among other
 organizations, the Association of Official Analytical Chemists (AOAC) and the American Society for Testing and
 Materials (ASTM) have been active in the field of collaborative testing and have published  guidelines of the proper
 procedure for conducting collaborative tests and evaluating the data obtained.^2"5) Publications of both  organizations
 were used extensively in planning and conducting the collaborative tests of this method to measure ozone.

      The first phase of the collaborative test involved the assembly of ten collaborators at a single site to measure
 ozone concentrations in ambient air to allow determination of the precision of the method. This phase of the test
 was conducted in Pasadena, California, during  the month of September, 1972. This report describes the results of
 that test. The second phase involved the measurement of synthetic ozone atmospheres by the collaborators working
 in their own laboratories  to determine the accuracy of the method.

 A.   Selection of  Location

      1.    Phase I (Precision)

           Since a synthetic mixture duplicating all of the characteristics of photochemical smog is difficult to
 prepare and standardize in the laboratory, it was decided to conduct an "on site" test in which participants would
 sample an ambient atmosphere under photochemical smog conditions. Since Los Angeles  has been the prime ex-
 ample of an urban photochemical smog problem for many years, and since this type of air pollution occurs more
 frequently in Los Angeles than in other cities, it was decided  to conduct this test in the Los Angeles area.

           In order to select a suitable location, long-term monitoring data obtained from the  Los Angeles
 County Air Pollution Control District were examined in order to determine  the frequency and severity of photo-
 chemical smog in different portions of this urban area.  This agency has maintained a comprehensive monitoring
 program for a number of years to measure oxidant and other  contaminants throughout Los Angeles County to
 provide information for a vigorous ongoing program of air pollution control.

           Of the various stations in this monitoring network, it was found that two stations showed the highest
 consistent levels of oxidant, one in Pasadena and one in Azuza.  Both of these cities are located north and north-
 east of the city of Los Angeles itself, close to the mountain range which  forms the northern boundary of the air
 basin and thus influences the degree  of air stagnation which frequently occurs. Because of the meteorology and
 topography of the area, it is not surprising that the Pasadena-Azuza area should consistently show the highest
 levels of oxidant throughout the entire area.

           So far as  time was concerned, it has long been known that the most severe smog season occurs in the
 summer and early fall. Therefore, it  was no surprise to find that the monitoring records indicated the most fre-
 quent occurrence of photochemical smog and the highest oxidant levels in the months of August, September,
 and October.

           With this background information as a guide, suitable time and place for the collaborative test was
next investigated. Through the cooperation of the California  Institute of Technology in Pasadena, arrangements
were made to conduct the test at an engineering laboratory building on  the campus during the period between
the summer and fall semesters.  Fortunately, both the location and the time of year were favorable with respect
to the probability of occurrence of high oxidant levels.

-------
      2.    Phase 11 (Bias)

           The locations of the Phase II test were fixed by the location of the individual collaborators, and
represented various geographic areas of the continental United States.

B.    Selection of Collaborators

      1.    Phase I (Precision)

           If a collaborative test is to achieve the desired objective, it is desirable that the participants in the test
be representative of the large group that will ultimately use the method being tested.  Since air pollution measure-
ments are made by many different groups, it is desirable to include in the group of collaborators a representative
mix of those people making the pollution measurements. Final selection of participants included two from federal
laboratories, five from state and local air pollution control agencies, two from industry, and one from a research
institution. A list of the participants completing the test is included in the  Acknowledgment.

           Even more important than the type of laboratory is the degree  of skill and experience of the persons
who participated. Each laboratory was asked to assign a person to this test who had previous experience with the
chemiluminescent method for measuring ozone and was competent in carrying out measurements by this method.
This places emphasis upon the capabilities of the method rather than the performance of the laboratories.

      2.    Phase 11 (Bias)

           The rationale for selection of collaborators in Phase II was the same as for Phase I.  For Phase II,
eleven collaborators were selected, nine of whom had participated in the first phase. Of the eleven, two were
from governmental agencies, six from local and state air pollution control agencies, two from industry, and one
from a research institution. Ten collaborators completed the test. A list of the participants is included in the
Acknowledgment.

C.    Providing Test Samples

      1.    Phase I (Precision)

           In the study, test samples were provided by two different methods, each of which is described in
detail below.

           a.    Ambient Atmospheres

                A schematic diagram of the sampling system is shown in Figure 1. In the ambient air sampling
mode, Valve A is closed and Valve B is opened. (Valve B is a safety pressure release for the ozone generation
system).  In this configuration, ambient air is brought in through a 3/4-in. diameter Teflon tube approximately
8-ft long extending through the window of the test room. The sampling line terminates on the outside of the
building with a 90° bend pointed downward for protection in case of rain.  All lines in contact with ozone are
either of glass or Teflon. No pumps are required, and the ozone instruments draw air through Teflon intake
lines from the common glass manifold. The glass manifold was designed, constructed,  and tested in the Houston
Laboratory prior to its installation in the test area.  It was also rechecked after installation.  A manometer on one
of the ports of the manifold can be observed periodically to check pressure  drops which were always less than
1/2-in. of water vacuum. The flow rate through the system in this manner is approximately 1 liter per minute
per instrument for a total of about 10 liters per minute.  A trap was installed in front of the manifold to coUect
condensate or particulate matter and keep it from entering the sample manifold.  This  trap proved to be unnec-
essary for this purpose since neither condensate nor particulate matter ever  accumulated.  A thermometer was
also incorporated in the sampling manifold to periodically monitor temperature of ambient air.  The ozone
instrument discharge lines were connected to a glass manifold  similar to the intake manifold to prevent ethylene
buildup in the test area.  This discharge manifold was connected by Tygon tubing approximately 30-ft long ex-

-------
                            Ozone
                          Generator
    Purified
   An Source
 Flowmeter
and Control
                                             t><] Valve A
       Ambient Air Sample Line
                              Valve B
                             -B-
Ozone Instrument
   (1 of 10)
                Sample Intake
                                            Cyclone Trap
                                               Mixer
                                             Sampling
                                             Manifold
                                     Discharge
                                     Manifold
                Sample Discharge
                                             to
                                            roof
tending to the roof of the three-
story building. The pressure in the
discharge manifold was consistently
under 1 inch of water.  This system
operated satisfactorily in every
respect.

           b.    Atmospheres Con-
                taining Supplemental
                Ozone
     FIGURE 1.  SCHEMATIC DIAGRAM OF SAMPLING SYSTEM
                          (Phase I)
                                                                                         Ambient atmospheres
                                                                         containing supplemental ozone were pro-
                                                                         vided by the same system shown in Fig-
                                                                         ure 1 with Valve A open and Valve B
                                                                         closed.  Operation of the ozone gener-
                                                                         ation system was required in this mode.
                                                                         Purified air was obtained from com-
                                                                         pressed air cylinders (dry grade, with
                                                                         dew point -75°F maximum).  Car-
                                                                         tridge type purifiers were used to
                                                                         remove oil, water, and particles larger
 than 12 microns. Dew points of — 100°F were thus achievable. The flow of purified air was controlled by a needle
 valve and a mass flowmeter on each of three ozone generators. Three ozone generators (NBS Technical Note 585) were
 required to provide the quantity needed for synthetic atmosphere generation. The purified air flow rate was adjusted
 to approximately 1 liter per minute. In this manner, approximately 9 liters per minute of ambient air were mixed
 with 1 liter per minute of a synthetic ozone-air mixture to provide an ambient air test sample containing supplemental
 ozone.  The cyclone trap in front of the manifold served as a mixing chamber and insured thorough mixing of ambient
 air and synthetic ozone mixtures prior to entering the sampling manifold. Adding supplemental ozone to ambient
 atmospheres was never incorporated during the daytime hours in order that normal  daily variations could be observed.
 As is obvious, the flow into and out of the sampling manifold was constant at all times, and the magnitude of this
 flow was equal  to the sum of each of the individual ozone instrument sampling flow rates.

                 The ozone generators were adjusted to bring the total ozone concentration in the sampling manifold
 to the desired level. No attempt was made to measure the amount of supplemental  ozone added. It was important to
 incorporate  this feature into the design of the sampling system in order to be able to achieve higher concentration
 levels for test purposes.  In the event that higher levels would occur naturally, then adding ozone would not be re-
 quired.  The site could be anticipated to experience ozone levels perhaps as high as 0.2 ppm under usual circumstances
 at this time of the year. Concentrations between 0.2 to 0.5 ppm could probably be achieved only by adding sup-
 plemental ozone. The quantity  of air through the ozone generation system was kept low in order to minimize
 dilution of other interfering substances present in ambient air.

                 No problems were encountered in this mode of operation, and the sampling system performed
 satisfactorily in every respect.

     2.    Phase 11 (Bias)

           It was necessary to have known values of samples in order to evaluate the bias or accuracy of the test
 method. These known values were obtained from calibrated ozone generating equipment furnished by the National
 Bureau of Standards (NBS). A set of generating equipment was supplied to each collaborator, and the specified
measurements of ozone samples were made by the collaborator at his own location.

           A schematic diagram of the ozone generating system is shown in Figure  2. The equipment between the
dotted lines was constructed and supplied by NBS.  The filter contains charcoal to remove hydrocarbons and
nitrogen dioxide, and a molecular sieve to remove water.

-------
    Cylinder
    Air Supply
     (0-30 PS I)
                                                                                 Needle Valve
Low Pressure
 Regulator
Filter
                            (10 PSI)
               Glass
             Rotameter
            Generator
                                      Manifold
                            Chemiluminescent
                                Instrument
                        FIGURE 2. SIMPLIFIED EQUIPMENT CONNECTION DIAGRAM
                                               (Phase II)

           Each ozone generator was calibrated by NBS as a unit using its specifically designated power supply, air
regulator and rotameter. The ozone concentrations produced by each generator were compared at several shutter
settings with the ozone produced by a standard generator using a Chemiluminescent ozone analyzer as a comparator.

           The standard generator was initially calibrated by the 1.0% neutral buffered potassium iodide method
(NBKI).  The generator was then used to generate N02 by gas phase titration (GPT) of NO (NO + 03 -> NO2 + 02)
where the concentration of N02 produced was based on the O3  concentration determined by NBKI. The response
of the NO2 produced by GPT was then measured with a Chemiluminescent NO2 instrument that had been calibrated
with a gravimetrically calibrated N02 permeation device (NBS Standard Reference Material 1629). Over the range
0.1 to 0.8 ppm NO2, the response to the two sources of NO2 was identical, within experimental error.  Thus, the
accuracy of the O3 concentration produced'by the standard is related to the accuracy of the permeation device
(±2%) and is independent of the NBKI method.

           The flow rate  through the systems was approximately 5 liters per minute. The flow rate (to maintain
calibrations) for each individual set of generating equipment was marked in red on the rotameter. The generator
output was corrected to 760 mm of mercury and 25°C. The individual collaborators then corrected for the de-
viation of ambient temperatures and pressures from the standard. The range of reported  temperature was from
20-29.8 C, and the range of atmospheric pressure was from 731-765.8 mm of mercury.

           The calibrated generator, pressure regulator, filter, rotameter and fittings were supplied as a unit in
order to maintain control  over the system as a "standard". The collaborator had only to  connect the air cylinder
and adjust the pressure and flow rate to the indicated level. Temperature and pressure measurements were made
by the collaborator during the  test and recorded. The volume of air sampled by the collaborators was then ad-
justed by taking into account the pressure and temperature reference conditions at the  time of calibration as
specified by the method.

           The ozone generators were the same type as were used in Phase I, and were capable of supplying
ozone concentrations from zero to 0.500 ppm with precision and stability in this mode of operation  Instructions
on the set up and operation of the generating equipment were sent to each collaborator.  The  before test calibration
data for each of the systems is  shown in Table I.  From Table I, the average of each of the five concentration
levels  over all labs is; 0.054, 0.084, 0.131, 0.273, and 0.481 ppm.

-------
TABLE I. CALIBRATION DATA (BEFORE TEST) OZONE
   BY LABORATORY BY CONCENTRATION LEVEL
                    (PHASE II)
Laboratory
Number
10
11
13
14
15
16
17
18
19
20
Ozone Concentration Level Number (ppm)
1
0.053
0.054
0.054
0.054
0.054
0.054
0.054
0.054
0.054
0.053
2
0.082
0.084
0.084
0.084
0.084
0.084
0.084
0.084
0.084
0.082
3
0.129
0.131
0.131
0.131
0.132
0.132
0.131
0.132
0.132
0.129
4
0.269
0.274
0.274
0.274
0.274
0.274
0.274
0.274
0.274
0.269
5
0.475
0.482
0.482
0.482
0.484
0.484
0.482
0.484
0.484
0.475
Average
Over 0.054 0.084 0.131 0.273 0.481
All Labs
TABLE II.  COMPARISON OF RESULTS OF INITIAL AND
  FINAL CALIBRATION AT SHUTTER SETTING OF 90
                    (PHASE II)
                                                               A comparison of the before and after cali-
                                                    bration is given in Table II. Section IV, A. of this report
                                                    contains a discussion of these results.

                                                    D.   Collaborative Test Procedure

                                                         1.    Phase I (Precision)

                                                               An immense amount of planning and coordi-
                                                    nation was required in order to conduct the test. Each
                                                    participant was provided with rather detailed instructions
                                                    regarding what was required. The responsibilities of a
                                                    participant can be divided into four general areas which
                                                    are (1) calibration work in his own laboratory prior to
                                                    coming to the test site; (2) transfer of equipment to and
                                                    from the test site; (3) activities at the test site; and (4)
                                                    processing of data after return from the test site.

                                                               Four different types of chemiluminescent
                                                    instruments were represented in the test and are listed
                                                    in Table III. A cross-section of available instruments
                                                    was desirable; however, it was not the purpose of the
                                                    experiment to compare instruments from different
                                                    manufacturers.  Because of the different instruments
                                                    used in the test, it was impossible to give specific
                                                    instructions applicable for all. The general recommen-
                                                    dations to collaborators were to follow specific instruc-
                                                    tions given in the manufacturer's manual for the instru-
                                                    ment, to make proper adjustments of flow controls and
                                                    zero and span controls, and to record whatever para-
                                                    meters were necessary  for future reference.

                                                               Each collaborator was instructed to perform
                                                    all possible calibration procedures in: his home laboratory
                                                    as if the instrument was being prepared for placement at
                                                    a monitoring site. Whatever supplemental calibration
procedures (either recommended by the instrument manufacturer or considered necessary by the collaborator)
that were essential were performed at the test site. Prior to coming to the test site, each collaborator was instructed
to prepare a potassium iodide calibration curve according to Section 8.1 of the method (see method in Appendix).
These data could be taken from an existing calibration curve then in use provided the curve was prepared in accor-
dance with Section 8.1.1  and provided that actual data points could be reported.  In other words, if a valid potas-
sium iodide calibration curve was in use in a collaborator's laboratory, its use was permitted for the purposes of
this test, and it was not necessary that another be prepared. It was not intended (but not precluded) that potas-
sium iodide determinations be made at the test site; therefore, the essentials for that determination were made
available at the test site should they be needed in any emergency case. The collaborators were instructed to
transfer the instrument calibration reference points to the test site by proper standardization of internal or
external ozone generators in order to minimize biasing the results.  Calibration procedures were to cover the
range 0 to .510 ppm (0 to 1000/ug/m3) which is the  range over which the precision information was desired.
The reference points were not significantly changed by shipment.

           Considerable  effort was required by a collaborator to prepare, pack, and ship his equipment to and
from the test site. Each collaborator was expected to bring the following items to the test site:
Generator
Number*
10
11
13
14
15
16
17
18
19
Initial
(ppm of 03)
0.49
0.49
0.49
0.48
0.45
0.53
0.48
0.48
0.50
Final
(ppm of O3 )
0.48
0.46
0.44
0.48
0.45
0.53
0.48
0.48
0.43
Change
in Percent
-2
-6
-10
0
0
0
0
0
-14
*Generator Number 20 was damaged in return
shipment.

-------
           1.
           2.
           3.
Chemiluminescent ozone monitor complete with sample ethylene and exhaust lines,
connecting cables, and ethylene pressure regulator.  An individual ethylene cylinder was
provided for each collaborator at the test site.

Recorder complete with connecting cable, chart paper, ink supply, and appropriate chart
speed motor, or gears to allow continuous recording for a period of five days.

Calibrated ozone generator (internal or external) with appropriate air filter, flowmeter,
and air pump if required.  Clean compressed air was available at the test site if necessary.
           4.    Miscellaneous tools to assemble and install equipment.

           5.    Other items considered necessary by the collaborator for the proper installation and
                operation of the ozone monitor.
TABLE III. LABORATORY INSTRUMENT TYPE AND METHOD
      OF CALCULATING HOURLY AVERAGES (Phase I)
Laboratory
Number
1
2
3
4
5
6
7
8
9
Instrument
Type
Meloy
McMillan
Bendix
McMillan
Bendix
REM
McMillan
Bendix
REM
Method of Calculating
Hourly Averages
A line representing hourly average
was visually constructed
Obtained by area averaging method
Visually constructed line of
best fit
Counting squares
Taken by visual averaging of
strip chart recording
Trace, cut, and weigh
Visual estimation of equal areas
Planimeter
Average of digital output
                                                   Upon arrival at the test site, the
                                         collaborator's first responsibility was to set up and
                                         check out his instruments.  Connection of the intake
                                         and discharge lines from each of the instruments to
                                         the sample system was handled by Southwest Re-
                                         search Institute personnel.  In general, there were
                                         no unusual problems; however, the instrument of
                                         one collaborator was damaged in shipment.  At first
                                         it was believed that the instrument could not be
                                         brought into operation.  However, a quick check
                                         indicated a broken electrical connection, a broken
                                         quartz tube in the ozone generator, and a bent fan
                                         cowl. Through the resources of the host, EPA
                                         personnel, Southwest Research Institute staff, and
                                         other collaborators, the instrument was repaired
                                         and in operation by mid-afternoon on Monday,
                                         the first day of the test.  Though noisy, the instru-
                                         ment performed satisfactorily throughout the test.
                                         The test area was somewhat small, and each collab-,
                                         orator was slightly crowded; however, no unusual
                                         problems were encountered. A partial view of the
                                         room and the sampling system is shown in Figure 3.
           During the test period, a collaborator's responsibilities were generally limited to daily zero and span
checks of his instrument, normally scheduled between 8 and 9 a.m. each morning of the test period. In general
a collaborator was present in the test area only during this time. All equipment was monitored by Southwest
Research Institute personnel, and if a malfunction occurred, the respective collaborator was requested to come
to the test site and make repairs.  The only problems of this type occurring were an occasional failure of a re-
corder pen.  These problems were not serious, and little or no data were lost as a result. The generation of test
atmospheres did not require any action on the part of the collaborators. All atmospheres to be analyzed  were
passed through the manifold, and no collaborator action was required during the change from one mode of
operation to another.

           So far as the collaborators were concerned, there was no distinction between ambient atmospheres
and ambient atmospheres containing supplemental ozone. In both cases, hourly averages were used.

           Eash instrument was in continuous operation, and during the daylight hours was measuring
ambient ozone concentrations. During the evening or nighttime hours, ozone supplemented atmospheres

-------
                  Photo by Don Ivers - CalTech
FIGURE 3.  PARTIAL VIEW OF THE TEST AREA SHOWING SOME TYPICAL
INSTRUMENTS AND THE SAMPLING SYSTEM  (PHASE I)

-------
were created and measured.  On one night, the instruments measured ambient air all night long while supplemented
atmospheres were measured during part or all of the remaining nighttime hours of the test.

            Southwest Research Institute personnel assisted collaborators, and dismantled and packed the sampling
system, and cleared and cleaned the test area.

            The collaborators took their recorder charts home with them to await1 instructions for data processing
and reporting.  The data from the instrument operated by Southwest Research Institute as a collaborator was in-
spected, and a series of hourly periods for calculation of hourly averages was specified. These periods were to
some extent randomly selected, excepting that they were usually specified between the hours or between the half
hours. They were selected so that they did not encompass any step changes created as a result of changing from
ambient atmospheres to supplemented atmospheres.

            A detailed set of instructions and data forms was transmitted to each collaborator.  These instructions
were supplied to guide the collaborator through the preparation and submission of his report. Each observation
to be reported was identified by the date and by start and finish times. A special data form greatly facilitated
data reporting and handling.

            There were 51 hourly averages specified for computation—32 for ambient and 19 for ozone supple-
mented samples. For these observations, the collaborator was instructed to report the hourly average between
the times shown.  The method allows the collaborator to use the averaging method of his choice; however, he
was requested  to report the method in detail.  The various methods are indicated in Table III.

            Collaborators were asked to report their results to the nearest 0.001 ppm if significant for their
equipment. If this sensitivity was not obtainable with their system, they were requested to report less significant
figures.  They were asked not to report trailing zeros if their system was not capable of producing results to the
nearest 0.001 ppm—for example, 0.07 should be reported instead of 0.070.

            The collaborator's final report consisted of the foil owing items:

            1.    Potassium iodide calibration curve.  Submitted on a special form.

            2.    Instrument calibration curve. This curve was to reflect zero offset, if any, and sensitivity
                 in terms of ppm per unit of instrument response so that Southwest Research Institute personnel
                 would be able to interpret the analog records.

            3.    Collaborative testing data form. Collaborators were encouraged to make comments, to explain
                 any unusual or atypical results, and to include notations regarding the individual results.

            4.    A  description of the method of obtaining hourly averages.

            5.    Original analog records. The original chart records were requested, and any collaborator who
                 wished to have these records returned after the completion of data analysis was asked  to
                 so specify.

            6.    Comments and criticisms.  Collaborators were invited and encouraged to submit any such
                 comments they felt important. Comments could involve the method, the equipment, the
                 facilities provided, the instructions, or the manner in which the collaborative test was planned
                 and conducted.

           The collaborators apparently experienced no problems in interpretation of their analog records and
calculation of average concentrations.  Nine of the ten participating laboratories submitted results.  As soon as
results were received by Southwest Research Institute, they were screened for any unusual observations-that is,
unusual in comparison with the results overall.  In one or two instances, collaborators were asked to recalculate


                                                    10

-------
 one or more suspicious looking averages.  The data were then coded into a computer compatible format and
 stored on the disc system of a time sharing computer service. This data file served as a basis for all subsequent
 statistical analysis.

       2.    Phase 11  (Bias)

            Detailed instructions were provided to each collaborator on the test procedure.  Portions of these
 instructions were unique to each collaborator (generator settings, concentration level, sampling order, etc.).

            The following equipment was to be supplied by each collaborator:

            a.    Chemiluminescent ozone monitor complete with sample, ethylene, and exhaust lines;
                 connecting cables; and ethylene pressure regulator and cylinder of ethylene.

            b.    Recorder complete with connecting cables, chart paper supply, ink supply, and appro-
                 priate chart motor speed.

            c.    Source of purified cylinder air for use with NBS calibrated ozone generator.

            d.    Absorbers, flowmeter, and spectrophotometer for potassium iodide determinations.

            Each collaborator was required to perform the following functions, in accordance with the test
procedure instructions and/or the test method:

            a.    Set up and operate the EPA supplied ozone generating equipment in accordance with the
                 instructions provided with the equipment.

            b.    Prepare a curve of absorbance of various standard iodine solutions versus calculated ozone
                 equivalents  according to Section 8.1 of the Method (Appendix).

           c.     Each day of the test prepare a calibration curve of instrument response versus ozone con'
                 centration as determined by the KI method per Section 8.2 of the Method, utilizing the
                 instrument manufacturer's instructions as necessary.

           d.    Sample and analyze the test atmospheres as per the test procedure instructions.

           At the completion of the test, the collaborators were to submit a final report consisting of the
following items:

           a.     Potassium iodide calibration curve.

           b.    Instrument calibration curve.  This curve was to reflect zero offset, if any, and sensitivity
                in terms of ppm per unit of instrument response so that Southwest Research Institute
                 personnel would be able to interpret the  analog records.

           c.    Collaborative testing data form. Collaborators were encouraged to make comments, to explain
                any unusual or atypical results.

           d.    A description of the method of calculating hourly average ozone concentrations from the
                analog records.

           e.    Original analog records.

           f.    Comments and criticisms.  Collaborators  were invited and encouraged to submit any such comments
                they felt were important.


                                                   11

-------
TABLE IV.  LABORATORY INSTRUMENT TYPE AND METHOD
    OF CALCULATING HOURLY AVERAGES (PHASE II)
Laboratory
Number
10
11
13
14

15
16
17
18
19

20
Instrument
Type
MEC
Bendix
Meloy
MEC

REM
REM
Bendix
Bendix
Bendix

Bendix
Method of Calculating
Hourly Responses
Visual averaging of hour readings
Visual averaging of hour readings
A straight line through chart
response was visually constructed.
Average values taken from midpoint
of line.
Visual observation of strip chart
reading for hourly run, subtraction
of offset, and value entered on
calibration curve and read off.
Arithmetic average of 60 print-outs
(one per minute) from recorder.
Visual observation and averaging
of hourly readings on chart records.
Visual observation and averaging
of hourly readings on chart records.
Visual averaging the hourly readings
on strip chart.
Visual observation of the chart
records.
Visual estimating the average over
the hour from the strip chart.
          Table IV presents a list by laboratory
number of instrument types and method of calculating
hourly averages of concentration level measurements
for those laboratories that completed the test.
                                                12

-------
                           III.  STATISTICAL DESIGN AND ANALYSIS
                                        (PHASE I, PRECISION)


 A.   Design of Experiment

      In collaborative testing, two general sources of variability can be readily detected. First, the variability
 between laboratories can be estimated. This is frequently the largest source of variability and is not under the
 control of the investigator. Second, the within-laboratory variability can be estimated. The separate sources
 which make up the within-laboratory variability cannot be estimated in this test.  The separate sources are lumped
 into a single variable including the variability between days and the variability between replicates, neither of which
 can be individually defined.

      In order to achieve the above goal, two distinct statistical analyses were used.  The results of both were
 coordinated into what was believed to be the best description of the precision of the method.

      The first method involves the computation of standard deviation for each respective sample.  Each sample
 standard deviation provides an estimate of the total process variability at the sample concentration level, and
 includes both the between-laboratories and within-laboratory components.  However, the first method yields
 no information regarding the within-laboratory or between-laboratories components of the  total variance. (6)

      The second method, the linear model analysis/5' yielded estimates of the total standard deviation (Sj),
 between-laboratory standard deviation (S^) and within-laboratory standard  deviation
      The linear model analysis also served as a good method in itself for detecting outliers. The method is
 described in detail in the respective reference, and the interested reader may consult that reference.

 B.   Results and Statistical Analysis

      The data yielded by the experiment for ambient and supplemented atmospheres are tabulated in Table V.
 These data have been edited to the extent that only samples for which all laboratories were able to produce an
 observation are recorded.  These data can be seen to consist for each laboratory of 29 observations for ambient
 atmospheres, and 19 observations for supplemented atmospheres.  Several preliminary analyses were made to
 determine how the data might best be grouped for analysis. In this preliminary examination, the linear model
 analysis was applied to the ambient atmosphere data alone, to the  supplemented atmosphere data alone, and to
 the ambient and supplemented atmosphere data combined. It resulted that the combination of ambient atmo-
 sphere data and supplemented atmosphere data would provide  the best analysis. The results from the subsequent
 analysis reinforced this conclusion. These analyses also revealed three sample outliers and one laboratory outlier.
 These outliers may be observed by referring to Table V.  In this table, the data are arranged in order of increasing
 sample means from top to bottom of the table and increasing laboratory means from left to right in the table.
 The high standard deviations for Samples  17 and 19 (lines 5 and 6 in the table) are obviously due to atypical
 results by Laboratory 3.  The "studentized" ratio outlier test(7'8) was used, and individual observations in the
 body of the table are followed by a dagger if they are significant outliers (a = 0.05) and an asterisk if they are
 highly significant outliers (a = 0.01).  The discarding of the two sample outliers does not cause any significant
 changes in results or interpretations because of the large  number of valid data points which are still available.
However,  the rejection of Laboratory 9 deserves more careful consideration.

      Further examination of the  data for Laboratory 9  in Table V reveals a total of 13 outliers among the 48
samples.  Five outliers occurred in the ambient sample data, and eight outliers occurred in the supplemented
atmosphere  data. (Sample Nos. 1  through 29 are ambient atmosphere data, and Sample Nos. 30 through 48 are
supplemented atmosphere data.) The total number  of outliers represents about 25 percent of the observations
by this laboratory. Likewise, the overall average obtained by Laboratory 9 is an outlier with respect to the set
of laboratory averages as shown  at the bottom of the table.  Two approaches were possible. The entire data for
Laboratory 9 could be eliminated  from the analysis, or every sample where an outlier is present could be  elim-
                                                    13

-------
 TABLE V. AMBIENT AND SUPPLEMENTED OZONE ATMOSPHERES
WITH OBSERVATIONS FOR EACH SAMPLE FOR EACH LABORATORY
                       (PPM)
Rank
Sample
Laboratory Number
9
5
7
8
6 | 4
3
1
2
Mean
Std
Dev
Ambient
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
32
33
35
29
3
18
2
19
17
4
28
1
9
8
7
6
5
10
27
15
20
26
16
11
21
22
12
25
24
14
23
13
0.013*
0.010
0.003
0.021
0.003
0.008
0.019
0.022
0.035
0.025
0.035
0.037
0.041
0.043
0.054
0.054
0.05 it
0.066
0.064t
o.oest
0.075
0.078
0.085
0.085
0.091
0.094
0.1151"
0.126
0.146
0.000
0.005
0.005
0.015
0.010
0.010
0.022
0.030
0.032
0.028
0.038
0.042
0.045
0.050
0.058
0.065
0.068
0.070
0.082
0.082
0.080
0.085
0.095
0.100
0.105
0.110
0.150
0.150
0.160
0.000
0.000
0.005
0.001
0.010
0.010
0.024
0.033
0.020f
0.038
0.052
. 0.057
0.052
0.057
0.061
0.062
0.070
0.075
0.080
0.080
0.080
0.090
0.099
0.094
0.100
0.108
0.146
0.146
0.155
0.000
0.004
0.002
0.014
0.008
0.010
0.030
0.028
0.033
0.030
0.042
0.050
0.048
0.053
0.062
0.066
0.065
0.071
0.083
0.085
0.085
0.087
0.100
0.101
0.111
0.112
0.153
0.155
0.166
0.000
0.005
0.005
0.015
0.012
0.010
0.025
0.029
0.035
0.040
0.053
0.053
0.056
0.060
0.066
0.068
0.072
0.073
0.092
0.095
0.097
0.090
0.105
0.108
0.117
0.120
0.167
0.162
0.182
-0.003
0.003
0.007
0.013
0.011
0.010
0.026
0.030
0.028
0.030
0.043
0.048
0.050
0.054
0.067
0.068
0.070
0.072
0.082
0.086
0.088
0.084
0.100
0.099
0.110
0.115
0.155
0.158
0.170
0.002
0.007
0.008
0.017
0.035*
0.045*
0.027
0.037
0.035
0.037
0.045
0.052
0.052
0.056
0.070
0.075
0.075
0.075
0.087
0.090
0.081
0.080
0.100
0.102
0.107
0.115
0.165
0.157
0.175
0.001
-0.005
0.006
0.007
0.013
0.008
0.029
0.033
0.030
0.033
0.046
0.053
0.053
0.057
0.065
0.071
0.070
0.073
0.086
0.088
0.090
0.089
0.102
0.104
0.112
0.117
0.160
0.157
0.172
0.000
0.005
0.005
0.015
0.015
0.015
0.025
0.040
0.035
0.030
0.045
0.050
0.050
0.055
0.065
0.080
0.075
0.080
0.095
0.090
0.090
0.095
0.110
0.110
0.130
0.125
0.165
0.175
0.175
0.002
0.004
0.005
0.013
0.013
0.014
0.025
0.032
0.032
0.033
0.044
0.049
0.049
0.054
0.063
0.068
0.068
0.073
0.084
0.085
0.085
0.087
0.099
0.100
0.109
0.113
0.153
0.154
0.167
0.005
0.005
0.002
0.006
0.009
0.012
0.004
0.005
0.005
0.005
0.006
0.006
0.005
0.005
0.005
0.008
0.007
0.004
0.009
0.009
0.007
0.005
0.007
0.008
0.011
0.009
0.016
0.013
0.011
Supplemental Ozone Added
27
28
29
30
31
34
36
37
38
39
40
41
42
43
44
45
46
47
48
37
36
38
34
35
39
40
42
41
46
47
48
43
33
44
32
31
30
45
Average
Std Dev
0.077
0.077
0.087t
0.103
0.098t
0.1 lit
0.128*
0.140t
0.1 48t
0.229
0.248
0.254
0.254
0.280
0.285
0.298
0.317t
0.31 9t
0.320
O.lllt
0.096
0.122
0.122
0.132
0.132
0.135
0.155
0.168
0.178
0.180
0.262
0.278
0.285
0.298
0.322
0.325
0.345
0.362
0.368
0.370
0.132
0.111
0.151
0.156
0.160
0.164
0.164
0.170
0.179
0.184
0.188
0.258
0.272
0.277
0.348
0.343
0.353
0.362
0.377
0.390
0.382
0.139
0.117
0.150
0.148
0.157
0.158
0.160
0.173
0.181
0.189
0.192
0.278
0.291
0.297
0.346
0.348
0.369
0.365
0.381
0.386
0.407
0.142
0.120
0.123
0.125
0.137
0.140
0.145
0.162
0.183
0.200
0.203
0.272
0.292
0.298
0.310
0.347
0.342
0.377
0.400
0.427
0.410
0.144
0.120
0.155
0.154
0.162
0.161
0.162
0.177
0.184
0.193
0.197
0.289
0.290
0.295
0.359
0.358
0.378
0.373
0.387
0.392
0.414
0.145
0.123
0.142
0.140
0.145
0.155
0.155
0.161
0.182
0.195
0.197
0.285
0.307
0.310
0.327
0.355
0.355
0.382
0.395
0.405
0.420
0.146
0.121
0.172
0.173
0.179
0.185
0.182
0.188
0.190
0.197
0.203
0.280
0.289
0.293
0.392
0.390
0.401
0.406
0.417
0.421
0.429
0.152
0.131
0.150
0.150
0.160
0.160
0.165
0.185
0.195
0.210
0.215
0.310
0.330
0.340
0.365
0.365
0.390
0.390
0.405
0.405
0.445
0.154
0.130
0.138
0.138
0.147
0.151
0.152
0.165
0.176
0.187
0.191
0.274
0.289
0.294
0.333
0.345
0.355
0.366
0.383
0.390
0.400
0.141
0.119
0.028
0.028
0.027
0.023
0.024
0.023
0.020
0.020
0.019
0.023
0.023
0.023
0.041
0.031
0.035
0.031
0.030
0.032
0.038
0.013

*Outlier, significant at a = 0.01.
f Outlier, significant at a = 0.05.
                         14

-------
 inated from the analysis, or every sample where an outlier is present could be eliminated from the analysis.
 In view of the fact that all but three outliers in the entire table occurred in Laboratory 9 data, it seemed most
 appropriate to delete Laboratory 9 data from all subsequent analyses. Therefore, all subsequent statistical
 analyses were made deleting Laboratory 9 and deleting Samples Nos. 17 and 19, leaving a total of eight labor-
 atories and 46 samples. As graphic evidence of the appropriateness of this decision, Figure 4, which includes
 data from Laboratory 9 is presented.  In this plot, each individual observation is plotted versus the respective
 average of each laboratory's observation for that sample. The numerals appearing in the plot represent the
 laboratory numbers.  Where two or more observations fall in the same plotting position, an "x" appears.
 For further clarification, a line of slope = 1 has been superimposed on the plot.  If agreement was perfect, all
 points would appear on this line.  The relative departures of the data from Laboratory 9 are now clearly evident.
                  0.5
                  0.4
                  0.3
                S 0.2
                  0.1
                                 0.1
                                              0.2           0.3

                                               Concentration, ppm
                                                                        0.4
0.5
                     FIGURE 4. PLOT OF INDIVIDUAL OBSERVATIONS VERSUS OVERALL
                                   AVERAGE FOR EACH SAMPLE (Phase I)

      Table VI is analogous to Table V and shows the corresponding results following the deletion of Laboratory
9 and Samples Nos. 17 and 19.  Comparison of the standard deviations shown in these two tables reveals the
degree of inflation of variances which would result if the data for Laboratory 9 were retained in the analysis.

      Each of the standard deviations in Table V is an estimate of the total variability for the method.  These
standard deviations (^jp) have  been plotted versus their respective average concentration in Figure 5.  This plot
illustrates the relationship of the total standard deviation to concentration. Superimposed on the plot is the line
of best fit, assuming a linear relationship and using the least squares method. The equation of this line  is as follows:
                                   = 0.0018 + 0.0542* concentration (ppm)
                   (1)
                                                    15

-------
 TABLE VI. AMBIENT AND SUPPLEMENTED OZONE ATMOSPHERES
WITH OBSERVATIONS FOR EACH SAMPLE FOR EACH LABORATORY,
       OMITTING LABORATORY 9 and SAMPLES 17 and 19
                       (PPM)
Rank

Sample

Laboratory Number
5
7
8
6
4
3
1
2
JM63.I1

Std
Dev
Ambient
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
29
30
32
29
3
18
2
4
1
28
9
8
7
6
5
10
27
15
20
26
11
16
21
22
12
25
24
23
14
13
0.000
0.005
0.005
0.015
0.022
0.032
0.030
0.028
0.038
0.042
0.045
0.050
0.058
0.065
0.068
0.070
0.082
0.080
0.082
0.085
0.095
0.100
0.105
0.110
0.150
0.150
0.160
0.000
0.000
0.005
0.001
0.-024
0.020f
0.033
0.038
0.052
0.057
0.052
0.057
0.061
0.062
0.070
0.075
0.080
0.080
0.080
0.090
0.099
0.094
0.100
0.108
0.146
0.146
0.155
0.000
0.004
0.002
0.014
0.030
0.033
0.028
0.030
0.042
0.050
0.048
0.053
0.062
0.066
0.065
0.071
0.083
0.085
0.085
0.087
0.100
0.101
0.111
0.112
0.155
0.153
0.166
0.000
0.005
0.005
0.015
0.025
0.035
0.029
0.040
0.053
0.053
0.056
0.060
0.066
0.068
0.072
0.073
0.092
0.097
0.095
0.090
0.105
0.108
0.117
0.120
0.162
0.167
0.182
-0.003
0.003
0.007
0.013
0.026
0.028
0.030
0.030
0.043
0.048
0.050
0.054
0.067
0.068
0.070
0.072
0.082
0.088
0.086
0.084
0.100
0.099
0.110
0.115
0.158
0.155
0.170
0.002
0.007
0.008
0.017
0.027
0.035
0.037
0.037
0.045
0.052
0.052
0.056
0.070
0.075
0.075
0.075
0.087
0.081
0.090
0.080
0.100
0.102
0.107
0.115
0.157
0.165
0.175
0.001
-0.005
0.006
0.007
0.029
0.030
0.033
0.033
0.046
0.053
0.053
0.057
0.065
0.071
0.070
0.073
0.086
0.090
0.088
0.089
0.102
0.104
0.112
0.117
0.157
0.160
0.172
0.000
0.005
0.005
0.015
0.025
0.035
0.040
0.030
0.045
0.050
0.050
0.055
0.065
0.080
0.075
0.080
0.095
0.090
0.090
0.095
0.110
0.110
0.130
0.125
0.175
0.165
0.175
0.000
0.003
0.006
0.012
0.026
0.031
0.033
0.033
0.045
0.050
0.051
0.056
0.064
0.069
0.070
0.073
0.086
0.086
0.087
0.088
0.101
0.103
0.112
0.115
0.158
0.158
0.169
0.002
0.004
0.002
0.005
0.003
0.005
0.004
0.005
0.005
0.005
0.004
0.003
0.004
0.006
0.004
0.003
0.005
0.006
0.005
0.005
0.005
0.005
0.009
0.006
0.009
0.008
0.009
Supplemental Ozone Added
25
26
27
28
31
33
34
35
36
37
38
39
40
41
42
43
44
45
46
37
36
38
34
35
39
40
42
41
46
47
48
43
33
44
32
31
30
45
Average
Std Dev
0.122
0.122
0.132
0.132
0.135
0.155
0.168
0.178
0.180
0.262
0.278
0.285
0.298
0.322
0.325
0.345
0.362
0.368
0.370
0.137
0.111
0.151
0.156
0.160
0.164
0.164
0.170
0.179
0.184
0.188
0.258
0.272
0.277
0.348
0.343
0.353
0.362
0.377
0.390
0.382
0.145
0.116
0.150
0.148
0.157
0.158
0.160
0.173
0.181
0.189
0.192
0.278
0.291
0.297
0.346
0.348
0.369
0.365
0.381
0.386
0.407
0.148
0.120
0.123
0.125
0.137
0.140
0.145
0.162
0.183
0.200
0.203
0.272
0.292
0.298
0.310
0.347
0.342
0.377
0.400
0.427
0.410
0.149
0.120
0.155
0.154
0.162
0.161
0.162
0.177
0.184
0.193
0.197
0.289
0.290
0.295
0.359
0.358
0.378
0.373
0.387
0.392
0.414
0.151
0.122
0.142
0.140
0.145
0.155
0.155
0.161
0.182
0.195
0.197
0.285
0.307
0.310
0.327
0.355
0.355
0.382
0.395
0.405
0.420
0.151
0.121
0.172
0.173
0.179
0.185
0.182
0.188
0.190
0.197
0.203
0.280
0.289
0.293
0.392
0.390
0.401
0.406
0.417
0.421
0.429
0.159
0.130
0.150
0.150
0.160
0.160
0.165
0.185
0.195
0.210
0.215
0.310
0.330
0.340t
0.365
0.365
0.390
0.390
0.405
0.405
0.445
0.160
0.129
0.145
0.146
0.154
0.157
0.159
0.171
0.183
0.193
0.197
0.279
0.294
0.299
0.343
0.353
0.364
0.375
0.391
0.399
0.410
0.150
0.121
0.017
0.017
0.015
0.016
0.014
0.012
0.008
0.010
0.011
0.016
0.018
0.019
0.031
0.019
0.026
0.018
0.017
0.019
0.024
0.007

•(•Outlier, significant at a = 0.05.
                         16

-------
                0.035 r
                                  0.1
                                                0.2            0.3

                                                  Concentration, ppm
0.4
              0.5
                          FIGURE 5.  INDIVIDUAL ESTIMATES OF TOTAL STANDARD
                            DEVIATION (STD) VERSUS CONCENTRATION (Phase I)

    With the possible exception of the assumption of linearity, this line represents a perfectly valid estimate of the
 total standard deviation of the method; however, this analysis reveals no information whatsoever regarding the
 within-laboratory standard deviation (S\yi) or the between-laboratory standard deviation (S^) of the method.

      The within-laboratory standard deviation and the between-laboratory standard  deviation for ambient air
 samples cannot be directly determined. However, because the ambient air quality standard calls for a one-hour
 average of not more  than 0.08 ppm, it is important that  a valid estimation of the within-laboratory and between-
 laboratory standard deviations of the method in this range be obtained.

      In order to accomplish this objective, an appropriate transformation must be indirectly determined since
 the replication error  for the ambient air data cannot be shown to vary with concentration. An appropriate trans-
 formation for the ozone supplemented atmosphere data  was sought by a trial and error method. The type of
 transformation sought was of the type in Equation 2 below where a and b are the intercept and slope of the line
 of best fit of replication error versus concentration.
                                                             G                                         (2)

The details and techniques of this transformation are described in detail by Mandel/5) The constants k and G
are arbitrary, but it is useful if they can be selected so that the transformed data have the same range as the
original data. The technique was to determine k and G so that at 7 = 0, z = 0 and atj = 0.510 ppm, z = 0.510 ppm;
therefore, k and G are directly related to a and b. After each trial, the reproducibility was  compared graphically with the
                                                    17

-------
 line of best fit for the individual estimates, and if satisfactory agreement was not obtained the constants a and
 b in Equation 2 were changed and another trial made.  After five or six trials, adequate agreement between the
 two methods was obtained.
 TABLE VII. MEANS, SLOPES, AND STANDARD
 ERRORS OF ESTIMATE FOR LINEAR MODEL
  ANALYSIS WITH DATA TRANSFORMATION
                 (PHASE I)
Laboratory
Number
5
7
8
4
6
3
1
2
Average
Mean
0.2064
0.2138
0.2168
0.2190
0.2193
0.2203
0.2257
0.2281
0.2189
Slope
0.9513
0.9855
0.9993
1.0137
0.9797
0.9891
1.0473
1.0341
1.0000
Std. Error
of Estimate
0.0071
0.0100
0.0044
0.0050
0.0115
0.0066
0.0110
0.0076
0.0089*
*Pooled Estimate.
   TABLE VIII. SUMMARY OF VARIANCE
     COMPONENTS FOR LINEAR MODEL
       ANALYSIS WITH DATA TRANS-
          FORMATION (PHASE I)
Component
Value
Between Laboratories
Vdi)
V(0)
V(6)
a
z0 (point of concurrence)
0.0867
44.57 X 10-'
12.09 X 10-"
10.57 X 10-'
-0.0274
Within Laboratories
V(\)
V(e)
V(n) = V(\) + K(e)/r)
0.04980
0.07875
0.12855
     Table VII shows the means, slopes and standard errors of
estimate for each laboratory for the linear model analysis.  Table
VIII shows the summary of results for variance components and
derived quantities for the linear model analysis. In all these tables,
the data were transformed in accordance with Equation 3 below.

        z = 0.31688 [£n(0.00255 + 0.02y) -A^O.00255]       (3)
                                                 or
                                                    z = 0.31688 &*.(! + 7.843 y)                         (3a)
                                                 The relationship of the standard deviation in the.y scale to
                                            the standard deviation in the z scale is shown in Equation 4 below.
                                                       = az (a + by)lkb
                                                 or
         ay = az (0.4024 + 3.1558 y),

where y is in ppm.
                                                           (4)
(4a)
                                                 The concept of a test result® is incorporated, and a test
                                            result is defined as a single observation of an hourly average concen-
                                            tration. Following this concept, the four sources of variability as
                                            calculated by the linear model analysis for several values of concen-
                                            tration were tabulated. Table VIII shows these sources of variability
                                            and their relative importance as determined by the linear model
                                            analysis.  The total variance in the z scale is shown along with the
                                            variance for each component and its respective percent of the total.
                                            The replication variance V(e) is  derived from the transformation
                                            parameters in accordance with Equation  5 below.
                                                                                                      (5)
Comparison of V(e) and F(X), the latter reflecting interferences in the method, reveals that they are approximately
the same magnitude.  Therefore, replication, even if possible, would not significantly improve the within-laboratory
precision. The means and slopes from Table VII were shown to be correlated and therefore V(S) appears in Table
IX. This term, indicating the variability of the slopes of the laboratory lines, is negligible throughout the table. At
concentrations below 0.15 ppm, the total variance is composed primarily of within-laboratory variation.  The be-
tween-laboratory variation is relatively small in this range. At higher concentrations, the between-laboratory vari-
ability becomes more predominant, and at concentrations above 0.3 ppm, the between-laboratory variation accounts
for more than one-half of the total variation.

      For thoroughness, the analysis of variance table is shown in Table X, and  from this, it can be shown that the
laboratory lines have a tendency to intersect or concur at a particular point near the origin but, at that point, con-
siderable scatter between the lines still exists. While this is of no direct importance to this study,  it would have
significance in the selection of standard samples for quality control in the use of the method.
                                                   18

-------
                TABLE IX SOURCES OF VARIABILITY AND THEIR RELATIVE IMPORTANCE FOR
                   THE LINEAR MODEL ANALYSIS WITH DATA TRANSFORMATION (PHASE I)
y

0
0.051
0.102
0.153
0.204
0.255
0.306
0.357
0.408
0.459
0.510
Z

0
0.107
0.186
0.250
0.303
0.348
0.388
0.423
0.455
0.483
0.510
Within-Laboratory
V(e)
40.2 X 10~6
40.2 X 10~6
40.2 X 10~6
40.2 X 10- 6
40.2 X 10' 6
40.2 X 10~6
40.2 X It)-6
40.2 X 10~6
40.2 X 10~6
40. 2 X 10~6
40.2 X 10~6
%
51.9
49.2
40.5
32.9
27.3
23.3
20.3
18.0
16.2
14.7
13,5
V(\)
25.4 x 10~6
25.4 X 10~6
25.4 X 10" 6
25.4 X 10~6
25.4 x 10~6
25.4 X 10~6
25.4 x 10~6
25. 4 x 10~6
25. 4 x 10~6
25.4 x 10~6
25.4 X 10~6

%
32.8
31.1
25.6
20.8
17.3
14.7
12.8
11.4
10.2
9.3
8.6
Between-Laboratory
(1+07)' K(M)
0.5 X 10~6
13. 2 X 10~6 '
33.3 X 10~6
56.1 X 10~6
79.6 X 10~6
103. OX 10~6
125.9X 10~6
148.2 x 10~6
169.8 X 10~6
190.7 x 10~6
210.9 X I0~6
%
0.7
16.1
33.6
46.0
54.2
59.7
63.5
66.3
68.3
69.9
71.]
72F(6)
11.3 X 10~6
3.0 X 10~6
0.2 X 10~6
0.2 X 10~6
1.7 X10 ~6
4.0 X 10~6
6.8 X 10~6
9.9 X 10"6
13.2X 10~6
16.7 X 10~6
20.1 X 10~6

°
14.6
3,6
0.3
0.2
1.1
2.3
3.4
4.4
5.3
6.1
6.8
Vf7\
' \L>)
77.5 X 10~6
81 7 X 10~6
99.2 X 10~5
121.9 X 10~6
146.9 X 10~6
172.5 X 10~6
198.2 X 10~6
223.6 X 10~6
248.6 X 10~6
272.9 X 10~6
296.6 X 10^6
       TABLE X. ANALYSIS OF VARIANCE TABLE FOR LINEAR MODEL
            ANALYSIS WITH DATA TRANSFORMATION (PHASE I)
Source
Laboratories
Concentrations
Laboratory X Concentration
Linear
Concurrence
Non-Concurrence
Deviation from Linearity
SS
14, 699.44 X 10~6
6,276, 065.62 X 10~6
25,449. SOX 10~6
5,256.47 X 10~6
4,139.75 X 10~6
1.116.72X 10~6
20,192.48 X 10~6
DF
7
45
315
7
1
6
308
MS
2,099. 92 X 10~6
139,468.12 x 10~6
80.79 X 10~6
750.92 x 10~6
4,139.75 X 10~6
186. 12X 10~6
65.58X 10-6
F (Con./Non-Con.) = 22.243 Probability = 100
Point of Concurrence = —0.274
F (Non-Conc./Resid.) 2,839 Probability = 99
Concurrence is not within experimental error.
    TABLE XI. LISTING OF TOTAL, BETWEEN-LABORATORY, AND WITHIN-
       LABORATORY STANDARD DEVIATION DERIVED FROM LINEAR
                       MODEL ANALYSIS* (PHASE I)
Z
0
0.107
0.186
0.250
0.303
0.348
0.388
0.423
0.455
0.483
0.510
f(n)
65.6 X 10~6
65.6 X 10~6
65. 6 X 10~6
65.6 X 10~6
65.6 X 10~6
65.6 X 10~6
65. 6 X 10~6
65.6 X 10~6
65.6 X 10~6
65. 6 X 10~5
65.6 X 10~6
V(Z)
77.5 X 10^6
81.7 X 10~6
99. 2 X 10~6
121. 9 X 10~6
146.9 X 10~6
172.5 X 10~6
198.3 X 10~6
223.7 X 10~6
248.6 X 10~6
272.9 X 10~6
296.6 X 10~6
Y
0
0.051
0.102
0.153
0.204
0.255
0.306
0.357
0.408
0.459
0.510
STM
0.0035
0.0051
0.0072
0.0098
0.0127
0.0159
0.0193
0.0229
0.0266
0.0306
0.0346
SL
0.0012
0.0022
0.0041
0.0066
0.0094
0.0125
0.0157
0.0193
0.0228
0.0266
0.0305
SWL
0.0033
0.0046
0.0059
0.0072
0.0085
0.0098
0.0111
0.0124
0.0137
0.0150
0.0163
*STM-Total Standard Deviation, Si -Between-Laboratory Standard Deviation,
and SWL -Within-Laboratory Standard Deviation.
           Table XI presents a list-
ing versus concentration level of the
total standard deviation (Sf^f), the
between-laboratories standard devi-
ation (S^), and the within-laboratory
standard deviation (S^j^). The con-
centration levels in  the z transforma-
tion are given in ppm.  The total
standard deviation was calculated,
using Equation 4a as follows:

STM = -JV(z) (0.4024 + 3. 1 558 v) (6)

The within-laboratory standard devi-
ation was calculated by the formula:
SWL =
              (0.4024 + 3. 1 558 y) (1)
                                                                        Finally, the between-laboratories
                                                                        standard deviation was calculated
                                                                        as follows:
                                                                                            l /2
                                                                                                         (8)
                                                                             In Figure 6, the calculated
                                                                        points from the linear analysis shown
                                                                        in Table XI for the total standard
                                                                        deviation (Sj-j^) are shown.  In order
                                                                        to obtain a good fit of the regression
                                                                        line to these points, it was necessary
                                                                        to divide the line into two segments.
                                                                        The first segment is from 0 to 0.153
                                                                        ppm and the other from 0.153 to
                                                                        0.510 ppm. Also, on Figure 6, the
STD regression line from Figure 5 is superimposed for comparison purposes. It is readily apparent that the results
of the linear analysis agrees very well with the individual estimates of the total standard deviation over the entire
range from 0 to 0.510 ppm.
                                                   19

-------
             .030
          c
          o
          '•p
          CD
          >  .020
          IV
          Q
          "2
          co
          -a
          c
          CD
          •(-<
          V)
             .010
                            Total Standard Deviation, Linear Model (STM)
                            Regression Line for S,...
                                               TM
                            Regression Line for Total Standard Deviation
                              from Plotted Data (SJD)
                                                                           = 0.0018 + 0.0542x
                               STM =0.003 +0.041
.10
                                           .20
                                        Concentration
                                                         .30
                                      .40
.50
               FIGURE 6. COMPARISON OF TOTAL STANDARD DEVIATION ESTIMATES VERSUS
                     CONCENTRATION LEVEL FROM LINEAR MODEL AND DATA POINTS
      Figure 7 presents the estimates of within-laboratory (Styj-J) and between-laboratories
The regression line for Sj^ was divided into two segments for a better fit.
                                                                                      standard deviation.
      The equations for the estimate of the standard deviations versus concentration level (x) in ppm follow:

                              STM = 0.0033 + 0.0412*,    0  < x <  0.153                              (9)

                              STM =-0.0024 + 0.07 llx,   0.153  
-------
I .030
Q.
C
o
   .020
C
to
*->
CO


   .010
                Between-Laboratories Standard Deviation, Linear Model (SL)

                Regression Line for S.


                Regression Line for Within-Laboratory Standard Deviation,
                  Linear Model (Sw  )
                                                       SWL = 0.0033 + 0.0255X
                      SL = 0.0008 + 0.0355X
       0
.50
                      .10          .20          .30           .40

                               Concentration Level, ^ (ppm)

FIGURE 7. BETWEEN-LABORATORIES AND WITHIN-LABORATORY STANDARD DEVIATION ESTIMATES
               VERSUS CONCENTRATION LEVEL (DERIVED FROM LINEAR MODEL)


Applying this technique to the estimated standard deviations, the LDL for the three cases can be estimated.



1.    The estimated LDL, taking into account the total standard deviation (using equation 9 with x = 0) is:


                     LDL(T) = (0.0033)(1.96) = 0.0065 ppm.


2.    The estimated LDL for between-laboratories (using equation 10) is:


                     LDL(L) = (0.0008)(1.96) = 0.0018 ppm.


3.    The estimated LDL for within-laboratory (equation  11) is:


                     LDL(WL) = (0.0033)(1.96) = 0.0062 ppm.
                                        21

-------
                           IV. STATISTICAL DESIGN AND ANALYSIS
                                           (PHASE 11, BIAS)
 A.   Purpose and Scope of the Experiment

      The purpose of the test was to establish the accuracy of the method using sampling periods of one hour, to
 agree with the requirements of the ambient air quality standard thereby providing bias determination and quanti-
 fication.

      The scope of the experiment was limited by the time constraints inherent in taking one-hour samples. In
 order to solicit an adequate number of collaborators, the scope of the work involved was limited to an amount
 that could be accomplished in a week.

      Each laboratory was required to analyze five separate concentrations.

      Because individual ozone generating equipment was supplied to each laboratory, the concentrations were
 not identical for all laboratories.  However, the deviation was less than 2 percent of the mean value for all labor-
 atories for all five concentration levels at the calibration points.  The expected mean concentrations were nominally
 0.054, 0.084, 0.131, 0.273, and 0.481 parts per million (ppm). These values are taken as the true concentration
 levels at all laboratories.

      The ozone generating equipment was calibrated by the National Bureau of Standards, and supplied to the
 collaborators. The calibration was checked by the National Bureau of Standards after the test was  completed.

      Table II shows the change in percent of the generator output at one setting (approximately at concentration
 level 5) from the first calibration. Generators 11, 13, and 19 showed a significant decline in  output. A thorough
 check of the data points reported by these laboratories over the four days on which data was taken (see Table XII)
 did not show a pattern of decreasing amounts for each concentration level over the four days.

      Referring to Table XVII, one  sees that laboratories 13 and 19 display a higher average over all concentration
 levels than the overall average across all laboratories for all concentration  levels. Thus it would appear that the
 generator output  decline occurred after the measurements. Laboratory 11 had the lowest average of all the labor-
 atories at four concentration levels. At concentration level 5, laboratory  11 had an average 44 percent below the
 overall average, and 29 percent below the next lowest average (laboratory 15).  Thus, it seems improbable that
 the generator output decline contributed significantly to this effect.  Therefore, no adjustment was made on the
 reported data for  the differences in generator output before and  after the test.

 B.   Design of the Experiment

     The experiment was designed  within the framework of the practical constraints of holding the collaborators'
expected workload to a week. Within this constraint, the experiment was structured so that analysis of variance
(ANOVA) techniques could be employed to obtain efficient estimates of the lower detectable limit.

     The ANOVA techniques used in this study were:

     1.   A two-factor ANOVA with data in the original scale  for each  concentration level separately on data
          taken  on the same day for concentration levels 1 and 5, and for the data taken across the three days
          for levels 1 through 5.

     2.   A linear model analysis^10) with data in the transformed scale.
                                                   22

-------
       TABLE XII. OBSERVED VALUES BY COLLABORATING
                   LABORATORIES (PHASE II)
Laboratory
Number
10
11
13
14
15
16
17
18
19
20
Ozone Concentration Level Number (ppm)

R
0.038
0.037
0.037
0.023
0.020
0.023
0.050
0.055
0.055
0.043
0.043
0.043
0.032
0.032
0.033
0.040
0.042
0.044
0.037
0.037
0.037
0.033
0.032
0.032
0.028
0.030
0.030
0.029
0.027
0.029
1
D
0.035
0.038
0.037
0.025
0.023
0.023
0.023
0.015
0.030
0.046
0.044
0.046
0.025
0.032
0.030
0.040
0.041
0.041
0.035
0.030
0.035
0.027
0.035
0.035
0.029
0.035
0.037
0.028
0.030
0.029
2
D
0.059
0.065
0.062
0.035
0.033
0.035
0.040
0.040
0.055
0.068
0.067
0.072
0.044
0.065
0.050
0.065
0.063
0.066
0.062
0.052
0.056
0.050
0.058
0.057
0.052
0.060
0.062
0.050
0.052
0.059
3
D
0.105
0.106
0.104
0.057
0.055
0.057
0.095
0.090
0.105
0.107
0.107
0.109
0-070
0.087
0.085
0.109
0.106
0.106
0.100
0.090
0.095
0.087
0.100
0.097
0.092
0.100
0.102
0.087
0.091
0.097
4
D
0.230
0.237
0.242
0.123
0.120
0.133
0.230
0.230
0.270
0.237
0.235
0.228
0.165
0.203
0.201
0.237
0.236
0.233
0.275
0.210
0.277
0.190
0.220
0.221
0.210
0.226
0.232
0.202
0.208
0.218
5
R
0.442
0.440
0.440
0.210
0.210
0.210
0.510
0.505
0.500
0.412
0.386
0.388
0.345
0.323
0.352
0.454
0.454
0.452
0.385
0.383
0.382
0.383
0.378
0.378
0.372
0.368
0.372
0.376
0.368
0.370
D
0.423
0.430
0.435
0.217
0.213
0.220
0.435
0.440
0.465
0.403
0.404
0.398
0.340
0.357
0.365
0.458
0.448
0.446
0.410
0.370
0.377
0.345
0.399
0.380
0.405
0.420
0.427
0.377
0.391
0.393
R -Replication (3 Readings at Concentration Levels 1 and 5 in One Day).
.D-Day (One Reading at Each Concentration Level Per Day for Three
Days).
      The experimental design is shown in
 Figure 8.  For the first ANOVA technique,
 each factor was represented in the following
 quantities:

      1.    Replicate data-laboratories
           (ten), replicates (three), with
           the replicates nested in the
           laboratories.

      2.    Data taken across days—labor-
           atories (ten), days (three), with
           the days nested in the labor-
           atories.

      The data from the ten laboratories were
 analyzed with the second ANOVA technique
 with three days and five concentration levels.
 The linear model^5^ views each of the five
 concentrations on each of the three days as
 individual materials.  Thus, each of the labor-
 atories has measured each of the 15 materials
 one time.

      Referring to Figure 8, each laboratory
 made a total of 21  one-hour average deter-
 minations.  For each of the first three days
 there were five separate concentration levels
 to be measured (levels 1 through 5), yielding
 a total of 15 determinations for the three
 days. On the fourth day, there were three
 sets of the first and fifth concentration
 levels to be measured, making a total of
 6 determinations.

     The order of the five concentration
 levels to be measured each  day was ran-
 domized within and over all laboratories.
 The replication measurements of the first
 and fifth concentration levels on the fourth
 day were made alternately  so that the error
 associated with setting the  ozone generation
slide scale would be incorporated in the
replication error.
      A detailed procedure was sent to each collaborator outlining the steps to be taken to carry out the necessary
details of the test.

C.    Preliminary Data Analysis

      Only one of the eleven collaborators (laboratory number 12) was unable to conduct the series of tests because
of reported ozone generating equipment problems. After several telephone conferences with the collaborator, it
was decided to abort the test and return the equipment to NBS. The equipment was checked at NBS and found to
be operating normally.
                                                    23

-------
                                                                      (same as LI )
                   D1
                                       D2
                                                          D3
                                                                            D4
            C5 C1  C2  C4 C3    C1  C4  C5 C3  C2    C3  C2 C4  C1 C5
                  L  — Laboratories
                  D  -Day
                  C  — Concentrations
                  R  - Replicates
C1
               C5
                                                                  R1   R2  R3
                                                                                R1  R2  R3
                           FIGURE 8.  DESIGN OF OZONE METHOD TEST (Phase II)

     The data provided by the other ten collaborators were checked to assure that the proper steps, as outlined in
the method of converting the monitor graph plots to hourly averages of ozone readings in parts per million were
accomplished. No obvious deviations from the prescribed method were noted.  The calculations which were required
by the method were checked for each collaborator,  and only one inconsistency was found.  One collaborator had
failed to correct his monitor readings by  the monitor calibration curves.  These readings were corrected.

      Several minor deviations from the test procedure were reported. These involved not following the prescribed
concentration level order for the five daily measurements. One other deviation was the absence of one determination
at a concentration level by a laboratory because  of an inadvertent wrong setting on the generator slide scale. In this
case, the average of the other two determinations was assumed, and the degrees of freedom was not adjusted since
it would have only a minor effect on the overall  analysis.

      Overall, the response from the collaborators was very good, and from the review of the data submitted, it
was apparent that each had  worked conscientiously and  diligently to conduct the test in accordance with the
procedure and method.

D.   Presentation of Data

      The observed values recorded during the tests by each of the laboratories are listed in Table XII.  The
expected concentration values based on the calibrated ozone generating equipment are given in Table I, along
with the averages of the expected concentration levels at each of the five points across all laboratories. The ob-
served values were adjusted  to compensate for the difference in the expected concentrations for the ten labor-
atories, and are displayed in Table XIII.  The formula used to calculate the adjusted values (A) was:

                                           A = O+(V-E),

where 0 is the observed value (from Table XII),  Vis the average value for a concentration level over all labor-
atories (from Table I), and E is the expected value for a concentration level for a laboratory (from Table I).
In this manner, the data from each laboratory were  adjusted to compensate for the minor variations in the
standards across the laboratories, making it feasible  to use the same expected value at a concentration level
for all laboratories.
                                                    24

-------
            TABLE XIII. OBSERVED VALUES (ADJUSTED)
                            (PHASE II)
Laboratory
Number
10
11
13
14
15
16
17
18
19
20
Ozone Concentration Level Number (ppm)

R
0.039
0.038
0.038
0.023
0.020
0.023
0.050
0.055
0.055
0.043
0.043
0.043
0.032
0.033
0.033
0.040
0.042
0.044
0.037
0.037
0.037
0.033
0.032
0.032
0.028
0.030
0.030
0.030
0.028
0.030
1
D
0.036
0.039
0.038
0.025
0.023
0.023
0.023
0.015
0.030
0.046
0.044
0.046
0.025
0.032
0.030
0.040
0.041
0.041
0.035
0.030
0.035
0.027
0.035
0.035
0.029
0.035
0.037
0.029
0.031
0.030
2
D
0.061
0.067
0.064
0.035
0.033
0.035
0.040
0.040
0.055
0.068
0.067
0.072
0.044
0.065
0.050
0.065
0.063
0.066
0.062
0.052
0.056
0.050
0.058
0.057
0.052
0.060
0.062
0.052
0.054
0.061
3
D
0.107
0.106
0.106
0.057
0.055
0.057
0.095
0.090
0.105
0.107
0.107
0.109
0.069
0.086
0.084
0.108
0.105
0.105
0.100
0.090
0.095
0.086
0.099
0.096
0.091
0.099
0.101
0.089
0.093
0.099
4
D
0.234
0.241
0.246
0.122
0.119
0.132
0.229
0.229
0.269
0.236
0.234
0.227
0.164
0.202
0.200
0.236
0.235
0.232
0.274
0.209
0.276
0.189
0.219
0.210
0.209
0.225
0.231
0.206
0.212
0.222
5
R D
0.448
0.446
0.446
0.209
0.209
0.209
0.510
0.505
0.500
0.411
0.385
0.387
0.342
0.320
0.349
0.451
0.451
0.452
0.384
0.382
0.381
0.380
0.375
0.375
0.369
0.365
0.369
0.382
0.374
0.376
0.429
0.436
0.441
0.216
0.212
0.219
0.434
0.439
0.464
0.412
0.386
0.388
0.337
0.354
0.362
0.455
0.445
0.443
0.409
0.369
0.376
0.342
0.396
0.377
0.402
0.417
0.424
0.383
0.397
0.399
R -Replication (3 Readings at Concentration Levels 1 and 5 in One Day).
D-Day (One Reading at Each Concentration Level Per Day for Three
Days).
      Since the variance across the five con-
centration levels was not homogeneous, a
transformation was necessary to establish
homogeneity so the concentrations could
be analyzed together. The transformed
data appear in Table XIV.

E.    Tests for Outlying Observations

      A preliminary review of the data
revealed that the observations reported
by laboratory 11 were noticeably lower
over concentration levels 2 through 5 than
the observations reported by the other nine
laboratories.  Telephone discussions with the
collaborator did not reveal any departure
from the procedure or test method.

      An analysis of variance was made on
each concentration separately over all labor-
atories on the adjusted "day" data (Table
XIII).  From this analysis the variation with-
in and between laboratories was obtained for
each concentration level.  Table XV is  a
summary of the analysis giving the cell
average (X), standard deviation (S) for each
laboratory at each concentration level, as
well as the within laboratory averages (X)
and standard deviations (S^^) for each
concentration level.

      Using the "Recommended Practice
for Dealing With Outlying Observations"
described in reference 9, a T' statistic
was calculated for each cell average for
the  "day" data as follows:
                                                                             T'=(X-X)lSx,
                                                                  where
      This statistic (71') was then compared to a t value at the 99 percentile (a = 0.01) which was taken from
Table 5 in reference 9, with« = 10.

                                                      = 3.37.
An a value of 0.01 was chosen to limit the probability of discarding valid data. All cell averages where T' is
equal to or greater than t are considered to be potential outliers.

      Table XVI presents the T' statistic for each laboratory at each level for the five concentration levels. Out-
liers  are marked with an asterisk. Laboratory 11  displayed outliers at all five concentration levels. Only  three
laboratories had no outliers.
                                                    25

-------
          TABLE XIV. TRANSFORMED DATA (PHASE II)
     An analysis of variance was then made
on each concentration level separately over
the remaining nine laboratories (excluding
the data from laboratory 11).  The labor-
atory standard deviation and cell averages
remain the same, only the within laboratory
standard deviation and concentration level
average changes. These new values, with
laboratory 11 data omitted are shown in"
the next to last row of Table XV. New
values for T' were calculated and are
shown in Table XVI. With laboratory 11
omitted, the critical value  of ?(0.01)(18d/)
with n = 9 is 3.35. T' values which exceed
t are marked with an asterisk.  Of the re-
maining nine laboratories, six had one or
more (three maximum) cells identified as
potential outliers. It was obviously unde-
sirable to drop all six laboratories from the
study.  Since no physical reason could be
found for the outliers, it seemed plausible
that the spread in the data may be a natural
phenomenon associated with the test method,
or nonconstant nonrandom differences be-
tween laboratories.
                                                                      The linear model as described by
                                                                         0) was utilized to investigate this
                                                                 possibility, since the linear model allows
                                                                 for nonconstant, nonrandom differences
                                                                 between laboratories.  As a preliminary
                                                                 step<6), the data in Tables XVII and XVIII
                                                                 were tabulated. From Table XVIII, one
                                                                 sees that the ratio of the cell standard
                                                                 deviation to average standard deviation
                                                                 over all laboratories for a given laboratory
                                                                 does not fluctuate very much over all five
                                                                 concentration levels. However, the average
                                                                 of the ratios over  the five concentration
                                                                 levels for each laboratory indicates there
                                                                 is a recognizable difference in the spread
of the individual concentration measurement results between laboratories. The data in Table XVIII seem to support
the probability of nonconstant, nonrandom differences between laboratories which, according to Mandel^^,
is not an uncommon occurrence.

     As a further check, a scale transformation of the data (see Table XIV) was made (to achieve homogeneity
of variances) in accordance with the methods outlined by Mandel, and the linear model was exercised.  The means,
slopes, and standard error of the estimate are presented in Table XIX. None of the remaining 9 laboratories show
values of both the mean and the slope that are  much smaller than all the remaining laboratories. This fact indicates
that no significant outliers exist in the data.

     Another argument for not censoring all data points marked as potential outliers in Table XVI (with laboratory
11 censored) can be made.  If laboratories 13 and 15 (each with 3 points marked as outliers) were censored from the
Laboratory
Number
10
11
13
14
15
16
17
18
19
20
Ozone Concentration Level Number (ppm)
1
R
0.093
0.091
0.091

0.115
0.124
0.124
0.101
0.101
0.101
0.079
0.081
0.081
0.095
0.099
0.103
0.089
0.089
0.089
0.081
0.079
0.079
0.070
0.074
0.074
0.074
0.070
0.074
D
0.087
0.093
0.091

0.059
0.040
0.074
0.107
0.103
0.107
0.063
0.078
0.074
0.095
0.097
0.097
0.085
0.074
0.085
0.068
0.085
0.085
0.072
0.085
0.089
0.072
0.077
0.074
2
D
0.135
0.145
0.140

0.095
0.095
0.124
0.147
0.145
0.153
0.103
0.142
0.115
0.142
0.138
0.143
0.136
0.118
0.126
0.115
0.129
0.128
0.118
0.133
0.136
0.118
0.122
0.135
3
D
0.205
0.204
0.204

0.188
0.181
0.202
0.205
0.205
0.208
0.148
0.175
0.172
0.206
0.202
0.202
0.195
0.181
0.188
0.175
0.194
0.190
0.183
0.194
0.197
0.180
0.185
0.194
4
D
0.338
0.344
0.348

0.334
0.334
0.366
0.340
0.338
0.332
0.273
0.310
0.308
0.340
0.339
0.336
0.369
0.316
0.371
0.298
0.325
0.317
0.316
0.331
0.336
0.314
0.319
0.328
5
R
0.475
0.474
0.474

0.505
0.502
0.500
0.456
0.441
0.442
0.416
0.402
0.420
0.477
0.477
0.477
0.441
0.440
0.439
0.438
0.436
0.436
0.432
0.430
0.432
0.440
0.435
0.436
D
0.465
0.469
0.471

0.468
0.470
0.483
0.456
0.442
0.443
0.412
0.423
0.428
0.478
0.473
0.472
0.455
0.423
0.436
0.416
0.447
0.437
0.451
0.459
0.463
0.440
0.448
0.449
R -Replication (3 Readings at Concentration Levels 1 and 5 in One Day).
D-Day (One Reading at Each Concentration Level Per Day for Three
Days).
                                                    26

-------
  TABLE XV. WITHIN LABORATORY AND CELL MEANS AND STANDARD
            DEVIATION FOR FIVE CONCENTRATION LEVELS
                    (Adjusted Day and Replication Data)
                               Phase II
Laboratory
Number
10
11
13
14
15
16
17
18
19
20
Within labs
(including
11)
Within labs
(excluding
11)
Within labs
(excluding
11,13,15)
Param-
eter
X
S
X
S
X
S
7
S
X
S
X
S
X
S
X
S
X
S
X
S
I
SWL
T
SWL
X
SWL
Concentration Level (Day Data)
(ppm)
1
0.0377
0.0015
0.0237
0.0012
0.0227
0.0075
0.0453
0.0012
0.0290
0.0036
0.0407
0.0006
0.0333
0.0029
0.0323
0.0046
0.0337
0.0042
0.0300
0.0010
0.0328
0.0035
0.0339
0.0037
0.0361
0.0027
2
0.0640
0.0030
0.0343
0.0012
0.0450
0.0087
0.0690
0.0027
0.0530
0.0108
0.0647
0.0015
0.0567
olooso
0.0550
0.0044
0.0580
0.0053
0.0557
0.0047
0.0555
0.0055
0.0579
0.0058
0.0604
0.0040
3
0.1063
0.0006
0.0563
0.0012
0.0967
0.0076
0.1077
0.0012
0.0797
0.0093
0.1064
0.0017
0.0950
0.0050
0.0937
0.0068
0.0970
0.0053
0.0937
0.0050
0.0932
0.0275
0.0973
0.0055
0.0999
0.0043
4
0.2403
0.0060
0.1243
0.0068
0.2423
0.0231
0.2323
0.0044
0.1887
0.0214
0.2343
0.0021
0.2530
0.0381
0.2060
0.0154
0.2217
0.0114
0.2133
0.0081
0.2156
0.0173
0.2258
0.0181
0.2287
0.0167
5
0.4353
0.0060
0.2157
0.0035
0.4457
0.0161
0.3953
0.0145
0.3510
0.0128
0.4510
0.0069
0.3847
0.0214
0.3717
0.0274
0.4143
0.0112
0.3930
0.0087
0.3858
0.0146
0.4047
0.0154
0.4065
0.0156
Concentration
Level Replica-
tion Data
1
0.0383
0.0006
0.0220
0.0017
0.0533
0.0029
0.0430
0.0000
0.0327
0.0006
0.0420
0.0020
0.0370
0.0000
0.0323
0.0006
0.0293
0.0012
0.0293
0.0012
0.0359
0.0014
0.0375
0.0013
0.0359
0.0010
2
0.4467
0.0012
0.2090
0.0000
0.5050
0.0050
0.3943
0.0145
0.3370
0.0151
0.4513
0.0006
0.3823
0.0015
0.3767
0.0029
0.3677
0.0023
0.3773
0.0042
0.3847
0.0071
0.4043
0.0074
0.3995
0.0059
data, the effects on the calculated
accuracy (bias) of the method would
be minimal, since from Table XVII
we see that laboratory 13 has one of
the higher averages over all concentration
levels, and laboratory 15 has one of the
lower averages.

      The bias data for both cases
(laboratory 11 censored, and labor-
atories 11, 13, and 15  censored) were
calculated, and are shown in Table XX.

      The T' statistic  was also recal-
culated with laboratories  11, 13, and
15 censored.  With seven laboratories,
      '(0.01)(14or/) = 3.33.

      From Table XVI, there are still
two potential outliers  in laboratory
14, and one each in laboratories 16, 17,
and 20.

      Since it is undesirable to delete
an individual concentration level data
point for a laboratory (without deleting
the entire laboratory data), and  in view
of the preceding arguments, it is felt
that the inclusion of all the data from
all laboratories except laboratory 11 in
the analysis provided a realistic estimate
of the accuracy (bias) of the method.

F.   Accuracy  (Bias) Determination
      In order to  determine the accuracy of
the method, the standard error of the treat-
ment means at each concentration level were
used with the t distribution to calculate a 95%
confidence interval on the mean at each con-
centration level, and also on the bias at each
concentration level.

L2=X+2.101 S%
     The confidence limits were established on the concentration level means by the use of the relation:
where  I, and L-y are the lower and upper limits, 2.101 is the value of the t distribution with 18 degrees of freedom
that encompasses 95% of the area, Sj[ is the standard error of the concentration level average (Sx = S/^/3), and X
is the average over all laboratories for a concentration level. The parameters X and S are taken from Table XV.  The
confidence intervals on the means are shown in Figure 9.
                                                    27

-------
     TABLE XVI. T' STATISTIC FOR IDENTIFICATION OF POTENTIAL OUTLIERS
Concen-
Level
Laboratories
10
11
13
14
15
16
All Laboratories
1
2
3
4
5

1
2
3
4
5

1
2
3
4
5
2.39
2.65
4.34*
2.43
5.87*
4.53*
6.59*
12.17*
9.15*
20.16*
5.03*
3.29
1.14
2.67
7.10*
6.18*
4.21*
4.78*
1.67
1.13
1.89
0.79
4.47*
2.70
4.12*
3.87*
2.86
4.22*
1.87
7.73*
17

0.25
0.36
0.59
3.74*
0.13
Laboratory 11 Censored
1.80
1.82
2.83
1.40
3.45*





9.07*
3.84*
0.20
1.59
4.62*
5.42*
3.31
3.25
0.63
1.39
2.29
1.46
5.52*
3.56*
6.05*
3.21
2.02
2.73
0.82
5.22*
0.24
0.37
0.72
2.61
2.25
18
19

0.25
0.32
0.15
0.97
1.67
0.24
0.77
1.25
0.64
3.38

0.72
0.86
1.14
1.90
3.72*
0.87
0.03
0.93
0.39
1.09
20

0.41
0.04
0.15
0.73
0.86

1.82
0.66
1.14
1.19
1.32
Laboratories 11, 13, and 15 Censored
0.97
1.54
2.59
1.21
3.20










5.84*
3.69*
3.13
1.46
1.16





2.88
1.83
2.62
2.26
4.62*
1.79
2.05
1.97
2.52
2.26
2.42
2.34
2.51
2.36
3.61*
1.57
1.04
1.17
0.73
0.82
3.90*
2.05
2.51
1.60
1.40
*Potential Outliers
TABLE XVII. ADJUSTED DATA, LABORATORY AND CONCENTRATION LEVEL AVERAGES,
              AND STANDARD DEVIATION AVERAGE (PHASE II)
Laboratory
Number
10
13
14
15
16
17
18
19
20
Average of
Averages
Average of
Standard
Deviations
Concentration Level (ppm)
1
0.0377
0.0015
0.0227
0.0075
0.0453
0.0012
0.0290
0.0036
0.0407
0.0006
0.0333
0.0029
0.0323
0.0046
0.0337
0.0042
0.0300
0.0010
0.0339
0.0030
2
0.0640
0.0030
0.0450
0.0087
0.0690
0.0026
0.0530
0.0108
0.0647
0.0015
0.0567
0.0050
0.0550
0.0044
0.0580
0.0053
0.0557
0.0047
0.0579
0.0051
3
0.1063
0.0006
0.0967
0.0076
0.1077
0.0012
0.0797
0.0093
0.1060
0.0017
0.0950
0.0050
0.0937
0.0068
0.0970
0.0053
0.0937
0.0550
0.0973
0.0047
4
0.2403
0.0060
0.2423
0.0231
0.2323
0.0047
0.1887
0.0214
0.2343
0.0021
0.2530
0.0381
0.2060
0.0154
0.2217
0.0114
0.2133
0.0081
0.2258
0.0145
5
0.4353
0.0060
0.4457
0.0161
0.3953
0.0145
0.3510
0.0128
0.4510
0.0069
0.3847
0.0214
0.3717
0.0274
0.4143
0.0112
0.3930
0.0087
0.4047
0.0139
Laboratory Average
Over All
Concentration Levels
0.1767
0.1705
0.1699
0.1403
0.1793
0.1645
0.1517
0.1649
0.1571
0.1639

                                28

-------
TABLE XVIII. RATIOS OF STANDARD DEVIATIONS TO
      AVERAGE STANDARD DEVIATIONS OF
      CONCENTRATION LEVELS (PHASE II)
Laboratory
Number
10
13
14
15
16
17
18
19
20
Concentration Level (ppm)
1
0.508
2.498
0.384
1.200
0,192
0.961
1.537
1.385
0.333
2
0.586
1.692
0.517
2.114
0.299
0.983
0.852
1.034
0.923
3
0.113
1.617
0.243
1.967
0.366
1.058
1.442
1.120
1.065
4
0.417
1.595
0.327
1.478
0.144
2.633
1.063
0.785
0.558
5
0.432
1.157
1.042
0.920
0.499
1.538
1.972
0.809
0.628
Average
of Ratios
0.411
1.712
0.503
1.536
0.300
1.435
1.373
1.027
0.588
TABLE XIX. MEANS, SLOPES AND STANDARD
    ERRORS OF ESTIMATE FOR LINEAR
    MODEL ANALYSIS. DATA IN TRANS-
       FORMED SCALE (PHASE II)
Laboratory
Code No.
15
18
20
13
19
17
14
10
16
Average
Mean
0.215
0.227
0.231
0.235
0.238
0.238
0.249
0.251
0.251
0.237
Slope
0.9334
0.9518
0.9931
1.1318
1.0078
1.0011
0.9243
1.0379
1.0190
1.000
Standard Error
of Estimate
0.0067
0.0023
0.0027
0.0072
0.0029
0.0148
0.0011
0.0033
0.0060
0.0070*
*Pooled Estimate
           TABLE XX. 95% CONFIDENCE INTERVALS ON MEANS AND BIAS (PHASE II)
Concen-
tration
Level
True
Value
00
Collab.
Mean
(X)
Collab.
Std. Err.
Mean (SJ)

1
2
3
4
5
0.054
0.084
0.131
0.273
0.481
0.0339
0.0579
0.0973
0.2258
0.4047
0.0021
0.0034
0.0032
0.0104
0.0089
95% Confidence
Interval
for Mean
Mean
Percentage
Difference
Difference
(*-M)
Bias (B)
Laboratory 11 Censored (18 dfl
0.0294 to 0.0383
0.0508 to 0.0650
0.0906 to 0.1 040
0.2039 to 0.2477
0.3860 to 0.4233
Laboratories 11,
1
2
3
4
5
0.054
0.084
0.131
0.273
0.481
0.0361
0.0604
0.0999
0.2287
0.4065
0.0016
0.0023
0.0025
0.0096
0.0090
0.0327 to 0.0395
0.0555 to 0.0654
0.0945 to 0.1052
0.2081 to 0.2493
0.3872 to 0.4257
37
31
26
17
16
-0.0201
-0.0261
-0.0337
-0.0472
-0.0763

95% CI on B

95% CI
on B as%
of True Value

-0.0246 to -0.0157
-0.0332 to -0.0191
-0.0404 to -0.0270
-0.0691 to -0.0253
-0.0950 to -0.0577
-45 to -29
-39 to -23
-31 to -21
-25 to -9
-20 to -12
13, and 15 Censored (14 df)
33
28
24
16
15
-0.0179
-0.0236
-0.0311
-0.0443
-0.0745
-0.0212 to -0.0145
-0.0285 to -0.0186
-0.0364 to -0.0258
-0.0649 to -0.0237
-0.0951 to -0.0539
-39 to -26
-33 to -22
-28 to -20
-24 to -8
-20 to -11
                                       29

-------
            .500
             .400
           Q.
             .300
          o
          ~QJ
           s
           I .200
             .100
             .050
AV
*7
                                                               f
                                                                   t
                                                                       /\
                                 /
                        /\
                      .050
                             .100
                                    .150
                                          .200
                                                 .250
                                                        .300
                                                               .350
                                                                     .400
                                                                             .450
                                                                                   .500
                                      Concentration Level 03 Expected (ppm)
   FIGURE 9. EXPECTED VERSUS AVERAGE OF OBSERVED VALUES (WITH 95% CONFIDENCE INTERVAL) FOR
                   FIVE CONCENTRATION LEVELS (PHASE II) (Laboratory Eleven Censored)

     The confidence limits on the bias^5^ were calculated in a similar manner, with:

                         B = X - n, and S(B) = Sx.

     The confidence intervals on the bias estimate (B) were then calculated by replacing X with X - p. in the
formula previously given, where n is the expected value for a concentration level taken from Table I.  The results
of these calculations are shown in Table XX. Figure 9 presents a plot of the average of the observed values, with
the 95% confidence interval versus the expected values for the five concentration levels (with laboratory 11 censored).

G.   Linear Model Analysis

     A transformation of the data is necessary to make the variances be approximately homogeneous over the five
concentration levels.  The method advocated by Mandel^5) was chosen for the transformation. The transformation
is of the form:

                         z = Kg*.(A + By) ~ G, where

                         A = 0.03047 and £= 0.0317.
                                                 30

-------
Setting G -A* A, and applying the constraints:

when                      y = 0, z = 0 and

                           .y = 0.500, z = 0.500,

then,

                           z = 0.274^(0.03047 + 0.03ITy) -A0.03047

or

                           z = 0.274 A(1.0+ lOAy),

where
                           y is the concentration level in  parts per million.

     The transformed data are shown in Table XIV, and the means, slopes and standard error of the means are
shown  in Table XIX.
                                                   31

-------
                              V.  SUMMARY AND CONCLUSIONS


A.   Phase I (Precision)

     A measure of the precision of the Method is the relative magnitude of the variance components in the
measuring process.  Equations 1, 9, 9a, 10, lOa, and 11 are expressions for the various standard deviations as
a function of concentration level, x (ppm).

                          STD  = 0.0018 + 0.0542*                                                  (1)

                          SJM  = 0.0033 + 0.0412x,     0
-------
                                     LIST OF REFERENCES


 1.    Youden, W.J., "The Collaborative Test," Journal of the AOAC, Vol. 46, No. 1, pp 55-62 (1963).

 2.    Handbook of the AOAC, Second Edition, October 1, 1966.

 3.    ASTM Manual for Conducting an Interlaboratory Study of a Test Method, ASTM STP No. 335,
      Am. Soc.  Testing & Math. (1963).

 4.    7977 Annual Book of ASTM Standards, Part 30, Recommended Practice for Developing Precision
      Data on ASTM Methods for Analysis and Testing of Industrial Chemicals, ASTM Designation:
      E180-67, pp 403-422.

 5.    Mandel, John, The Statistical Analysis of Experimental Data, John Wiley & Sons, New York,
      Chapter 13, pp 312-362 (1964).

 6.    Mandel, John, "Repeatability and Reproducibility," Materials Research and Standards, MTRSA,
      Vol. 11, No. 8, p8.

 7.    7977 Annual Book of ASTM Standards, Part 30, Recommended Practice for Dealing With Outlying
      Observations, ASTM Designation:  E178-68, p 436.

 8.    Dixon, W.J., "Analysis of Extreme Values," Ann. Math. Stat., 21, 488-506 (1950).

 9.    1968 Book of ASTM Standards, Part 30, Recommended Practice for Dealing with Outlying
      Observations, ASTM Designation EE 178-68, pp 445-447.

10.    Mandel, J., and Lashof, P.W., "The Interlaboratory Evaluation of Testing Methods,"
      ASTM Bulletin 239, pp 53-61 (1959).
                                                 33

-------
                          APPENDIX

REFERENCE METHOD FOR THE MEASUREMENT OF PHOTOCHEMICAL
       OXIDANTS CORRECTED FOR INTERFERENCES DUE TO
            NITROGEN OXIDES AND SULFUR DIOXIDE

  Reproduced from Appendix D, "National Primary and Secondary Ambient
        Air Quality Standards," Federal Register, Vol. 36, No. 84,
                  Part 11, Friday, April 30, 1971
                              35

-------
  8.  Calibration.
  8.1  Calibration  Curve.
linearity of  the  detector resp
operating flow rate and  temg
pare a calibration curve and
furnished with the
                i  the
                it the
                 Pre-
             the curve
            Introduce
zero gas and set the
a recorder reading
gas and adjust the
the proper value
on  0-58  mg./m.'
standard  at
           to indicate
       Introduce span
    control to Indicate
    recorder scale  (e.g.
:e, set  the 46 mg./m.'
     of  the  recorder
chart).  Recherf zero and span until adjust-
ments  are Jro longer necessary.  Introduce
intermeciyce calibration gases and plot  the
values obtained.  If  a smooth curve is  not
obtaln/B,  calibration   gases  may  need

    ' Calculations.
      Determine the concentrations directly^
'from  the calibration curve. No calculatlo;
are necessary.
  9.2  Carbon  monoxide  concentratipjff in
mg./m.3 are converted to p.p.m. as

       p.p.m. CO = mg. CO/m.'XOj

  10.  Bibliography.
  The Intech NDIR-CO Anawzer by Frank
McElroy.  Presented  at  t^r  nth  Methods
Conference In  Air Pollujfcn, University of
California, Berkeley,  Cajfl., April 1, 1970.
  Jacobs,  M. B.  et  afT J.AJ>.C.A. 9, JVC. 2,
110-114, August 195^
  MSA LIRA Infrfed Gas and Liquid Ana-
lyzer  Instructiojr Book, Mine Safety Appli-
ances Co., Pltjlmirgh, Pa.
  BeckmanJEstructlon 1635B, Models 216A,
315A andV^A Infrared Analyzers, Bookman
Instrument Company, Fullerton, Calif.
  Conifcuous CO Monitoring System, Model
A 5afl7lntertech Corp., Princeton, N.J.
   Ifendlx—UNOR  Infrared  'Gas  Analyzers.
  lonceverte, W, Va.
                 ADDENDA

  A. Suggested  Performance  Specific
for NDIR Carbon Monoxide Analyz —

Range (minimum)	   0-58 mC/m."
                            (/SOp.P-r,  .
Output  (minimum)	   Oj>,  100,  1,000,
                          rs.000  mv. full
                            scale.
Minimum detectable  senF  0.6  mg./m.' (0.5
  sitlvity.           S
                  X^        p.pjm.).
Lag time  (maximum)—   15 seconds.
Time  to  90  parent re-   SOseconoba.
  sponse (majmium).
Rise  tlme/90  percent   15 seconds.
  (maximum).
Fall  tUCe,  90  percent   15 seconds.
  (manmum).
Zerj^urift (maximum)—  8  percent/week,
                            not to exceed
                            1  percent/:
                            hours.
Span drift (maximum) „
     3  percent/J
       not  to ,
       1
Precision  (minimum)—   ±Q*
Operational period (min-   8
  imum);
Noise (maximum)	f±0.5 percent
Interference   equivalent   1  percent of full
  (maximum).    ^T       scale.
Operating tempeXture   5—40° C.
  range (mlnirmun).
Operating humJHlty range   10-100 percent.
  (mlnlmur
Linearity  i&aximum de-   1  percent of full
  viatic^                  *cale-
             ( Definitions o/  Performance
SgjFiflcations:
  nge—The mlntmirin and
  urement limits.
     RULES AND  REGULATIONS

Output—Hectrtcal signal whlcli^rpropor-
  ttonal  to the measurement; iKended for
  connection to readout or dipt processing
  devices. Usually expressed^* mHltrolts or
  mllllamps full scale at y^ven Impedance.
Pull Scale—The maximi^E meaBormg limit
  for a given range.
Minimum Detectable^ensltdvlty—The «nuU-
  eat amount of ^^ut  concentratk>n
  can be detecte^as the concentration ap-
  proaches
Accuracy—Tb/rdegree of agreement
  a measurjjTvalue and the true nlo*; an-
  ally  exaAssed as  ± percent of fun  mle
Lag  Tlj^—The  time Interval from * Btep
          in Input concentration *t tn0 in-
         nt Inlet to  the first corresponding
     nge in the instrument output.
     to  90 percent Response—The time ln^
  terval  from a  step change  in tbe In]
  concentration  at the  Instrument lnl<'
  a reading of 90 percent of the
  recorded concentration.
Rise Time (90 percent)—The Inyrval be-
  tween Initial response time and^lme to 90
  percent response after a ste^rlncrease in
  the inlet concentration.
Fall  Time (90 percent)—^^ie  Interval be-
  tween  initial responseyHme and  time to
  90 percent  response ^rter a  step decrease
  in the Inlet concen^Ktlon.
Zero Drift—The change in instrument out-
  put over a staiJB time period, usually M
  hours,  of un^njusted  continuous opera-
  tion, when^^he  input concentration is
            uy  expressed as  percent full
  scale.   ~

               SAMPLE INTRODUCTION
                                                                                       Span Drift—The change in Inst^^nent out-
                                                                                         put over a stated time  perlojf usually  34
                                                                                         hoars, of unadjusted oonwiuous  opera-
                                                                                         tlioi,  when the-  input oo^entratlon  IB a
                                                                                         stated upscale vmlue;  u^Bally expressed  as
                                                                                         percent full icale.  *
                                                                                       Precision—The  degrea^f agreement between
                                                                                         repeated measurements of the same con-
                                                                                         centration, eipj^&ed  as the average devia-
                                                                                         tion of the single results from the mean.
                                                                                       Operational Benod—The period of time over
                                                                                         which th^lnstrument can be expected to
                                                                                                rtinattended within specifications.
                                                                                       Noise-pontaneous deviations from a mean
                                                                                             put not caused, by Input concentration
                                                                                           -nges,
                                                                                        __-ferenoe—An undeslred positive or :
                                                                                         tive output caused by  a substance
                                                                                         than the one  being measured.   .
                                                                                       Interference  Equivalent—The  porj^n  of
                                                                                         indicated input  concentration djR to the
                                                                                         presence of an toterf erent.    >
                                                                                       Operating  Temperature Rarj^^-The  range
                                                                                         of  ambient temperatures^^er which  the
                                                                                         Instrument   will  meetyKll performance
                                                                                         specifications.
                                                                                       Operating  Humidity Bulge—The range  of
                                                                                         ambient relative hnildlty over which the
                                                                                         Instrument   wll^nmeet  all  performance
                                                                                         speclflcatlone.^r
                                                                                       Linearity—Tnj^naximum deviation between
                                                                                         an actuaVlnstrumerLt reading  and  the
                                                                                         readlng^redlcted by a straight line drawn
                                                                                         betwejn  upper  and  lower  calibration
                                                                                                      ANALYZER SYSTEM
                                               SAMPLE IN .
                                                            PRESSURE BEi-Ji
                                                              AND F1LT-
                                               SPAM
                                               AND
                                            CALIBRATION^
                                               GAS.
                                                                                                         n. ANALYZER
                                                                                                                    FLOWHETER
                                                                                               -S-J
                                                                                                  VALVE
                                                                      Figured. Carbon marexte analyzer f
                                           APPENDIX D—REFERENCE  METHOD  TOE. THE
                                             MEASUREMENT OF PHOTOCHEMICAL OXTDANTS
                                             COBKECTED  TOB INTERTEBKNCES  I>trE  TO
                                             NITROGEN OXHTES AND STTU^m I>K>KIDE

                                             I. Principle and Applicability.
                                             I.I  Ambient air  and  ethylene are de-
                                           livered simultaneously to  a  mixing zon»
                                           where the ozone in the air reacts with the
                                           ethylene to emit light which Is detected by
                                           a photomultlpUer tube. The resulting photo-
                                           current is amplified and is either read  di-
                                           rectly or displayed on a recorder.
                                             1.2  The  method Is applio»Me to the oon-
                                           tlnuotiB measurement of  oBoae 1n ambient
                                           air.
                                             2. Sanye and Sensitivity.
                                             2.1  The  range Is 8.8 ^g. O/m.' to greater
                                           than  1960  eg. O,/m."  (0.005  p.p.m. O, to.
                                           greater than 1 p.pjn. O3).
                                                                     2.2  The MnsitiTlty Is fl.fi pf. Os/m.3 (0.005
                                                                   p.pjn. Os).
                                                                     3. Interference*.
                                                                     3.1  Other oxidizing and reducing species
                                                                   normally found in ambient air do not Inter-
                                                                   fere.
                                                                     4. Precision and Accuracy.
                                                                     4.1  Tbe average deviation from the mean
                                                                   of repeated single measurements does not ex-
                                                                   ceed 5 percent of the mean of the measure-
                                                                   ments.
                                                                     4^  The method  is accurate  within it?
                                                                   percent.
                                                                     6. Apparatus.
                                                                     5.1  Detector OeM. Figure Dl is a  drawing
                                                                   of a typical detector oen showing flow paths
                                                                   at gases, the mining aone, and placement of
                                                                   •the  photomultlpller tube. Other flow paths
                                                                   in which the air and ethylene streams  meet
                                   FEDERAL  REGKX*, VOL 36.  NO. 84—NHDAV, AHW. 30,  1971
                                                              37

-------
              RULES  AND REGULATIONS
at s. point near the photomultipller tube are
also allowable.
  5.2  Air  Flowmeter. A device  capable of
controlling air flows between 0-1.5 I/mm.
  5.3  Ethylene Flowmeter. A device capable
of  controlling  ethylene  flows between 0-50
ml./mln. At any flow In this range, the device
should be  capable  of  maintaining constant
fow rate within ±3 ml./mln.
  5.4  Air   Inlet  Filter.  A  Teflon  filter
i-apable of removing all particles greater than
o microns in diameter.
  5.6  Photomultiplier Tube. A high  gain
  •w dark current  (not  more than  1X10-°
.  mpere)  photomultiplier  tube  having  Its
maxirmim  gain at  about 430 nm.  The  fol-
 owing  tubes  are  satisfactory:  RCA 4507,
;^CA  8575,  EMI 9750, EMI 9524, and EMI
S'536.
  5.6 High  Voltage Power Supply.  Capable
•>f delivering up to  2,000  volts of r*jula+«i* f*
  5.7 Direct Current Amplifier. Capable of
full scale amplification of currents from 10-10
TO 10-' ampere; an electrometer is commonly

  5.8 Recorder. Capable of full scale display
"f voltages from the DC amplifier. These volt-
ages commonly are  in the 1 millivolt to 1-volt
•ange.
  5.9  Ozone Source  and  Dilution  System.
The ozone source consists of a quartz tube
into which ozone-free air is Introduced  and
then  irradiated with a very stable low pres-
sure mercury lamp. The level of irradiation is
controlled by an adjustable aluminum sleeve
which fits around  the lamp.  Ozone concen-
trations are varied by  adjustment of  this
sleeve. At a fixed level of irradiation, ozone is
produced at a constant rate. By  carefully
controlling the flow of ah- through the quartz
sube, atmospheres are generated  which con-
tain constant  concentrations of ozone.  The
 evels of ozone in  the test atmospheres are
determined by the neutral buffered potas-
sium iodide method  (see  section  8).  This
ozone source and dilution  system is shown
schematically In Figures D2 and D3, and has
been described by Hodgeson, Stevens,  and
Martin.
  5.10  Apparatus for Calibration
  5.10.1  Absorber.  All-glass  Impingers  as
shown In Figure D4 are recommended.  The
impingers may be purchased from most ma-
jor glassware  suppliers.  Two absorbers In
series are needed to insure complete collec-
  sn of the sample.
  5.10.2  Air Pump.  Capable of drawing 1
i;ter/minute through the  absorbers.  The
pump should be equipped with a needle valve
on  the inlet side to regulate flow.
  5.10.3  Thermometer.  With  an  accuracy
of  ±2°  C.
  5.10.4  Barometer. Accurate to  the nearest
mm. Hg.
  5.10.5  Flowmeter. Calibrated metering de-
vice for measuring  flow up to 1 liter/minute
within  ±2  percent.  (For  measuring  flow
/trough impingers.)
  5.10.6  Flowmeter.  For measuring  airflow
past the lamp;  must be capable of measuring
flows  from 2 to 15  liters/minute within ±5
.- ercent.
  5.10.7  Trap. Containing glass wool to pre-
set needle valve.
  5.10.8  Volumetric  Flasks.  25,  100,  500,
,  ,000  ml.
  5.10.9  Bunt. BO  ml.
  5.10.10  Pipets. 0.5,  I,  2, 3, 4,  10,  25, and
50 ml. volumetric.
  5.10.11  Erlenmeyer  Flasks. 300 ml.
  5.10.12  Spectrophotometer.  Capable   of
measuring  absorbance at 352 nm,  Matched
1 -cm. cells Should be used.
  6. Reagents.
  6.1  Ethylene. C.  P. grade (minimum).
  6.2  Cylinder Air. Dry grade.
  <5.3  Activated Charcoal Trap. For filtering
- "Under air.
          6.4  Purified Water. Used for all reagents.
        To distilled or deionized water In an all-glass
        distillation apparatus, add a crystal of potas-
        sium permanganate and a crystal of barium
        hydroxide, and redistill.
          6.5  Absorbing Reagent.  Dissolve  13.6 g.
        potassium dihydrogen phosphate  (KH-,PO(),
        14.2 g. anhydrous disodlum hydrogen phos-
        phate  (NaJHPO,)  or 35.8 g.  dodecahydrate
        salt (Na.,HPO412H.!O), and 10.0 g.  potassium
        Iodide  (KI) in purified water and dilute to
        1,000 ml. The pH  should be  6.8±0.2. The
        solution Is stable for several weeks, if stored
        in a glass-stoppered amber bottle In a cool,
        dark place.
          6.6  Standard  Arsenious  Oxide Solution
        (0.05 N). Use primary standard grade  arse-
        nious oxide (As2O3).  Dry 1  hour  at 105° C.


       '• stoppered weighing bottle. Dissolve in 25 ml
        1 N sodium hydroxide in a flask or beaker on
        a steam bath. Add 25 ml. 1 N sulfuric acid.
        Cool, transfer quantitatively to a 1,000-ml.
        volumetric flask, and*dilute to volume. NOTE:
        Solution  must be neutral to litmus,  not
        alkaline.
              Normality
          6.7  Starch Indicator Solution  (0.2 per-
        cent). Triturate 0.4 g. soluble starch and ap-
        proximately 2 mg. mercuric iodide (preserva-
        tive) with a little water. Add the paste slowly
        to 200 ml. of boiling water. Continue boiling
        until the solution is clear, allow to cool, and
        transfer to a glass-stoppered bottle.
          6.8  Standard Iodine Solution  (0.05 If).
          6.8.1  Preparation.  Dissolve 6.0 g. potas-
        sium iodide (KI) and 3.2 g. resubllmed iodine
        (I2)  In 10 ml. purified water. When the Iodine
        dissolves, transfer the solution to a 500-ml.
        glass-stoppered  volumetric  flask. Dilute to
        mark  with purified  water  and  mix thor-
        oughly. Keep solution in a dark brown glass-
       • stoppered bottle  away  from light,  and re-
        standardize as necessary.
          6.8.2  Standardisation. Pipet accurately 20
                                               8.1.1   Into  a series of 25 ml.  volumetric
                                             flasks, plpet 0.5, 1, 2, 3, and 4 ml. of diluted
                                             standard Iodine solution  (6.9). Dilute each
                                             to the  mark  with absorbing reagent. Mix
                                             thoroughly, and  Immediately read the ab-
                                             sorbance of each at 352 nm.  against  unex-
                                             posed absorbing reagent as the reference.
                                               8.1.2  Calculate the concentration  of the
                                             solutions as total #g. Os as follows:
                                                     Total 0g. 0,= (N) (96) (V,)
                                               N=Normallty I2 (see 6.8.2), meq./ml.
                                               V,=Volume of diluted standard  Ia added,
                                                     ml. (0.6,1,2,3,4).
                                             Plot  absorbance versus total ^g. O3.
                                               8.2 Instrument Calibration.
                                              . 8.2.1   Generation of Test Atmospheres. As-
                                             semble the apparatus as shown In Figure D3.
                                             The  ozone concentration produced by the
                      ing. Accurately weigh+o generator can be varied  by changing the DO-
                     S''from a small glass-   sition of the adjustable sleeve. For calibra-
                                             tion  of ambient  air analyzers,  the  ozone
                                             source should be capable of producing ozone
                                             concentrations in the  range 100  to  1,000
                                             jug./m.8  (0.05 to 0.5 p.p.m.) at a flow rate of
                                             at least 5 liters per minute. At all times the
                                             airflow through the generator must  be great-
                                             er than the total flow required by the sam-
                                             pling systems.
                                               8.2.2   Sampling and Analyses of  Test At-
                                             mospheres. Assemble  the KI sampling  train
                                             as shown In  Figure  D4. Use ground-glass
                                             connections upstream from  the Implnger.
                                             Butt-to-butt connections with Tygon tubing
                                             may  be  used. The manifold distributing the
                                             test  atmospheres must be  sampled  simul-
                                             taneously by the KI sampling train and the
                                             Instrument to be calibrated.  Check assem-
                                             bled  systems  for leaks.  Record the Instru-
                                             ment response  in  nanoamperes  at   each
                                             concentration  (usually six). Establish these
                                             concentrations  by analysis, using the  neu-
                                             tral  buffered  potassium iodide  method  as
                                             follows:
                                               8.2.2.1  Blank. With ozone lamp off, flush
                                             the system for several  minutes  to remove
                                             residual ozone. Pipet 10 ml.  absorbing re-
                                             agent into each absorber. Draw air from the
                                             ozone-generating system through the sam-
                                            pling train at 0.2 to  1  liter/minute  for  10
                               wt  As203  (g.)
                                   49.46
ml. standard arsenious oxide solution into a
300-ml.  Erlenmeyer  flask.  Acidify slightly  mtautes.  Immediately  transfer the  exposed
with 1:10 sulfuric acid, neutralize with solid  solution to a clean 1-cm. cell. Determine the
sodium bicarbonate, and add about 2 g. ex-
cess. Titrate with the standard iodine  solu-
tion using 6 ml. starch solution as indicator.
Saturate the  solution with carbon dioxide
near the end point by adding 1 ml. of 1:10
sulfuric acid. Continue the tltratlon to the
first appearance of a blue color which per-
sists for 30 seconds.
               ml. AssOaX Normality
 Normality  Ia=	
                         ml. la
  6.9  Diluted Standard Iodine. Immediately
before use, pipet  1 ml. standard iodine solu-
tion into a 100-ml. volumetric  flask and
dilute  to volume with absorbing reagent.
  7. Procedure.
  7.1  Instruments can be constructed from
the components given here or may be  pur-
chased. If commercial Instruments are used,
follow the specific Instructions given in the
manufacturer's manual.  Calibrate  the in-
strument as directed in section 8. Introduce
samples into the system under the same con-
ditions of pressure and flow rate as are used
In calibration. By proper adjustments of zero
and span controls, direct reading of ozone
concentration Is possible.
  8. Calibration.
  8.1  KI Calibration Curve. Prepare a curve
of  absorbance  of various Iodine  solutions
against  calculated  ozone  equivalents  u
follows:
                                                    absorbance at 352  nm.  against  unexposed
                                                    absorbing  reagent  as the reference.  If the
                                                    system blank gives an absorbance, continue
                                                    flushing the ozone generation system until
                                                    no absorbance is obtained.
                                                      8.2.2.2  Test Atmospheres. With the ozone
                                                    lamp operating, equilibrate the system  for
                                                    about 10 minutes. Pipet 10 ml. of absorbing
                                                    reagent Into each absorber and collect sam-
                                                    ples for 10  minutes  in the concentration
                                                    range desired for  calibration.  Immediately
                                                    transfer the solutions from the two absorb-
                                                    ers to clean 1-cm. cells. Determine the ab-
                                                    sorbance of each at  3S2 nm. against unex-
                                                    posed absorbing reagent as the reference. Add
                                                    the absorbances of the two solutions  to ob-
                                                    tain total absorbance. Read total /ig.O3 from
                                                    the calibration curve (see 8.1). Calculate to-
                                                    tal volume of air sampled corrected to ref-
                                                    erence conditions of 28° C. and 760 mm. Hg.
                                                    as follows:
                                                                      P.    298
                                                             VE=VX	 X	X10-3
                                                                     760  t+273
                                                      VK  = Volume  of ah" at  reference  condi-
                                                             tions, m.3
                                                      V    =Volume of  air at' sampling  condi-
                                                             tions, liters.
                                                      P   = Barometric  pressure  at  sampling
                                                             conditions, mm. Hg.
                                                      t   = Temperature at sampling conditions
                                                             °C.
                                                      10-s=Converslon of liters to m.'
FEDERAL REGISTER,  VOL. 36, NO. 84—FRIDAY,  APRIL 30,  1971
                          38

-------
                          RULES  AND  REGULATIONS
  Calculate ozone concentration In p.p.m. as
follows:
                  Ag.O.
         p.p.m. O»=	X 6.10 X10-4
                    v»
  8.2.3  Instrument  Calibration  Curve. In-
strument response from the photomvutlpller
tube is ordinarily In  current or voltage. Plot
the  current,  or  voltage  If  appropriate,
(y-axls)  for  the test atmospheres  against
ozone concentration  as determined  by the
neutral buffered  potassium Iodide method,
In p.p.m. (x-axis).
  9. Calculations.
  9.1  If a recorder Is used  which has been
properly zeroed and spanned, ozone concen-
trations can be read directly.
  9.2  If the DC amplifier is read directly,
the reading  must be  converted to ozone
concentrations using  the Instrument calibra-
tion curve (8.2.3).
                                            9.3  Conversion between p.pjn.  and ng./
                                          m.3 values for ozone can be made as follows:
                                                  p.p.m. Os=
                                                              . O«
                                                                  X 5.10x10-'
                                            10. Bibliography.
                                            Hodgeson, J. A., Martin, B. E., and Baum-
                                          gardner, R. E., "Comparison of Chemlluml-
                                          nescent Methods for Measurement of  At-
                                          mospheric  Ozone", frepf'iKr, Ei*4tr r\ Antl^Hi.*!
                                          S^n/wii*,^.^*,* f. , Cc><.tt.T II 7O.
                                            Hodgeson, J. A., Stevens, B. K., and Martin,
                                          B. B., "A Stable  Ozone Source Applicable as
                                          a Secondary Standard for Calibration at At-
                                          mospheric Monitors", P,-*pfini -Vf Ti-ytiO, l**4rv.-ntsn
                                          Scciety of Ammrict, Inlfrnit^iji C'*.>ffff/a:ti Sc £*t>'6i°f
                                          Cfiicjfu.Jtfinft, Otftber /tTI.
                                            Nederbragt, Q. W., Van der Horst, A.,  and
                                          Van Duljn, J., Nature 206, 87 (1986).
                                            Warren, Q. J., and  Babcock, G., Rev.  Set.
                                          Instr.41,aBO (1970).
                    SAMPLE AIR IN
EXHAUST
 6mm

                                              ETHYLENE IN
                                   s^    6mm
                              •<-  10mm

                                6 mm

             2 mm
                         * j <"**""* 2 nuw
            PHOTOMULTIPLIER TUBE
                                             PYREX CONSTRUCTION
                                               • EPOXY SEALED OPTICALLY FLAT
                                                PYREX WINDOW ON END
                       Figure Dt.  Detector cell.
             FEDHAl UGISTM. VOL 36,  NO. 84—FRIDAY, APRIL 30,  1971

-------
                     RULES AND REGULATIONS
6-In. PEN-RAY
   LAMP
  ALUMINUM
BOX ENCLOSURE
                         Figure D2. Ozone source.
     —7  FLOW METER
       I (0-10 liters/mln)
           FLOW
        CONTROLLER
        CYLINDER
           AIR
                                              SAMPLE
                                        ^fLJfLJlL

                               VENT-=*    MANIFOLD
                     5 liters/min
                                                           OZONE
                                                          SOURCE
                  Figure D3. Ozone calibration atr supply, source, and
                          manifold system.
        FEDERAL REGISTER, VOL. 36, NO. 84—FRIDAY, APRIL 30, 1971
                             40

-------
                                                   .RUBBER TUBING
 FUOWMETER
                                                                              TO AIR
                              Figure D4. Kl sampling (rain.
                ADDENDA

  A. Suggested Performance  Specifications
/or Atmospheric Analyzers for Hydrocarbons
Corrected for Methane:
Range (minimum)	
Output (minimum)	
Minimum detectable sen-
  sitivity 	
Zero drift (maximum)	
Span drift (maximum) _-
Precision  (minimum) —
Operational period (mini-
  mum)  	
Operating   temperature
  range (minimum)	
Operating humidity range
  (minimum)	
Linearity  (maximum)	
                                      full.
0-5 p.p.m. THC.
0-5 p.p.m. CHt.
0-10   mv.
  scale.
0.1  p.p.m.
0.1 p.p.m. CHj
Not   to
  1   perc^it/24
  ho;
Not   tof exceed
  1   j^rcent/24

      ercent.

    '  C.
  -100 percent.
  percent of full
  scale.
                          of Performance

                    \ and maximum meas-
  B. Suggested Defli
Specifications:
Range—The minlm^
  urement limits.
Output—Electrical signal  which Is propor-
  tional to the^measurement; intended for
  connection Jp readout or data processing
  devices.  Usually expressed as millivolts or
  mllllamps^full scale at a given Impedence.
Full ScaleAThe maximum measuring limit
  for a gwen range.
        l Detectable Sensitivity—The small-
  est Tmount of input concentration  that,
         detected as the concentration  apj
     aches ^ero.
                                           Lag
         -The degree of agreement between
         ured value and the true value;  us;
        expressed at  ± percent of full :
          -The time interval from  a
   Eange in input concentration at tha
   .rument inlet to  the first  corresponding
      age in the instrume
                                  *ine In-
  terval from a step change in the iAut con-
  centration  at the  InstrumentyrUet to  a
  reading of  90 percent  of the yitimate re-
  corded concentration.
Rise Time (90  percent)—Th^ interval  be-
  tween Initial response tlmXand time to 90
  percent response after a/step decrease in
  the inlet concentratioi
Zero Drift—The change ^T Instrument output
  over a stated time period, usually 24 hours,
  of  unadjusted contiguous operation, when
  the input  concentration is  zero;  usually
  expressed as perc»t full scale.
Span Drift—The^iange In instrument out-
  put over a st^ed  time period, usually 24
  hours, of unadjusted continuous operation,
  when the yput concentration is a stated
  upscale vafce; usually expressed as percent
  full scale
Frecislou^The degree of agreement between
        pd measurements of the same con-
        ;ion. It Is expressed as the avert
           of the single- results from
     an.
   'rational Period—The period of timjfover
   /hich  the instrument can be expected to
  operate nnatteu-.led within neciflc^ions.
Moise—-Spontaneous deviations froWT a mean
  output not caused by input  concentration
  changes.
                                                                                        Iti -C-rfcrenoe-  An undesired pq
                                                                                         tive output  caused by a spb.siance other
                                                                                         than the one being measured.
                                                                                        Interference Equivalent—-Whe portion of in-
                                                                                         dicated input concenj/atlon due  to the
                                                                                         presence of an inte:
                                                                                        Operating Temperatuj^ Range—The range of
                                                                                         ambient temperances over which the in-
                                                                                         strument will me^ all performance specifi-
                                                                                         cations.
                                                                                                          Operating  Humidity  rfange— Tlie  ra^ige of
                                                                                                            ambient relative l^nldlty over whicn the
                                                                                                            instrument   willVrneet  all  performance
                                                                                                            specifications.
                                                                                                          Linearity—The^axtmum deviation between
                                                                                                            an actual ^mstrument reading  and  the
                                                                                                            reading pyldicted by a straight line drawn
                                                                                                            between/ipper and lower calibration points.
                                                                                                                                         7^
                                                                                                                                    -n.
                                                                                           He
                                                                                         SUPPLY
                                                                                                                                   SUPPLY

                                                                                                                             ffiPLE OUT    I


                                                                                                                           ' SAMPLE IK
                                                                                                                         FLAME
                                                                                                                        DETECTOR
                                                                                                                           I L
                                                                                                                        WPPER COLUIW
                                                                                                                                 HYDROGEN
                                                                                                                                GENERATOR
                                                                                                                                                                   UCTROMETER
                                                                                                                           I BACKFLUSH VALVE

                                                                                                                            VENT


                                                                                                                            Ht PURGE
                                                                                                                       Figuie E1  Typical SKVI diagram.
                                                                                                                                                           -NICKEL
                                                                                                                                                           REACTOR
APPENE/K P -REFERENCE METHOD FOR
  THEybETERMINATION  OF  NITROGEN ,
  DIOXIDE  IN  THE  ATMOSPHERE
         SAMPLING METHOD)

   f. Principle and Applicability.
   .1  Nitrogen dioxide Is collected by>Tub-
 illng air through a sodium hydroxicysolu-
'tlon to  form  a  stable solution ofyfodium
nitrite. The nitrite ion produced dujrng sam-
pling is determined colorimetricaU^by react-
ing  the  exposed  absorbing  ryfgent  with
phosphoric acid,  sulianilamlde,  and  N-l-
naphthylethylenediamine dlUydrochlorlde.
  1.2  The method Is applldfble to collection
or 24-hour samples  in  tm?  field and sub-
sequent  analysis in the Mooratory.
  2.  Range and SensitMiity.
                                                                                                                                     f'L.l  The range of the analysis Is OM
                                                                                                                                     g. NO;/ml.  With 50 ml. absorbing r
                                                                                                                                                                           1.5
                                                                                                                                                                          ent
and a sampling rate of 200 ml./mln^for 24
hours,  the range of the  method  If 20-740
Ag./m.= (0.01-0.4 p.p.m.) nltrogenyjoxide.
  2.2  A  concentration  of  0.04 A. NO;'ml.
will  produce an absorbance f 0.02 using
1-cm. cells.
  3. Interferences.
  3.1  The Interference o^sulfur  dioxide is
eliminated by converting! to' sulfurtc  acid
with hydrogen  peroxlde/before analysis. (I)
  4. Precision, Accuram, and Stability.
  4.1  The relatlve^fandard d;vlatlons are
14.4  percent and JH.6  percent  at nitrogen
dioxide concentryons of 140 «g./m.a (0.072
p.p.m.) and 20oyf./m.' (0.108 p.p.m.), respec-
tively,  based mi an automated analysis  of
                                                                                        C
                                                                                        m
                                                                                        
-------
                                                    TECHNICAL REPORT DATA
                                         (Please read Instructions on the reverse before completing)
 1. RETORT NO.
      l£l'A-650/4-75-016
   TITLE AND SUBTITLE
      Collaborative Study of Reference Method for Measurement
      of Photochemical Oxidants in the Atmosphere
      (0/one-Ethylene Chemiluminescent Method)
                   3. RECIPIENT'S ACCESSION-NO.
                   5. REPORT DATE
                       February, 1975(date of preparation)
                   6. PERFORMING ORGANIZATION CODE
 7. AUTHOH(S)

      H.C. McKee, R.E. Childers, V.B. Parr
                   8. PERFORMING ORGANIZATION REPORT NO.

                       01-2811(1975)
 9. PERFORMING ORGANIZATION NAME AND ADDRESS
      Southwest Research Institute
      8500 Culebra Road
      San Antonio, Texas  78284
                                                                                        10. PROGRAM ELEMENT NO.
                   11. CONTRACT/GRANT NO.

                       CPA 70-40
 12, SPONSORING AGENCY NAME AND ADDRESS
      Methods Standardization & Performance Evaluation Branch,
      Quality Assurance and Environmental Monitoring Laboratory, NERC
      Research Triangle Park, North Carolina  27711
                                                                                        13. TYPE OF  REPORT AND PERIOD COVERED
                       Test Report-Mar 70-Sep 75
                   14. SPONSORING AGENCY CODE
 15. SUPPLE!V1£NTARY NOTES
 16. ABSTRACT
      This report contains information on collaborative tests to determine the precision and bias of the reference method for measurement of
      photochemical oxidants as published by the Environmental Protection Agency in the Federal Register, April 30, 1971.
      In the first phase test, ten collaborators were assembled at a common site to measure (from a common source) ambient and ozone sup-
      plemented concentrations over a range from 0 to 0.510 ppm (0-lOOOjug/m3). The data were analyzed to derive an estimate for the
      precision of the method. Estimates of the standard deviations of the process for total (Sy), between-laboratory (S^) and within-labor-
      atory (S\fj[} were derived, as well as.for the lower detectable limit (LDL). The results were:
                      ST     0.0033 + 0.0412 x      0*SS (T'nis Report)
  UNCLASSIFIED	
20. SECURITY CLASS (This page)
  UNCLASSIFIED
21  NO. OF PA~,ES
    51
                                                                                                            22, PRICE
EPA Form 2220-1 (9-73)
                                                                 43

-------