~N /I
rapes Copy
EPA-600/4-78-004
January 1978 Environmental Monitoring Serfes
SUMMARY OF AUDIT PERFORMANCE MEASUREMENT
OF SO-,NOrCO, SULFATE, NITRATE -1976
"- L L , ;
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine 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 (STAR)
7 Interagency Energy-Environment Research and Development
8 "Special" Reports
9 Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL PROTECTION TECH-
NOLOGY series This series describes research performed to develop and dem-
onstrate instrumentation, equipment, and methodology to repair or prevent en-
vironmental degradation from point and non-point sources of pollution. This work
provides the new or improved technology required for the control and treatment
of pollution sources to meet environmental quality standards
This document is available to the public through the National Technical Informa-
tion Service. Springfield, Virginia 22161.
-------
600478004
SUMMARY OF AUDIT PERFORMANCE
MEASUREMENT OF S02, N02, CO, SULFATE, NITRATE
1976
by
Steven M. Bromberg, Berne I. Bennett and Robert L. Lampe
Quality Assurance Branch
Environmental Monitoring and Support Laboratory
Research Triangle Park, North Carolina 27711
ENVIRONMENTAL MONITORING AND SUPPORT LABORATORY
QUALITY ASSURANCE BRANCH
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Monitoring and Support
Laboratory, U.S. Environmental Protection Agency, and approved for publication.
Mention of trade names or commercial products does not constitute endorsement
or recommendation for use.
-------
ABSTRACT
The Quality Assurance Branch of the Environmental Monitoring and Support
Laboratory, Research Triangle Park, North Carolina conducts an ambient air
audit program. Measurement principals for sulfur dioxide, nitrogen dioxide,
carbon monoxide, sulfate and nitrate are audited on a semi-annual basis.
Unknown samples were sent to requesting laboratories. The analytical results
were returned to the Quality Assurance Branch for evaluation. After pro-
cessing, an individual report was returned to each participant.
This report contains a summary of audit results obtained during 1976.
m
-------
CONTENTS
Abstract i i i
Fi gures v
Tab! es vi
Acknowledgment viii
1. Introduction 1
2. Program Coordination 3
3. Statistical Approach 5
4. Audit Materials
S02 8
N02 9
CO 10
S04 - N03 10
5. Results
S02 12
N02 18
CO 25
S04 31
N03 38
6. Summary 45
References 46
Appendices
iv
-------
FIGURES
Number Page
1 Example of Individual Report 4
-------
TABLES
Number Page
1 Mean and Standard Deviation of SCL Acceptance Analysis 9
2 Mean and Standard Deviation of N02 Acceptance Analysis 9
3 Mean and Standard Deviation of CO Acceptance Analysis 10
4 SO^ and NO., Sample Values 11
5 S02 Agency Distribution 12
6 S02 Sample and Target Ranges 13
7 S02 Absolute Percent Difference 14
8 Study 0576 S02 Survey Statistics 15
9 Summary of S02 Analytical Methods Used by Outlier Labs 17
10 Mean and Standard Deviation of S02 Results by Analytical Method 17
11 N02 Agency Distribution 18
12 N02 Analytical Methods Used During Study 0676 19
13 N02 Sample and Target Ranges 19
14 N02 Absolute Percent Di fference 21
15 Study 0676 N02 Survey Statistics 22
16 Summary of NOp Analytical Methods Used by Outlier Labs 23
17 Mean and Standard Deviation of N02 Results by Analytical Method 24
18 CO Agency Distribution 25
19 CO Analytical Methods Used 26
20 CO Sample and Target Ranges 26
21 CO Absolute Percent Difference 28
-------
TABLES (CONTINUED)
Number Page
22 Study 0476 CO Survey Statistics 29
23 Summary of CO Analytical Methods Used by Outlier Instruments 30
24 Mean and Standard Deviation of CO Results by Analytical Method 30
25 S04 Agency Distribution 31
26 S04 Analytical Methods Used 32
27 SO, Sampl e and Target Ranges 33
28 SO. Absolute Percent Difference 34
29 Study 0476 S04 Summary Statistics 35
30 Summary of S04 Analytical Methods Used by Outlier Labs 36
31 Mean and Standard Deviation of S04 Results by Analytical Method 37
32 NOo Agency Distribution 38
33 N03 Analytical Methods Used 38
34 NO, Sampl e and Target Ranges 39
35 N03 Absolute Percent Di fference 40
36 Study 0476 NOg Summary Statistics 41
37 Summary of N03 Analytical Methods Used by Outlier Labs 42
38 Mean and Standard Deviation of N03 Results by Analytical Method 43
-------
ACKNOWLEDGMENTS
The authors would like to thank all the participants for their coopera-
tion during the past year. Also due a word of thanks are the diligent
programers of the Statistical and Technical Analysis Branch, EMSL, for
developing the data management systems necessary to store and summarize
the audit data.
-------
SECTION 1
INTRODUCTION
Since 1972 the Environmental Protection Agency (EPA) has been engaged in
a Performance Audit program of various monitoring groups throughout the
United States and in several foreign countries. The audit program is one
part of an overall quality assurance program and the result should not be
construed as an absolute indicator of data quality. Used along with informa-
tion obtained from an internal quality control program however, the conclusions
can be quite meaningful.
The purposes of the audit program are twofold: The first, from a partici-
pant standpoint, is most important. Agencies are furnished a means of rapid
self-evaluation of the specific operation under study. The second objective
of the program is to provide EPA with a continuing tndex of the quality of
data being reported to air quality data banks.
The program is being coordinated through the 10 EPA Regional Offices (RO)
by the Quality Assurance Branch (QAB) of the Environmental Monitoring and
Support Laboratory (EMSL), Environmental Research Center, Research Triangle
Park, North Carolina 27711. Comments, questions, or applications to partici-
pate in the Program, should be sent to the above address.
Previous to this publication, individual reports have been written at the
conclusion of each audit. Because of the increasing number of audits being
conducted each year, and because much of the background material contained in
each report is repetitive, it was decided to publish one report each year.
-------
Thus, beginning with this report, an annual publication will include all
survey results for the year.
Users of the information contained in this report should take note of
some limitations imposed in gathering the data. With the exception of the
audit for carbon monoxide (CO), the surveys checked only a portion of the
entire system. The sulfur dioxide (S02), nitrogen dioxide (NX^)> sulfate-
nitate (S0,-N03) surveys examined only the analytical portion of the system.
No restrictions were placed on the methodology used by the participants (other
than that the method used had to be compatible with the audit samples). How-
ever, to the extent possible, the various methods used were documented.
The following sections include discussions of the program operation,
descriptions of the audit materials, the statistical approach used to analyze
the data, and the results. The Results Section includes data from April -
September 1976.
-------
SECTION 2
PROGRAM COORDINATION
Participants in the audits were selected by the Regional Quality Control
Coordinator in each of the 10 regions. Once a potential participant has re-
ceived audit samples for a particular pollutant, he is automatically notified
of subsequent surveys for that pollutant. Additional participants are obtained
through individual requests and through notification in the Journal of the Air
Pollution Control Association (JAPCA). Participants are assigned an identifi-
cation number which remains with the agency through all audits for all
pollutants.
After the audit roster is completed for a particular survey, instructional
materials and unknown samples are mailed. The participants are allowed 5 to 6
weeks in which to return their results. After these results are entered into
a data bank, individual reports are returned to the participants. This report
indicates the acceptable ranges for each sample as well as the value reported
by the agency. Figure 1 is an example of an individual report for S02-
-------
INTER-LABORATORY STUDY RESULTS
< MAY 1976 1
POLLUTANT - St>2
123456
Laboratory Name
Laboratory Address
UNITS - MICROGAMS PER CUBIC METER
eE.eQBT.CQ
1571 ..... - ---- 3*. ICO ------ -------------- 32.280 10.259 - 29. CIS «*3.523
254<0 ------------- 70.830 ------- 68.229 ---- 76.170 ...... 57.760 86.633
---- 3«tUl --------------- 70.6SU -------------- -------- 67.755 76.&OH 57.76C 86.&3S
WM92 .- - - 1155.000 ------------- 130.591 138.668 107.70«» 161.555
5691 ----- ------- 176. E60 ------- 180.063 -- 169.676 ....... 117.896 221. £43
FIGURE 1. Example of Individual Report*
-------
SECTION 3
STATISTICAL APPROACH
Of primary concern at the onset of the program was the establishment of
some type of acceptance criteria to enable participants to judge their per-
formance. Initially, there was little guidance and the choices of acceptable
limits were somewhat arbitrary. As more data have been gathered through the
audit program more realistic acceptance limits have been set.
For convenience, two acceptable ranges have been defined. The Sample
Range contains variability attributable to the sample material. This estima-
tion was based on the repeated analysis of several samples from each concen-
tration range by one laboratory. These ranges should be goals of each agency;
it is not reasonable to expect all laboratories to fall within these ranges
on any given analysis. However, falling within these ranges repeatedly in-
dicates a laboratory with excellent precision and accuracy. Falling outside
the Sample Range does not indicate problems.
The Target Range has been constructed to include sample variability and
between laboratory variability. A laboratory falling outside the Sample Range
but within the Target Range should feel comfortable with the results. If,
however, the results fall outside the Target Range, the accuracy of the analysis
should be considered suspect.
As can be seen from Figure 1, and as mentioned above, two sets of limits
are given for each sample value. The Target Range is based in part on past
performance of all participants. The percent difference between the reported
-------
results and the true value (result determined by QAB) was established for past
surveys. Using this percent difference for all samples over all surveys an
average difference was determined. This average difference was used to calcu-
late the Target Ranges.
To establish the Sample Range it was decided the probability (P) of %D
(as defined below) should not be greater than p%, more than 5% of the time.
i P nr I observed value - true value , v ,nn ^ 0/1 M
i.e.. pt-l - true value - ' x 10° > p%^ ^ 5%
Defining P such that P> t/ nf. n s
\ UD $ n~ i / "JIJI"L
where:
t = "Students t"
-^- = coefficient of variation determined by QAB for each
x sample concentration
%D = observed value - true value x 100
true value
x = mean of samples analyzed by QAB
s = standard deviation of sample mean
In using the above assumption, it is desired to determine what percent differ-
ence should be exceeded in only one case out of 20 for a particular laboratory,
then:
By knowing the number of samples, the sample mean and standard deviation, the
p% can be determined. It is this percentage that is used to construct the
Sample Range.
When evaluating results based on the acceptable ranges, one must take
into consideration the uses of the data. For some purposes the Sample Range
-------
may be too wide, for other uses the Target Range may be too narrow. Thus,
judgement should be exercised when comparing ranges with results.
A preferable method of evaluating performance "after the fact" is to
compare individual results with the respective tables entitled "Absolute Per-
cent Difference" found in the Results Section. These tables are frequency
distribution tables of the percent difference between EPA and reported results,
The relationship of individual performance to the overall study performance
can be determined using these figures.
Two outlier tests were used to select data for analysis. The first test
was used to screen results that were grossly in error. To be eliminated from
further analysis using this technique, a laboratory had to report all samples
outside the Target Ranges for the respective pollutants. This data appears
only in the appendix and is marked with an asterisk (*). No statistical anal-
yses in this report contain these data.
A second outlier test, using more traditional techniques, was also used.
Data greater than +2 standard deviations from the study mean (mean of all re-
ported data for each concentration) were identified as outliers. These data
are identified in the "outlier removed" columns of selected tables. Other
data appearing in table columns labeled "all data" or containing no specific
designation, contain results subjected to the first outlier test, but not the
second.
Several summaries have been used to condense the large amounts of data
into a more manageable form. The Tables in the Appendix are sequential list-
ings of all data by sample concentration for each audit. The values at the
top of the table are statistical summaries of the sequentially listed data.
-------
SECTION 4
AUDIT MATERIALS
AMBIENT SULFUR DIOXIDE SAMPLES
The commercially produced sample material consisted of freeze-dried
mixtures of sodium sulfite and potassium tetrachloromercurate (TCM) contained
in 5-ml glass vials with inert screw cap closures. A set of samples was com-
prised of five vials containing from 8.4 to 57.6 micrograms (yg) of S0?
equivalent per vial. Vials were chosen in place of ampoules to facilitate
quantitative transfer of the contents. Difficulty had been experienced in
transferring liquid from the small ampoules due to the surface tension..
The samples were placed in freezers immediately upon delivery with the ex-
pectation that a low temperature would preserve the integrity of the material.
Initial EPA analyses were performed in January, 1976. Subsequent reanalyses
after several months demonstrated that freezing did not completely stabilize
the sulfite content in that the S02 levels continued to gradually decline.
The analyses were performed by the reference method for the determina-
tion of S02 in the atmosphere (pararosaniline method). The sample, when
dissolved in 0.04N potassium tetrachloromercurate (TCM) forms a dichloro-
sulfitomercurate complex. The complex is reacted with pararosaniline and
formaldehyde to form intensely colored pararosaniline sulfonic acid. The
absorbance of the solution is measured spectrophotometrically at 548 nm.
Ten samples from each concentration were analyzed. Table 1 lists the
results of these analyses. The values contained in Table 1 are those used
-------
in determining the acceptable ranges reported to participants. The values
listed are based on the assumption that the sample was collected in 50 ml of
absorbing reagent and the total sample air volume was 300 liters.
TABLE 1. MEAN AND STANDARD DEVIATION OF S02 ACCEPTANCE ANALYSIS (vg/m3)
Mean
Stand. Dev.
Cone. 1
36.27
1.77
Cone. 2
72.87
1.77
Cone. 3
71.70
1.93
Cone. 4
134.63
1.77
Cone. 5
184.87
2.10
AMBIENT NITROGEN DIOXIDE SAMPLES
The commercially produced samples consisted of 4 ml of aqueous sodium
nitrite in 5 ml glass vials with inert screw cap closures. A set of samples
consisted of 5 vials. The samples, when mixed with absorbing reagent, simu-
lated ambient samples ranging from 0.25 yg/ml to 1.02
EPA analysis of 10 samples from each concentration was performed using
2
a candidate method for the determination of N02 in ambient air. Measurements
were made on a Varian Model 635 UV-Vis spectrophotometer at 540 nm. Aqueous
calibration standards were prepared to encompass a linear range from 0 to
1.6 yg/ml of NOp. Table 2 lists the results of these analyses. The values
contained in Table 2 are those used in determining the acceptable ranges
reported to participants. The values listed are based on the assumption
that the sample was collected in 50 ml of absorbing reagent.
TABLE 2. MEAN AND STANDARD DEVIATION OF N02 ACCEPTANCE ANALYSIS (yg/ml)
Mean
Stand. Dev.
Cone. 1
0.25
.004
Cone. 2
0.41
.008
Cone. 3
0.41
.005
Cone. 4
0.83
.006
Cone. 5
1.02
.003
-------
AMBIENT CARBON MONOXIDE SAMPLES
The samples consisted of commercially produced mixtures of CO and artifi-
cial air. Mixtures also contained methane and 350 ppm of C02. Specially treated
aluminum cylinders were used to improve gas stability. Sample concentrations
ranged from 3 ppm to 45 ppm of CO. Each participant received a set of 3
cylinders, one from each concentration level.
Analysis by EPA of cylinders from each concentration was performed using
a Bendix Model 8501 NDIR analyzer. National Bureau of Standards SRM gases
were used as the reference standards. Three SRM's at nominal concentrations
of 9, 46, and 95 ppm CO were used as the primary reference gases.
Ten samples from each concentration were analyzed. Table 3 lists the
results of these analyses. The values contained in Table 3 are those used in
determining the acceptable ranges reported to participants.
TABLE 3. MEAN AND STANDARD DEVIATION OF CO-ACCEPTANCE ANALYSIS (ppm)
Mean
Stand. Dev.
Cone. 1
3.43
0.02
Cone. 2
20.15
0.01
Cone. 3
40.60
0.07
SULFATE/NITRATE SAMPLES
The commercially prepared samples consisted of 3/4" x 8" glass fiber
filter strips with depositions of potassium sulfate (KpSO.) and lead nitrate
PbN03. Filter strip samples were comprised of combinations of differing SO*
and N03 concentrations. Each strip was packaged in a plastic envelope. The
concentration of sulfate ranged from 0 (clank) to 40 yg/m . Nitrate concen-
o
trations ranged from 0 (blank) to 15 yg/m. Concentrations were calculated
assuming that the samples were collected on an 8" x 10" filter with a total
air volume of 2000 m .
10
-------
It was felt that the gravimetric preparation of the solution deposited
onto the filter strips was more accurate than analysis using existing proce-
dures. Thus, the values accepted as "True-Values" are the vendor certifica-
tions. Verification analyses insured that the accuracy and precision of
the samples was acceptable. Table 4 lists the concentrations of the samples
used during the audit. The values listed are based on the assumption that
the filter strip was 1/12 of the total filter and the total air volumed
o
sampled was 2000 m .
TABLE 4. SULFATE AND NITRATE SAMPLE VALUES (ug/m3)
Sulfate
Nitrate
Cone. 1
0.00
0.00
Cone. 2
4.47
6.22
Cone. 3
21.53
7.64
Cone. 4
39.29
11.27
11
-------
SECTION 5
RESULTS
AMBIENT SULFUR DIOXIDE
Participant Characteristics
Sulfur dioxide study number 0576 began in May 1976. Out of 166 sample
sets requested by participants, 143 sets of data were returned, for a re-
sponse rate of 86%. Table 5 indicates the monitoring agency type distribution.
TABLE 5. S02 AGENCY DISTRIBUTION
Agencies Requesting Samples
Agencies Returning Data
Foreign
2
2
EPA
10
7
State
60
55
Local
76
67
Private
18
12
Total
166
143
Methods used to analyze the samples were grouped into three broad cate-
gories: pararosaniline-manual, pararosaniline-automated, and other. Of the
143 monitoring groups responding to the study, 109 used the manual pararosani-
line technique, 32 used the automated pararosaniline method, and 2 used other
methods. It should be noted that laboratories tend to define the method used
in very general terms. Thus, an agency reporting to have used the manual
pararosani1ine technique may not have used the method exactly as it appears
in 40 CFR - 50.11^.
Acceptable Ranges
As described in a previous section, two ranges were used as one
means of judging performance. The Sample Ranges and Target Ranges for
12
-------
S0? 0576 are listed in Table 6. The ranges apply to sample values in
increasing concentration.
TABLE 6. S02 SAMPLE & TARGET RANGES
Sample Range
Target Range
Cone. 1
± H%
+ 20%
Cone. 2
+ 5.5%
+ 20%
Cone. 3
+ 6.1%
+ 20%
Cone. 4
+ 3.0%
+ 20%
Cone. 5
+ 2.6%
+ 20%
The Sample Range percentages were determined by the method described
under Statistical- Approach, the Target Range percentages were determined
using previous audit data. It was determined from earlier studies that the
average percent difference between reported results and EPA determined values
for all sample concentrations across all agencies was +_ 20 percent. Thus,
this value has been used for all Target Ranges.
Using the above criteria, a tabulation was made of the number of agencies
reporting results within the ranges. The noteable figure revealed by this
tabulation is that 67 laboratories (47%) were not able to get at least one
value within the Sample Ranges. Four agencies reported results in which all
five results were within the Sample Ranges. The equivalent results for the
Target Range showed that 14 laboratories reported all 5 samples outside the
Target Ranges, while 53 laboratories (37%) reported results in which all 5
samples were within the Target Ranges.
Data Summary
Using the Target Ranges as one means of eliminating non-representative
data it was decided that the laboratories reporting all 5 results outside
the Target Ranges (i.e. had all results greater than +_ 20 percent of the EPA
values) would be removed from the data base and excluded from further evalua-
tions. The Target Ranges are broad enough that, unless the determination is
13
-------
totally out of control, at least one value should fall within the ranges if
the laboratory is performing similarly to most of the study population. Thus,
if the agency is not performing in a similar manner, its data should not be
included in the summaries.
The remainder of the discussion will be concerned with the reported
results minus the 14 data sets that met the exclusion criterion. These data
points are marked in the Appendix by an asterisk (*).
Table 7 is a frequency distribution of the percent difference between
the reported and EPA values for each sample concentration. The differences
were calculated by the following equation:
absolute percent difference = Deported vajlue^EPA valuej x m
The frequency distribution was then constructed and appears below. For ex-
ample, 50% of the reported results for sample Concentration 1 were less than
or equal to an absolute percent difference of 17 percent. It should be noted
that the "All Samples" line is not an average of the numbers appearing above it,
but is the distribution resulting when all the data is examined together re-
gardless of its concentration.
TABLE 7. S02 ABSOLUTE PERCENT DIFFERENCE
Cone. 1
Cone. 2
Cone. 3
Cone. 4
Cone. 5
All Samples
No.
127
126
128
128
129
638
Min.
.63
.24
.07
.05
.10
.05
10%
4
2
2
3
1
2
30%
9
6
7
5
4
6
50%
17
12
10
9
7
10
70%
28
15
16
11
9
15
90%
47
27
26
16
15
28
Max.
94
94
73
66
71
94
Mean
22
14
13
10
8
13
The data in Table 7 are very useful for laboratories trying to determine
their performance relative to other participants. For example, only 10% of
14
-------
the labs reporting results for Concentration 1 had a percent difference of
4.0% or less while 50% of the laboratories reported a percent difference of
17% or less for the same concentration. The Table also indicates that the
average percent difference for all laboratories for all samples was 13%.
All data received (with exception of previously omitted data) were grouped
according to concentration. The results for each sample value are listed in
the Appendix in increasing concentration. The summary statistics which appear
in Table 8 and at the top of each listing in the Appendix do not include the-
14 data sets identified as outliers.
Table 8 lists summary statistics based on the reported data. The "out-
lier removed" columns contain data on which the outlier criterion was applied;
all data greater than +_ 2 standard deviation were removed. It should be noted
that the variation in the number of samples from concentration to concentration
is due to laboratory accidents and damaged samples.
TABLE 8. STUDY 0576 S02 SURVEY STATISTICS(^g/m3}
Number
True- Value
Mean
Median
Range
Std. Dev.
Coef. Var.
Skewness
2
Accuracy
Concentration 1
all
iata
127
36.3
30.8
31.0
58.0
8.6
27.8
1.11
14.5
outlier
removed
124
36.3
30.0
30.8
40.9
7.0
23.5
0.0
-15.1
Concentration 2
all
data
126
72.2
64.4
65.1
96.8
11.2
17.3
-1.5
-9.9
outlier
removed
122
72.2
65.2
65.3
38.8
8.0
12.2
0.0
-9.5
Concentration 3
all
data
128
72.2
67.6
65.7
98.3
11.7
17.3
1.4
-9.0
outlier
removed
124
72.2
66.9
65.7
52.9
8.9
13.3
1.0
-9.0
Concentration 4
all
data
128
134.6
125.1
124.5
109.5
.14.1
11.2
-1.4
-7.5
outlier
removed
126
134.6
126.1
124.9
56.0
11.4
9.0
0.4
-7.2
Concentration 5
all
data
129
184.9
176.1
175.0
173.7
18.8
10.7
-1.7
-5.3
outlier
removed
128
184.9
177.1
175.4
92.0
15.4
8.7
1.0
-5.2
1
Skewness - A statistic which indicates the lack of
For a normal distribution this value is
symmetry in a distribution.
near zero.
"Accuracy -
Median - True Value
True Value
x 100
15
-------
Several observations can be made from the data. The mean and median
agree very well. This indicates that the distribution is normally distributed
(also indicated by skewness) and that the median is a good predictor of the
study average. It is also apparent, both from the accuracy figure and from a
comparison of the mean or median with the True-Value, that a bias exists over
the entire concentration range. Also, the coefficient of variation indicates,
for the lower concentration values, a great amount of variability. Large
variability is to be expected at the very low concentration, but is some
what surprising at the mid-value.
The bias mentioned in the previous paragraph is probably due to an
inaccurate analytical run. However, the quality control information does
not indicate this was the case. There were not enough samples remaining
after the survey to permit an investigation of the discrepancy.
Each laboratory data set was plotted against the corresponding EPA data
set and the slope and intercept from the linear regression line was deter-
mined. The mean of 129 slopes was 1.018 with a standard deviation of .089.
The mean intercept was 5.9 with a standard deviation of 10.6. From this data
it can be seen that the average slope is very near ideal with little vari-
ability. However, the intercept is high with much scatter. When the high
intercept of 5.9 is added to the mean or median of each concentration, the
result is very close to the True-Value.
Analytical Method Summary
A check was made to determine if the 14 data sets excluded from the sum-
mary were related to the method used for analysis (i.e., was any method
responsible for most of the data outliers). The following table (Table 9)
resulted.
16
-------
TABLE 9. SUMMARY OF S02 ANALYTICAL METHODS USED BY OUTLIER LABS
Method
Pararosam 1 i ne-Manual
Pararosani 1 i ne-Automated
Other
Total No.
Using Method
109
32
2
No. Identified
as Outliers
10
4
-
Percent of
Total as Outlie-s
9
9
-
As can be seen, no one method contributed an unusually large portion of the
outlier data.
To determine if a particular analytical method produced biased results,
Table 10 was developed. This Table contains the mean and standard deviation
of each sample concentration for each method used.
TABLE 10. MEAN AND STANDARD DEVIATION OF S02 RESULTS BY ANALYTICAL METHOD (yg/m3)
Manual
Automated
True-Value
Cone. 1
Mean
31.11
29.53
Std. Dev.
9.05
6.89
36.27
Cone. 2
Mean
63.78
66.32
Std. Dev.
12.22
6.61
72.20
Cone. 3
Mean
67.78
66.88
Std. Dev.
12.88
6.14
72.20
Cone. 4
Mean
124.56
126.88
Std. Dev.
14.84
11.34
134.63
Cone. 5
Mean
174.90
179.96
Std. Dev
19.69
15.57
184.87
There is no apparent bias between the mean of the manual and automated
methods. However, the standard deviations of the automated method are much
lower than those of the manual method. Thus, those agencies using an automated
system should be able to report more precise data than those using a manual
technique.
SUMMARY
The S02 Audit Survey 0576 began in May, 1976. One hundred sixty-six
laboratories requested to participate, of which 143 returned results for a
17
-------
response rate of 86%. Foreign, EPA, State, locaT and private laboratories
submitted data.
Three analytical methods were used. One hundred and nine respondents
used the manual pararosaniline method, thirty-two used the automated para-
rosaniline method and two agencies used other methods.
The overall results show a bias between the study results and the EPA
determined results. This may be due an analytical error on the part of EPA.
The overall average percent difference between the reported results and the
EPA results is _+ 13 percent.
The average of all slopes (EPA vs. reported) was 1.018 with a stand-
ard deviation of 0.089. The average intercept was 5.9 with a standard
deviation of 10.6. This high intercept is consistent with the approximately
6 yg/m difference between the study mean and the EPA determined values.
AMBIENT NITROGEN DIOXIDE
Participant Characteristics
Nitrogen dioxide study number 0676 began in June 1976. A total of 119
laboratories requested samples. Of those samples sent to monitoring groups,
100 sets of data were returned for a response rate of 84%. Table 11 indicates
the monitoring agency type distribution.
TABLE 11. N02 AGENCY DISTRIBUTION
Agencies Requesting Samples
Agencies Reporting Data
Foreign
2
2
EPA
10
8
State
45
38
Local
55
50
Private
7
2
Total
119
100
Methods used to analyze the samples were grouped into 6 broad categories.
Table 12 lists the analytical methods used and the number of respondents using
the method.
18
-------
TABLE 12. N02 ANALYTICAL METHODS USED DURING STUDY 0676
Method
Saltzman-Manual
Saltzman-Automated
Sodium Arsenite-Manual
Sodium Arsenite-Automated
TGS-ANSA - Manual
Other
Number Agencies
Using Method
5
1
65
24
1
4
It should be noted that agencies tend to define the analytical method used
in very general terms. A laboratory reporting the use of the manual sodium
arsenite procedure as the method of choice may have used that procedure with
various modifications. Thus, Table 12 should be interpreted as 65 labor-
atories used a procedure approximating the manual sodium arsenite method.
Acceptable Ranges
As described in a previous section, two performance ranges were used
as one means available for judging performance. The Sample Ranges and
Target Ranges for N02 0676 are listed in Table 13. The ranges apply to
sample values in increasing concentration.
TABLE 13. N02 SAMPLE AND TARGET RANGES
Sample Range
Target Range
Cone. 1
1 12%
+ 24%
Cone. 2
+ 7%
+ 14%
Cone. 3
+ 7%
+ 14%
Cone. 4
± 5%
+ 10%
Cone. 5
+ 2%
+ 4%
19
-------
The Sample Ranges were determined by the method described under Statis-
tical Approach. The Target Ranges were obtained by doubling the Sample Range
percentages.
Using the above criteria, a tabulation was made of the number of agencies
reporting results within the ranges. Eleven laboratories (11%) were not able
to report at least one value within the Sample Ranges. Eight of these 11 agen-
cies also reported all samples outside the Target Ranges. Sixteen laboratories
(16%) reported all 5 results within the Sample Ranges. Fifty-one laboratories
(51%) reported all results within the Target Ranges.
Data Summary
Using the Target Ranges as one means of eliminating non-representative
data, it was decided that any laboratory not reporting at least one value
within the Target Ranges would be considered an outlier. The Target Ranges
are sufficiently broad that, unless the analysis is totally out of control,
at least one value would fall within the ranges if the laboratory is perform-
ing similarly to most of the study population. Thus, if the agency is not
performing in a similar manner, Its data should not be included in the sum-
maries. Eight agencies met this criteria for outlier rejection and are
eliminated from further data summaries. The values identified as outliers
are noted in the Appendix by an asterisk(*).
Table 14 is a frequency distribution of the percent difference between
the reported and EPA values for each sample concentration. The differences
were calculated by the following:
. , . . .... .reported value - EPA value^, inn
absolute percent difference = |K EPA value '
20
-------
The frequency distribution v;as then constructed and appears below. It should
be noted that the "All Samples" line is not an average of the numbers appearing
above it, but is the distribution resulting when all the data is examined
together, regardless of the concentration.
TABLE 14. N02 ABSOLUTE PERCENT DIFFERENCE
Cone. 1
Cone. 2
Cone. 3
Cone. 4
Cone. 5
All Samples
No.
91
92
91
91
90
455
Min.
.40
.00
.00
.12
.10
.00
10%
1.2
0.5
1.0
0.6
0.5
0.7
30%
4.8
1.7
1.7
1.6
1.6
1.7
50%
8.7
2.4
3.2
2.5
2.4
3.2
70%
12.7
4.8
5.1
4.4
4.4
6.5
90%
28.2
11.4
11.4
8.4
15.3
15.1
Max.
78.2
287.4
104.8
45.3
86.4
287.4
Mean
12.7
7.8
6.2
4.5
6.7
7.7
Table 14 is very useful for laboratories trying to determine their
performance relative to the other participants. For example, only 10% of
the labs reporting results for Concentration 1 had a percent difference of
1.2% or less while 50% of the laboratories reported a percent difference of
8.7% or less for the same concentration. The Table also indicates that the
average percent difference for all laboratories for all samples was 7.7%.
All data received (with exception of the previously omitted data) were
grouped according to concentration. The results for each sample value are
listed in the Appendix in increasing concentration. The summary statistics
which appear in Table 15 and at the top of each listing in the Appendix do
not iie-lude the 8 data sets identified as outliers.
21
-------
Table 15 lists summary statistics based on the reported data. The
"outlier removed" columns contain data on which the outlier criterion was
applied. All data greater than +2 standard deviation were removed. It
should be noted that the variation in the number of samples from concentration
to concentration is due to laboratory accidents and damaged samples.
TABLE 15. STUDY 0676 N02 SURVEY STATISTICS (yg/ml)
lo.
True- Value
fean
ledian
tenge
5td. Dev.
loef. Var.
>kewness
p
\ccuracy
Concentration 1
all
data
91
.25
.23
.24
.38
.04
18.71
.01
-6.35
outlier
removed
88
.25
.23
.24
.19
.03
13.12
-.10
-6.15
Concentration 2
all
data
92
.41
.42
.41
1.36
.13
30.46
8.42
-1.09
outlier
removed
91
.41
.41
.41
.24
.03
7.72
-1.79
-1.21
Concentration 3
all
data
91
.41
.41
.41
.58
.06
13.46
5.01
-1.21
outlier
removed
90
.41
.41
^ -41
.22
.03
7.60
-1.33
-1.33
Concentration 4
all
data
91
.83
.82
.81
.51
.06
7.87
3.01
-1.45
outlier
removed
89
.83
.81
.81
.26
.04
5.01
-.01
-1.57
Concentration E
all
data
90
1.02
1.01
1.01
1.26
.15
15.08
-3.21
-.88
outlier
removed
88
1.02
1.03
1.01
.63
.09
8.62
1.38
-.73
1
Skewness - A statistic which indicates the lack of symmetry in a distribution.
For a normal distribution this value is near zero.
"Accuracy -
Median - True Value
X 100
True Value
The EPA determined True-Value and the median and mean of the study
results agree well. This is indicated by the "Accuracy" values. The study
population is normally distributed as indicated by the "Skewness". All of
the summary statistical values for the duplicate samples (concentrations 2 and
3) are nearly identical.
Each laboratory data set was plotted against the corresponding EPA data
set and the slope and intercept from the linear regression was determined.
The mean of 92 slopes was 0.961 with a standard deviation of 0.191. The mean
22
-------
intercept was 0.031 with a standard deviation of 0.106. The large standard
deviation of both the slope and intercept should be noted. This scatter in-
dicates a large amount of imprecision when the entire body of data is observed.
Analytical Method Summary
A check was made to determine if any relationship existed between the
eight laboratories whose data were omitted from the summaries and the ana-
lytical method employed (i.e. was any method responsible for most of the data
outliers). The following table (Table 16) resulted.
TABLE 16. SUMMARY OF N02 ANALYTICAL METHODS USED BY OUTLIER LABS FOR N02 0676
Method
Saltzman - Manual
Saltzman - Automated
Sodium Arsenite - Manual
Sodium Arsenite - Automated
TGS-ANSA - Manual
Others
Total No.
Using Method
5
1
65
24
1
4
No. Identified
as Outliers
1
0
4
2
1
0
Percent of
Total as
Outliers
20%
6%
8%
100%
As can be seen, no particular analytical method was responsible for an
unusually large portion of the outlier data.
To determine if a particular analytical method produced biased results,
Table 17 was developed. This Table contains the mean and standard deviation
of each sample concentration for each method used.
23
-------
TABLE 17. MEAN AND STANDARD DEVIATION OF NO- RESULTS BY ANALYTICAL METHOD (yg/tnl)
Method
Saltz.-Man.
Saltz.-Auto.
Sodi um-Arsen . -Man .
Sod i urn-Arsen . -Auto .
TGS-ANS - Auto.
Other
True- Value
Cone. 1
Mean
.214
.164
.236
.228
12.45
.228
Std. Dev.
.045
.040
.028
.121
0.25
Cone. 2
Mean
.411
.460
.426
.405
20.88
.423
Std. Dev.
.007
.157
.017
.033
0.41 . ,
Cone. 3
Mean
.408
.361
.415
.403
20.55
.425
Std. Dev.
.014
.066
.017
.041
0.41
Cone. 4
Mean
.807
.821
.828
.810
41.55
.810
Std. Dev.
.033
.076
.032
.023
0.83
Cone. 5
Mean
1.00
.99
1.01
1.00
49.83
.96
Std. Dev.
.033
.181
.064
.075
1.02
Table 17 indicates that the automated sodium arsenite and the Manual
Saltzman methods are the most precise of all the methods checked. These
methods are also the most accurate at the lower concentrations while the
manual sodium arsenite method is the most accurate at higher concentrations.
The manual Saltzman method proved to be more precise than the manual sodium
arsenite method.
Summary
The N02 Audit Survey 0676 began in June 1976. One hundred nineteen
laboratories requested to participate, of which 100 returned results for a
response rate of 84%. Foreign, EPA, State, local, and private laboratories
submitted data.
Six analytical methods were used. Sixty-five percent of the laboratories
used the manual sodium arsenite procedure and 24% of the laboratories used the
automated sodium arsenite method.
24
-------
The overall agreement between the reported values and the EPA values
was good. No bias was evident and the results were normally distributed.
The average percent difference between the EPA and reported results was only
7.7%.
Of the two analytical methods predominately used during the audit, the
automated sodium arsenite gave more precise results.
The average of all slopes (EPA vs. reported) was 0.961 with a standard
deviation of 0.191. The average intercept was 0.031 with a standard devia-
tion of 0.106.
AMBIENT CARBON MONOXIDE
Participant Characteristics
Carbon monoxide study number 047-6 began in April 1976. A total of
150 laboratories requested samples. Of those samples sent to monitoring
groups, 97 agencies responded with data (for a response rate of 65%). A
total of 238 instruments was checked. The response rate was considerably
below that of other surveys conducted during the year. Table 18 indicates
the monitoring agency type distribution.
TABLE 18. CO AGENCY DISTRIBUTION
Agencies Requesting Samples
Agencies Returning Data
Foreign
1
0
EPA
8
3
State
61
36
Local
76
57
Private
4
1
Total
150
97
Methods used to analyze the samples were grouped into 3 categories.
Table 19 lists the analytical methods used and the number of instruments
using the method.
25
-------
TABLE 19. CO ANALYTICAL METHODS USED DURING STUDY
Method
NDIR
FID
Other
Number of Instruments
Using Method
197
33
8
Most of the instruments classified as "other" use an electrochemical
method of detection.
Agencies generally employ a liberal interpretation of the analytical
method used. However, in the case of CO analyses, the methods are clearly
defined. Unknowns that could effect results and are considered part of the
analytical system are purity of zero air and accuracy of calibration standards,
Acceptable Ranges
As described in a previous section, two performance ranges were used as
one means of judging performance. The Sample Ranges and Target Ranges for
CO 0476 audit are listed in Table 20. The ranges apply to sample concentra-
tions in ascending order.
TABLE 20. CO SAMPLE AND TARGET RANGES
Sample Range
Target Range
Concentration 1
+ 14.6%
+ 14.6%
Concentration 2
+ 2.54%
± 10%
Concentration 3
+ 1.25%
+ 10%
The Sample Ranges were not determined by the method described under
Statistical Approach. Because filling of the cylinders was done so precisely,
the standard deviations of the verification analyses were small. Thus, using
the method described under Statistical Approach resulted in unreasonably small
26
-------
Sample Ranges for all concentrations. Rather than use unrealistic values,
the Sample Ranges were set at the QAB determined value plus or minus 0.5 ppm.
The Target Ranges were determined from previous CO audit results.
Larlier studies indicated that the average percent difference between re-
ported results and EPA determined values for all sample concentrations was
4- 10 percent. Thus, this value has been used for all Target Ranges with the
exception of Concentration 1. In that case the Sample Range was used because
of its larger value.
Using the above criteria, a tabulation was made of the number of agencies
reporting results within the ranges. Seventy-eight instruments (33%) were not
able to report at least one value within the Sample Ranges. Twenty-one of
these 78 instruments also reported all samples outside the Target Ranges. Thirty-
four instruments (14%) reported all results within the Sample Ranges. One
hundred nine instruments (46%) reported all results within the Target Ranges.
Data Summary
Using the Target Ranges as one means of eliminating non-representative
data, it was decided that any instrument not reporting at least one value
within the Target Ranges would be considered an outlier. The Target Ranges
are broad enough that, unless an instrument is totally out of control, at
least one value would fall within the ranges. This should occur if the in-
strument is performing similarly to most of the study population. If the
instrument is not performing in a similar manner, its data should not be
included in the summaries. Twenty-one instruments from 14 agencies met this
criteria for outlier rejection and are eliminated from further data summaries.
The values identified as outliers are indicated in the Appendix by an
asterisk (*).
27
-------
Table 21 is a frequency distribution of the percent difference between
the reported and EPA values for each sample concentration. The differences
were calculated by the following:
Absolute Percent Difference = ^portedvalu- EPA value
The frequency distribution was then constructed and appears below. It should
be noted that the "All Samples" line is not an average of the numbers appearing
above it, but is the distribution resulting when all data Is examined together
regardless of concentration.
TABLE 21. ABSOLUTE PERCENT DIFFERENCE
Concentration 1
Concentration 2
Concentration 3
All Samples
No.
216
217
217
650
Win.
0.0
0.0
0.0
0.0
10%
2.0
0.7
0.2
0.7
30%
5.2
1.7
1.2
2.0
50%
10.8
3.7
2.5
4.2
70%
17.6
5.7
4.1
7.9
90%
37.0
10.0
8.4
16.6
Max.
80.8
14.4
20.2
80.8
Mean
15.1
4.4
3.5
7.7
Table 21 is very useful for laboratories trying to determine their per-
formance relative to the other participants. For example, only 10% of the
instruments reporting results for Concentration 1 had a percent difference of
2.0% or less while 50% of the instruments reported a percent difference of
10.8% or less for the same concentration. The Table also indicates that the
average percent difference for all instruments for all samples was 7.7%.
All data received (with the exception of the previously omitted data)
were grouped according to concentration. The results for each sample value
are listed in the Appendix, in increasing concentration. The summary statistics
which appear in Table 22 and at the top of each listing in the Appendix do
not include the 21 data sets identified as outliers.
28
-------
Table 22 lists summary statistics based on the reported data. The
"outlier removed" column contains data on which the outlier criterion was
applied. All data greater than ;+ 2 standard deviation were removed. It
should be noted that the variation in the number of samples from concentra-
tion to concentration is due to laboratory accidents and damaged samples.
TABLE 22. STUDY 0476 CO SURVEY STATISTICS (ppm)
Number
True-Value
Mean
Median
Range
Std. Dev.
Coef. Var.
Skewness
2
Accuracy
Concentration 1
all
data
216
3.43
3.74
3.70
5.11
0.67
17.86
-0.02
7.87
outliers
removed
211
3.43
3.75
3.70
3.70
0.56
14.84
0.41
7.87
Concentration 2
all
data
217
20.15
20.57
20.50
5.48
1.06
5.18
0.06
1.74
outliers
removed
217
20.15
20.57
20.50
5.48
1.06
5.18
0.06
1.74
Concentration 3
all
data
217
40.60
41.28
41.00
13.31
1.87
4.54
0.57
0.99
outliers
removed
215
40.60
41.21
41.00
11.01
1.74
4.22
0.12
0.99
1
Skewness - a statistic which indicates the lack of symmetry in a
distribution. For a normal distribution this value is
near zero.
Median - True Value
"Accuracy -
X 100
True Value
The EPA determined True-Value and the median of the study results agree
well. This is indicated by the "Accuracy" values. The study population is
normally distributed as indicated by the "Skewness".
Each instrument data set was plotted against the corresponding EPA data
set, and the slope and intercept from the linear regression were determined.
29
-------
The mean of 217 slopes was 0.99 with a standard deviation of 0.05., The
mean intercept was 0.27 with a standard deviation of 0.87.
Analytical Method Summary
A check was made to determine if any relationship existed between the 21
instruments whose data were omitted from the summaries and the analytical
method employed (i.e. was any one method responsible for most of the data
outliers). The following table (Table 23) resulted.
TABLE 23. SUMMARY OF CO ANALYTICAL METHODS USED BY OUTLIER INSTRUMENTS
Method
NDIR
FID
Other
Total No.
Using Method
197
33
8
No. Identified
as Outliers
18
3
0
Percent of Total
as Outliers
9%
9%
M
As can be seen, both of the predominantly used methods contributed equally
to the outlier data.
The data in Table 23 indicate that there is no relationship between the
method used and the number of outliers found.
To determine if a particular analytical method produced results that
contained a bias, Table 24 was developed. This table contains the mean and
standard deviation of each sample concentration for each method used.
TABLE 24. MEAN AND STANDARD DEVIATION OF CO RESULTS BY ANALYTICAL METHOD (ppm)
Method
NDIR
FID
Other
True-Value
Concentration 1
Mean
3.76
3.57
4.51
Std. Dev.
.66
.62
.52
3.43
Concentration 2
Mean
20.66
19.85
21.35
Std. Dev.
.95
1.11
.77
20.15
Concentration 3
Mean
41.20
41.44
41.46
Std. Dev.
1.69
2.40
.60
40.60
30
-------
The data indicate that the FID method was the most accurate. The NDIR
method was more precise than the FID. However, the "Other" category was the
most precise of all the instruments.
Summary
The CO Audit Survey 0476 began in April 1976. One hundred fifty lab-
oratories requested samples, of which 97 returned data for a response rate of
65%. Two hundred thirty-eight instruments were checked during the audit.
Foreign, EPA, State, local, and private laboratories submitted data.
Three analytical methods were used. Eighty three percent of the instru-
ments used the NDIR principle and 14% used the FID method.
The overall agreement between the reported values and the EPA values was
good. No significant bias was evident and the results were normally distri-
buted. The average percent difference between the EPA results and reported
results was 7.7%.
The average of all slopes (EPA vs. Reported) was 0.99 with a standard
deviation of 0.05. The mean intercept was -0.27 with a standard deviation .
of 0.87.
HIGH-VOL SULFATE
Participant Characteristics
The sulfate study number 0476 began in April 1976. Out of a total of 61
laboratories requesting samples, 46 responded with data (for a response rate
of 75%). Table 25 indicates the monitoring agency type distribution.
TABLE 25. S04 AGENCY DISTRIBUTION
Agencies Requesting Samples
Agencies Returning Data
Foreign
0
0
EPA
3
2
State
35
27
Local
17
13
Private
6
4
Total
61
46
31
-------
Methods used to analyze the samples were grouped into 9 broad categories.
Results were received from laboratories using 4 methods and these are listed
in Table 26.
TABLE 26. SQ4 ANALYTICAL METHODS USED DURING STUDY
Method
Methyl thymol blue
Barium Chloride -
Barium Chloride -
- Automated
Manual
Automated
Sulfa-ver - Manual
Number Agencies
Using Method
11
18
3
14
It should be noted that some agencies tend to define the analytical methods
used in very general terms. A laboratory reporting the use of the automated
methythymol blue procedure as the method of choice may have used that pro-
cedure with various modifications. Thus, Table 26 should be interpreted as 11
laboratories used procedures approximating the automated methyl thymol blue
procedure.
Acceptable Ranges
As described in a previous section, two performance ranges were used as
one means of judging performance. The Sample Ranges and Target Ranges for
SO. 0476 are listed in Table 27. The ranges apply to sample concentrations
in ascending order.
32
-------
TABLE 27. S04 SAMPLE AND TARGET RANGES
Sample Range
Target Range
Cone. 1
+_ 5%
+ 15%
Cone. 2
+ 5%
+ 15%
Cone. 3
+ 5%
+ 15%
Cone. 4
+ 5%
+ 15%
Neither the Sample Ranges nor the Target Ranges were determined using
the methods described under Statistical Approach. Both ranges were arbitrarily
set at the values listed in Table 27. As more audits are conducted and more
data become available, the ranges will be refined.
Using the above criteria, a tabulation was made of the number of agencies
reporting results within the ranges. Twelve agencies (26%) were not able to
report at least one value wfthtn the Sample Ranges. Seven of these 26 labs
also reported all samples outside the Target Ranges. Two agencies reported all
values within the Sample Ranges while two labs reported 4 or 5 samples within
Sample Ranges. Seventeen labs (37%) reported 4 or 5 samples within the
Target Ranges.
Data Summary
Using the Target Ranges as one means of eliminating non-representative
data, it was decided that any laboratory not reporting at least one value
within the Target Ranges would be considered an outlier. The Target Ranges
are sufficiently broad that, unless a laboratory is totally out of control,
at least one value would fall within the ranges; this should occur if the
laboratory is performing similarly to most of the study population. If the
lab is not performing in a similar manner, its data should not be included in
the summaries. Seven laboratories met this criterion for outlier rejection
and are eliminated from further data summaries. The values identified as out-
liers are indicated in the Appendix by an asterisk (*).
33
-------
Table 28 is a frequency distribution of the percent difference between
the reported and EPA values for each sample concentration. The differences
were calculated by the following:
Absolute Percent Difference = |Rep°rtejjp]!a}Su; ^ ^~l x 10°
The frequency distribution was then constructed and appears below. It should
be noted that the "All Samples" line is not an average of the numbers appear-
ing above it, but is the distribution resulting when all data is examined to-
gether regardless of concentration.
TABLE 28. S04 ABSOLUTE PERCENT DIFFERENCE
Concentration 1
Concentration 2
Concentration 3
Concentration 4
All Samples *
No.
32
75
38
77
190
Min.
.00
.05
.08
.00
10%
2.2
0.8
1.2
1.6
30%
5.9
3.1
4.0
4.0
50%
12.8
5.2
6.3
7.4
70%
20.8
7.3
10.9
15.1
90%
40.3
13.0
22.9
28.2
Max.
66.4
17.8
75.9
75.9
Mean
17.4
6.6
10.8
12.6
*Does not include Concentration 1. (See Text)
Table 28 is very useful for laboratories trying to determine their per-
formance relative to other participants. For example, only 10% of the labs
reporting results for Concentration 2 had a percent difference of 2.2% or less
while 50% of the laboratories reported a percent difference of 12.8% or less
for the same concentration. The Table also indicates that the average per-
cent difference for all laboratories for all samples (except Concentration 1)
was 12.6%.
Concentration 1 was a blank and was not included in the "All Samples"
distribution. Because small concentration differences result in large percent
differences, it was felt that the numbers would unduly distort the study results.
All data received (with the exception of the previously omitted data)
were grouped according to concentration. Results for each sample value are
34
-------
listed in the Appendix in increasing concentration. The summary statistics
which appear in Table 29 and at the top of each listing in the Appendix do
not include the 7 data sets identified as outliers.
Table 29 lists summary statistics based on reported data. The "outlier
removed" column contains data on which an outlier test has been applied. All
data greater than +2 standard deviation were removed. It should be noted
that the variation in the number of samples from concentration to concentration
is due to laboratory accidents, damaged samples, and inclusion of duplicate
samples.
TABLE 29. STUDY 0476 S04 SUMMARY STATISTICS (yg/m3)
Number
True- Value
Mean
Median
Range
Std. Dev.
Coef. Var.
Skewness
Accuracy
Concentration 1
all
data
27
0.00
1.18
1.01
3.59
0.95
80.31
0.75
--
outlier
removed
26
0.00
1.09
0.88
2.56
0.83
76.46
0.53
Concentration 2
all
data
75
4.47
4.74
4.60
5.84
1.03
21.63
0.13
2.91
outlier
removed
74
4.47
4.79
4.60
4.64
0.96
20.15
0.50
3.02
Concentration 3
all
data
38
21.53
21.43
21.08
7.30
1.76
8.22
0.05
-2.09
outlier
removed
38
21.53
21.43
21.08
7.30
1.76
8.20
0.05
-2.09
Concentration 4
all
data
77
39.29
38.67
38.52
43.62
6.76
17.47
2.07
-1.96
outlier
removed
75
39.29
37.87
38.20
27.00
4.73
12.48
0.06
-2.77
1
Skewness - a statistic which indicates the lack of symmetry in a distri-
bution. For a normal distribution this value is near zero.
"Accuracy -
Median - True-Value
True-Value
X 100
With the exception of Concentration 1, the EPA determined "True-Value"
and the mean and median of the study results agree well. This is indicated
by the "Accuracy." The greatest difference is approximately 3%. The poor
agreement between the EPA and reported values for Concentration 1 is an
35
-------
indication of the minimum detectable limits of the various analytical methods
used during the audit.
Each laboratory data set was plotted against the corresponding EPA data
set, and the slope and intercept from the linear regression were
determined. The mean of 39 slopes was 1.05 with a standard deviation of
0.159. The mean intercept was -0.52 with a standard deviation of 1.15. The
contrast between the relatively small bias of 5% and small standard deviation
of the slope and large intercept and standard deviation should be noted. The
large intercept and standard deviation again indicated the difficulty in
making measurements near the lower detection limit.
Analytical Method Summary
A check was made to determine whether any relationship existed between
the seven laboratories that submitted data that were omitted from the sum-
maries and the analytical method employed (i.e., was any method(s) respon-
sible for most of the data outliers). Table 30 resulted from this check.
TABLE 30. SUMMARY OF S04 ANALYTICAL METHODS USED BY OUTLIER LABS
Method
Methyl thymol Blue - Automated
Barium Chloride - Manual
Barium Chloride - Automated
Sulfa-Ver - Manual
Total No.
Using Method
11
18
3
14
No. Identified
As Outliers
2
2
1
2
Percent of Total
As Outliers
18%
11%
33%
14%
The data in Table 30 does not indicate that the gross outliers were related
to analytical method.
To determine if a particular analytical method produced biased results,
Table 31 was developed. This Table contains the means and standard deviations
of each sample concentration for each method used.
36
-------
TABLE 31. MEAN AND STANDARD DEVIATION OF S04 RESULTS BY ANALYTICAL METHOD (yg/m3)
Method
MT3 - Auto.
Bad - Man.
Bad - Auto.
Sulfa-Ver - Man
True-Value
Concentration 1
Mean
0.62
0.91
1.72
0.76
Std. Dev.
0.45
1.03
.39
1.17
0.00
Concentration 2
Mean
5.47
4.72
5.80
4.80
Std. Dev.
0.38
1.10
1.39
1.14
4.47
Concentration 3
Mean
20.71
21.40
22.90
21.77
Std. Dev.
1.33
7.58
1.27
2.20
21.53
Concentration 4
Mean
36.89
38.91
41.21
39.26
Std. Dev.
3.83
4.24
1.09
10.60
39.29
The automated procedures were more precise than the manual methods. The
MTB automated method was more precise than the automated Bad method. The
Sulfa-Ver manual method was the most accurate, but the least precise. The
manual Bad was second to the Sulfa-Ver in accuracy.
Summary
The S04 Audit Survey 0476 began in April 1976. Sixty-one laboratories
requested samples, of which forty-six returned results for a response rate
of 75%. EPA, State, local, and private laboratories submitted data.
Four analytical methods were used. Twenty-four percent of the labor-
atories used the automated methyl thymol blue procedure, 40% the manual barium
chloride procedure, 6% the automated barium chloride procedure, and 30% the
manual Sulfa-Ver method.
The overall agreement between the reported values and the EPA values was
good. No bias was evident and the results were normally distributed. The
average percent difference between the EPA results and reported results was
12.6%. This figure does not include the blank data.
All analytical methods appeared to be approximately equal when comparing
accuracy and precision.
37
-------
The average of all slopes (EPA vs. reported) was 1.05 with a standard
deviation of 0.159. The average intercept was -0.52 with a standard deviation
of 1.15.
HIGH-VOL NITRATE
Participant Characteristics
The nitrate study number 0476 began in April 1976. A total of 61 lab-
oratories (same as 0476 sulfate participants) requested samples. Of those
samples sent to monitoring groups, 29 agencies responded with data. The filter
containing the nitrate sample also contained the sulfate sample. Thus, all
laboratories requesting samples performed sulfate analyses, but only 29 out
of 46 responding performed nitrate analysis. Table 32 indicates the monitoring
agency types.
TABLE 32. N03 AGENCY DISTRIBUTION
Agencies Requesting Samples
Agencies Returning Data
Foreign
0
0
EPA
3
1
State
35
18
Local
17
8
Private
6
2
Total
61
29
Methods used to analyze the samples were grouped into 5 broad categories.
Table 33 lists the analytical methods used and the number of respondents using
the method.
TABLE 33. N03 ANALYTICAL METHODS USED DURING STUDY
Method
Cadmium reduction -
Cadmium reduction -
Hydrazine reduction
Hydrazine reduction
Manual
Automated
- Manual
- Automated
Other
Number Agencies
Using Method
1
12
2
3
11
38
-------
It should be noted that some agencies tend to define the analytical methods
used in very general terms. A laboratory reporting the use of the automated
cadmium reduction method as the procedure of choice may have used that method
with various modifications. Thus, Table 33 should be interpreted as 12 lab-
oratories used procedures approximating the automated cadmium reduction method.
Acceptable Ranges
As described in a previous section, two performance ranges were used as
one means of judging performance. The Sample Ranges and Target Ranges for
N03 0476 are listed in Table 34. The ranges apply to sample concentrations in
ascending order.
TABLE 34. N03 SAMPLE AND TARGET RANGES
Sample Range
Target Range
Concentration 1
+ 5%
+ 15%
Concentration 2
+_ 5%
+ 15%
Concentration 3
+ 5%
+ 15%
Concentration 4
+_5%
+ 15%
Neither, the Sample Ranges nor the Target Ranges were determined using the
methods described under Statistical Approach. Both ranges were arbitrarily
set at the values listed in Table 34. As more audits are conducted and more
data become available, the ranges will be refined.
Using the above criteria, a tabulation was made of the number of agencies
reporting results within the ranges. Eight agencies (28%) were not able to
report at least one value within the Sample Ranges. Five of these 8 labs also
reported all samples outside the Target Ranges. Six agencies reported 5 or 6
samples within the Sample Ranges. Seventeen labs C59%) reported 5 or 6 sam-
ples within the Target Ranges.
Data Summary
Using the Target Ranges as one means of eliminating non-representative
data, it was decided that any laboratory not reporting at least one value
39
-------
within the Target Ranges would be considered an outlier. The Target Ranges
are sufficiently broad that, unless a laboratory is totally out of control,
at least one value would fall within the ranges; this should occur if the
laboratory is performing similarly to most of the study population. If the
lab is not performing in a similar manner, its data should not be included
in the summaries. Five laboratories met this criteria for outlier rejections
and are eliminated from further data summaries. The values identified as
outliers are indicated in the Appendix by an asterisk (*) .
Table 35 is a frequency distribution of the percent difference between
the reported and EPA values for each sample concentration. The differences
were calculated by the following:
Absolute Percent Difference = ^ EPA value x 1(XJ
The frequency distribution was then constructed and appears below. It should
be noted that the "All Samples" line is not an average of the numbers appear-
ing above, but is the data examined together regardless of the concentration.
TABLE 35. N03 ABSOLUTE PERCENT DIFFERENCE
Concentration 1*
Concentration 2
Concentration 3
Concentration 4
All Samples**
No.
47
47
23
117
Win.
0.16
0.26
0.71
0.16
10%
1.3
0.9
1.0
1.2
30%
2.9
3.1
3.3
3.2
50%
5.1
5.2
5.8
5.2
70%
7.7
9.3
7.3
8.2
90%
30.4
22.2
10.9
30.8
Max.
99.4
97.6
90.8
99.4
Mean
13.88
14.2
12.9
13.8
*Blank sample, no percent difference calculated.
**Does not include Concentration 1.
Table 35 is very useful for laboratories trying to determine their per-
formance relative to other participants. For example, only 10% of the labs
reported results for Concentration 2 that had a percent difference of 1.3%
or less while 50% of the laboratories reported a percent difference of 5.1%
40
-------
or less for the same concentration. The Table also indicates that the average
percent difference for all laboratories for all samples (except for Concen-
tration 1) was 13.8 percent.
All data received (with the exception of the previously mentioned data)
were grouped according to concentration. Results for each sample value are
listed in the Appendix in increasing concentration. The summary statistics
which appear in Table 36 and at the top of each listing in the Appendix do not
include the 5 data sets identified as outliers.
Table 36 lists summary statistics based on reported data. The "outlier
removed" column contains data on which an outlier test has been applied. All
data greater than +_ 2 standard deviations were removed. It should be noted
that the variation in the number of samples from concentration to concentra-
tion is due to laboratory accidents, damages samples, and inclusion of dupli-
cate samples.
TABLE 36. STUDY 0476 N03 SUMMARY STATISTICS (vg/m3)
Number
True-Value
Mean
Median
Range
Std. Dev.
Coef. Var.
Skewness
Accuracy
Concentration 1
all
data
22
0.00
45.69
0.12
999.95
213.15
466.52
4.07
outliers
removed
21
0.00
0.25
0.11
0.79
0.24
97.96
1.18
Concentration 2
all
data
47
6.22
6.32
6.10
11.00
1.75
27.69
0.91
-1.93
outliers
removed
43
6.22
6.27
6.10
3.87
0.68
10.88
2.19
-1.93
Concentration 3
all
data
47
7.64
7.67
7.62
13.50
2.12
27.66
0.73
-0.26
outliers
removed
43
7.64
7.61
7.62
4.14
0.78 ..
10. 30'
-0.37
-0.26
Concentration 4
all
data
23
11.27
11.01
10.80
18.90
3.00
27.26
0.85
-4.17
outliers
removed
21
11.27
10.91
10.80
4.70
0.93
8.54
-1.31
-4.17
1
Skewness - a statistic which indicates the lack of symmetry in a
distribution. For a normal distribution this value is
near zero.
"Accuracy -
Median - True Value
True-Value
x 100
41
-------
With the exception of Concentration 1, the EPA determined True-Value
agreed well with the study mean and median. This is evidenced by the Accu-
racy Figure for each concentration. As with the Sulfate Audit, the poor
agreement between the blank (Concentration 1) and reported values indicates
the minimum detectable limits of the methods used for analysis.
Each laboratory data set was plotted against the corresponding EPA data
set and the slope and intercept from the linear regression was determined.
The mean of the 24 slopes was 1.11 with a standard deviation of 0.78. The
mean intercept was 0.168 with a standard deviation of 1.70. The large mean
slope is greatly influenced by the slope of one lab of 4.58. Eighty percent
of the labs had a slope of 1.05 or less.
Analytical Method Summary
A check was made to determine whether any relationship existed between
the 5 laboratories that submitted data that was omitted from the summaries
and the analytical method employed (i.e., was any method(s) responsible for
most of the data outliers). Table 37 resulted from this check.
TABLE 37. SUMMARY OF N03 ANALYTICAL METHODS USED BY OUTLIER LABS
Method
Cadmium reduction -
Cadmium reduction -
Hydrazine reduction
Hydrazine reduction
Manual
Automated
- Manual
- Automated
Other
Total No.
Using Method
1
12
2
3
11
No. Identified
As Outliers
1
3
_ M
1
Percent of Total
As Outliers
100
25
0
0
9
The one laboratory that used the manual cadmium reduction reported all 6 sam-
ples outside the Target Ranges. Three of the 12 that used the automated
cadmium reduction method reported outlier results.
42
-------
To determine if a particular analytical method produced biased results,
Table 38 was developed. This table contains the means and standard deviations
of each sample concentration for each method used.
TABLE 38. MEAN AND STANDARD DEVIATION OF N03 RESULTS BY ANALYTICAL METHOD
Method
Cadmium Reduction - Auto.
Hydrazine Reduction - Man.
Hydrazine Reduction -Auto.
Other
True Value
Concentration 1
Mean
0.14
0.33
0.05
.24
Std. Dev.
0.11
0.34
0.05
.24
0.00
Concentration 2
Mean
6.99
6.12
4.36
6.34
6
Std. Dev.
2.00
0.04
2.35
0.89
22
Concentration 3
Mean
8.43
6.90
5.35
7.85
Std. Dev.
2.52
0.69
2.97
0.68
7.64
Concentration
Mean
12.14
9.34
8.19
11.20
Std. De'
3.56
2.17
4.86
0.67
11.27
The hydrazine reduction methods were biased low when compared to EPA.
The automated method had a greater bias than did the manual method. The
automated cadmium reduction procedure was biased high when compared to EPA.
With the exception of the blank, the "Other" category was the most accurate
and precise of the methods used.
Summary
The NO., Audit Survey 0476 began in April 1976. Sixty-one laboratories re-
quested the S04 - N03 sample combination, of which 29 returned nitrate data.
EPA, State, local and private laboratories submitted results.
Four distinct analytical methods were used with 38% of the reporting labs
using "Other" methods. Forty-one percent used the automated cadmium reduction
method.
The overall agreement between the reported values and the EPA values was
good. No bias was evident and the results were normally distributed. The
average percent difference between the EPA results and the reported results
was 13.8*.
43
-------
The average of all slopes (EPA vs. reported) was 1.11 with a standard
deviation of 0.28. The average intercept was 0.17 with a standard deviation
of 1.70.
Three of the five laboratories identified as outliers were also outliers
during the sulfate audit. This indicates that part of the problem could have
been in the extraction process.
44
-------
SECTION 6
SUMMARY
The quality assurance audits conducted by the QAB during April - September
1976, contained participants from foreign countries, federal, state, and
local agencies, and private groups. The number of participants ranged from
61 for the NO- audit to 166 for SC^. The response rate for return of data
was approximately 80%.
All audit samples, with the exception of CO, were mailed to participants
by QAB. Carbon monoxide samples were drop shipped from the vendor's facilities.
No unusual problems were encountered with any samples with the exception of
S02- Analysis of the results indicated these samples may have been improperly
analyzed.
Results show that, in general, approximately 10% of the participants
could be classified as good (withing +_ 15% of EPA value). Approximately 15%
of the results indicated an immediate need for technical assistance. The re-
mainder of the results was mixed. Many of the lab results usually contained
one or two values greater than +_ 50% of the QAB value.
Consistency seemed to be a major problem with many laboratories. The
ability to analyze within narrow limits for all samples can only be accom-
plished by a few labs. A concentrated effort directed at evaluating the
analytical system and adopting appropriate quality control procedures would
eliminate much of the imprecision and bias.
45
-------
SECTION 7
REFERENCES
1. Code of Federal Regulations (40 CFR) 50.11 Appendix A, pp. 5 -11.
2. Federal Register, Vol. 38, No. 110, June 8, 1973, pp. 15175 - 15176,
3. Duncan, A. Cheson. Quality Control and Industrial Statistics.
Richard D. Irwin, Inc., Homewood, 111., 1965. pp. 55 - 56.
46
-------
INTrR-LACPRATOPY STUDY
UN-
CAMPLF NUM3FR - 1
M 127
TRUE-VALUE 3R.27
MF^N 1C. 75
MEDIAN 31.00
DATA IN ASCENDING
) tij £
1.91 *
~? OP *
1 r C7 *
12.23
) 1 ? 00
' i f. O3
13.11 *
f fa ni
) 16.10
1C 50
1^.28 *
) i a 7n
18.31
) 18. 9C *
1 °. rn
19.71
) T.rr
71. AT *
) r?.cc
C2.39
27.11 *
} ;r^
RANrr
VARIANCE
jTD. DEV.
COEF. VAR
ORDER
O-T on
21.00
71 -1 £
21 ..70
25.00
** ^ n ?
25.20
or: -rn
25.10
25 ^ 8
2T.59
? E- - ? P
2F.2C
2E .20
2f .20
2r 50
2E.G7
2E.&9 *
zr .91
27.23
27.34
7 r, . r r
23.31
5 a.nu
73. 3G
3,57
27.35
o Q ero
28.82
7 ft n?
23.82
0 3 B'5
o Q nn
29.01
*> Q Cft
&. _* w~»
30.00
30.00
30.07
30.58
3 1.00
31.00
3 1*03
31. 1C
3 i. 31
3i.11
31.1% -.
7 it ii
A ; . £iu
3 i . 7C
31. 7C
32. TC
.1 p.rn
r.T.n,Pprp,
C.T. (LOWER)
SK FU NF ^
-------
INTCR-LADORATPRY STUDY
57E
POLLUTANT S02
^AMPIT MUMRFR 2
M 1?G
TRUE-VALUE 72.20
MEAN 64.42
MEDIAN 65. C6
DATA IN ASCENDING
4.36 *
5.93 *
C « 3Z *
2G.67
i -i r- C7 *
£-cT*»-f "
26.70
29. 6H *
a* c r, *
1 ^ 1 1 Jtt-
ft 5. DC
i ftp.nc
ft 7. 5C
U7T77
43.18 *
|i O ilO
««-;-« -i
*~ r . T
r T , c 7
n. 33 *
) CT.33
) r ft . r ?
rr.c2
) r «: . 7 r.
)
)
> - -
RflNRF
VARIANCE
STDr DEV,
CCEF. VAR
ORDER
^7^64
58. 03
58 «1Q
58. 3C
CQ it ~T
59.92
6C.CO
6C.D3
c r . ? f.
60.26
cn.50 _ -
6D.78
Gl.CO
n ,pn
C1.22
61. 3 C
C1.7C
r i - 7 r. . . _ .
E2.CC
C2.87
C2 .30
qr ^R
124. 74
11.17
17.34
4^2^92
G3.CC
63. CO
63.33
63.42
63.68
63.71
63.74
G4.ll
G ft . 50
c 5 . rn
65.13
G5.5C
£5"rc
GS.P.C
&5.8C
G5.-4
S
C. I. (UPPER
C.T.. (LOWER
SK FU MF ^-"C -
AC CU RA CY
6-^64?-
CE .11
C6 .67
67 .83
rs .or
68 .06
~ee-»i " - -
es .12
68 .4C
C9 .Cl
C9 .52
--sa^S's
re .r, °
70 .cr
7D .74
-7C .7 U
7C .33
.
MTTFR
I pf . 3 7
) G2.47
1 'tE
-9.88
71 Cn
71.30
71. 6P
72.61
73. Or
75.98
7S.9B
,.75 .98 -
_£ll£ _.
8C.4S
31.2"
81.91
.83 -.LI
9? .8^
m.i9
1 09 . 3 3 *
-
_ - - - -
._ ._ -
--
-------
INTER-LACOPATORY STUDY
57G
POL1 UTANT ^O?
«:AMPI *~ NUMRTR ~*
N 128
TRUE-VALUE 72. 2C
MEAN 57.61
MEDIAN 65.71
DATA IN ASCENDING
tof> *
4.95 *
c. 15 *
17.62 *
29.40 *
36.55 *
45.40 *
It C pp
49.18 *
P r, r n
5C.57
5 1 77
52.13
CT ^ £ip
E7.13
r4, tfP
r -,.c?
5f.SC
r H.17
r. r 7c
5 7 . 2C
5 3 . jr
r- q C T
IINTT*; - MTrRDRRAMS PFR ttlPTP MFTFR
RANGF
VARIANC
STD. DE
COEF, V
ORDER
cr n o
U i... V U U
60.26
60 .26
6C.26
cp ii o
GO. 50
60,53
en on
b U o U
60.83
c-t rn
61.22
ci ^70
61.76
C-l ,C Q
G2.0C
G2 .32
G2.50
G7 .70
G2.88
G2.93
63. TC
63.27
S7.33
^ "7 ii *>
r T r f
t» ^ . ^ ._.
93.33
E 13G.33
V. 11.63
AR. 17.27
c /i nn -- - -
G4.5D
6 4. 8C
64.85
r h f\-i
65. HC
r *i pn
u o I u
65.50
f c trn
U D * DU
65.50
65.50
65.50
G 5. PC
GS.^7
G5 67
65.75
G5.38
GG.TD
£ L .rn.
GG.ll
GG.ll
Go. 23
L o * . 'u
G6.G7
C7.CC
r. 7_rn
r.Tr
-------
INTFR-LA90RATORY STUDY
Pni 1 IJTflNT
.- - KI CROCR-AHSPER^-aiSIC-HEX ER-
SAMPt r NUMBFR - 1
N
128
1 77 ,
TRUE-VALUE 134.63
MrAN 125 1
MEDIAN
12*4.
VARIANCE 198.21
-STP »_DE V» 1^.03-
COEF. VAR. 11.25
C.I. (LOWER) 122.SG
-SK EW NE-SS t. 42-
AC CURACY -7.54
DATA IN ASCENDING ORDER
T 5 7^
11.12 *
1 ' 20 *
25.17 *
'I 5 53
75.00
8Il ->C *
85.33 *
»c an *
94.10 *
98.37
1C1.75
in?_7n *
104. CO
i r a.?? *
IP i . sr
1C K. go
11 C i Or
lie. 24
111 .75
112. 5G
1 1 7-RE
113.50
1 1 & C7
114.79
IT r . n
j-ic in
HE. 52
lie 07
117.23
1 1 T 'I 1
117.90
TIT pn
117.97
1 1 R nn
118,00
118 .^2
118,52
118.17
119.53
119.68
12P.CC
12C .00
12C.18
IT in
12C.52
120.70
121. CC
1?1 .; Q
121 .49
1 *> 1 K1^
JL 4L JL V*J
121.86
122. CO
1 ° ** 10
122.10
1 ? ? 1 R
122.50
123.11
12 3. 33
123.79
12 3. «C
123.30
124.00
124.04
124.76
1 2 5 . CC
125.PC
125.17
125. 7C
125.76
125. 76
i2c.cr
12G.7C
1 2 7 » "C
1 77 7f1
1 27 .4 C
1 "'7 P 7
1 28 .0 0
128 .38
1 ?fl T n
128 .38
1 °B 3ft
128 ,53
1 29 .0 P
129 .OH
1 23 .1 6
1 30 .r o
1 3C .1 1
. .1.31 .O.C - -
131 .CD
131 .rn
in .or
171 .20
1 T? .^ n
1 32 .F 3
.132 -»7C. .. - - _
1 3? .9 «
-1 T'7 *~ O
134 .7C
1 7F rn
13G.GO
1 TC. n q
13G .24
1 TO 7ft
138. 8G
1 7Q . 08
139.38
i 11 qp
143.00
1 43 . 0 0
143.67
143,38
145.82
149.34
149 . 34
150.70
154.53
- 154-,-7r:-
7 HI . 7 ^ *
323.18 *
-icnc-.cn-*-
i c cr . o r
-------
JL7JL
PQ1LUTANT -
.UNI:
_=.HICROGR-AH-S-PEB-CUEIC METER-
N
TRUE-VALUE
MrAN
MEDIAN
129
184.87
i7G.in
175.00
RANRF
VARIANCE
STD. DFV.
COEF. VAR.
173.R7
354.70
1 8^.83
10.63
T T MlPprpi
C.T. (LOWER)
SK FU NF *"S
ACCURACY
1 7q »«;
172.35
-1 66
-5.34
DATA IN ASCENDING OROFR
1
16
1 7
32
r" 7
114
jr-r
135
4 TO
.j a
141
1 fa *>
145
147
1ST
153
1^7
1*7
-i c 7
153
158
* r o
u. .j .*
1G1
Ir2
* r ->
.27 *
77 *.
.72 *
3~r
.52 *
-T -» J.
Toi *
.00
CT -If
of *
.60 *
cn *
.35 *
.34 *
.83
.12
. 2P
,2C
>r
.20
.52
.7E
.7^
.cr
162.48
162 .70
163.53
165
165
IfTK
16G
166
1G6
1G6
167
1F.T
167
.00
f O
»1U
.83
8Q
3
.31
5 C
.66
9.2- -
.63
.fifl
.68
1G8.0T
lCO-7r
1G8
1C?
17T
17C
L-r
17T
17f
.6C
.CC
.nc
.CO
"^r
.61
O -"l
' L> *--
1
1
1
1
1
70^
71.
7-1*
71.
T 1
S3
00
40
rt-l
172. CC
IT ^> or*
/ z. ULt
172.20
1
1
1
3_
1
i
1
_1
72.
:z-2-»-
72.
75.
73,
73.
175.
176.
- - 17&.
176.
1
1
7G.
7C.
92
92
92
16
33
82
CC
rc
S3- _ - -
rn
7C
o »
17 7aPC,
177
177
178
178
i TO
178
179
181
1 ft!
iei
183
- 1-83-
183
183
133
183
184
1£4
..!££
1 C7
188
18J2
.42
.30
.67
n n
.u u
.63
r r
.C4
_r r
.70
.00
.7'
»4 0
.4H
5 1
.24
.£-3
*r!.,i
n ^
»'-! -'-
... -
r f^
3--9Q-
191
191
191
1 *W-
*
*
191.
1 O1
j. nx *
194.
i nit
194
1 <1S
196
19S
196
1 99
2PO
*> 03
208
-2-12
212
215
217
2?G
227
352
- 1CC£
or
CO
05
2E
.Of;
6P
03
68
r. r,
.34
,e;n. -..-
-*
50
12
1?
3C
22
22
.4'
«E1
*
-
*
*
or
f-t *
ijq *
cn *.--
-------
INTER-LABORATORY STUDY
POLLUTANT - N02
SAMPLE NUMBER - 1
N 91
TRUE-VM-Ut .25
M*~fi*I ?"*
n . « . 4 *£.-.!
UAIA 1 N A^ULNUINU OKUL
.06
.1C
.11 *
.12 *
.15
.16
.16 *
.16
.17 *
I/
.17
.18
.18
.13
.2C
.2r
.21
676
RANGE
VAHIANC
STD. DE
COEF. V
R
.21
.21
.21
.21
.22
.22
.22
.22
.22
.22
.22
.22
.22
.22
.23
".23
.23
UNITS - HTCROGRAHS
.38
E .no
V. .04
AR. 13.71
.23
.23
.23
.23
.23
.23
.23
* f. *J
.23
I 74
.24
.74
,~Z4_
.24
.24
."4
PER MTLLILITER
C.T.(UrPFR)
C.I. (LOWER)
SKEW NESS
ACCURACY -6
.2%
ill
\ll
.2%
.25
.25
.25
.25
.25
.25
'.2*
125
.25 1
.24
.22
.01 -
.35
2S
.26
.26
.26
.27
.27
.27
.27
.23
.28
.23
.28
.31
.33
.35 *
.4T *
7.4T *
-------
INTER-LABORATORY STUDY
POLLUTANT - N02
SAMPLE NUM3ER -
N
TRUt VALUL
MFAN
MEDIAN
.... , nfl T. j^f, flffrMMT
2
32
.41
.42
.41
Km npnr
G7G
UNITS - MICROGRAMS
RANOF 1.36
VARIANCE .02
STC. DEV. .13
COEF. VAR. 3d. 46
o
PTR
C. '.
C.T.
SKFW
ACCU
MILLTLITER
(LTPTR)
{LOWER)
NESS T
RA CY -1
.45
.39
.09
.19 *
.2*»
.27 *
.29 *
.29
.35 *
.3G
.35
.37
.37
.37
.37
.33
.33
.38
.33
.38
.39
.39
.39
.39
.39
.39
.40
.40
.40
.40
.40
.40
.4C
T4T3 " " "
.40
.40
.40
.4C
.40
.4C
.41
.41
.41
.41
.41
.*!
.41
.41
.41
.41
.41
.41
.41
.41
"."«.
.41
. 4r
.41
.41
.41
.41
.41
.41
.41
.41
.41
.42
.42
.42
.42
.42
.42
.42
.42
.42
~ .42
.42
.42
.42
.42
.42
.43
.43
.4?
.43
.43
.47
.43
.43
.44
.44
.45
.45
.4B
.46
.48
.48
.48
.60 *
.79 *
.81 *
1 .60
7C.S8 *
-------
INTER-LABOPATORY STUDY
POLLUTANT - N02
SAMPLE NUMBER - 3
N 91
TRUE-VALUE .41
MEAN .41
HCUJAN " .41
fl*iA Irt AM.LNUJ.IMU UKUt.
.13 *
. Zb
.27 *
. 29
.30 *
3s *
.35
. 36
.36
..Sb
.37
.37
.38
.38
.33
.39
~~rq ~~~ ~
67G
RANGE
VARIANC
STD. DE
CCET. V
H
.39
.39
.39
.39
.39
.39
.39
.40
.40
.40
4G
.40
.40
.40
.4C
«4tT
.40
~ .40
UNITS - MICROGRAKS
.58
E .00
V. .06
Aft. 13. 46
.40
.Til
.41
.41
.41
.41
.41
.41
W
.41
.41"
.41
.41
.41
. 41
.41
.41
PER MILLILITE
C.T. (UPPER)
C.I. (LOWER)
SKEW NESS
AC CU RA CY
.41
1
.42
.42
.42
.42
.42
.42
.42
.4?
.4 Z
.42
.47
.42
.42
.43
it r
74 T~
R
.4?
.40
5.01
-1.21
.43
.43
.43
.44
.44
.45
HO
.46
___
Ho
ZT5
*to
.60 *
on
.35
~ "T 3 ^ *
2C .55 *
-------
IN7Er?-LACORATCPY STUDY
POLLUTANT - N<">2
SAMPLE NUMBER - 1
N 91
TSUC-VALUe .3?
MEAN .32
" MLUiAN ' .81
PRTft TN ArCENPINU ORDER
.37 *
.65 *
} .65 *
.69
.70
) .73
.73
.71
) .76
.76
.76
) *77
.77
.77
) .78
.73
.78
. . /y
.79
79
676
UNITS
RANCE
VARIANCE
STD. DEV.
COEF. VAH.
.79
.79
.79
.80
.80
.80
.80
.80
.80
.80
.80
.80
.30
.8~C
.80
.81
.31
~;~FT ~ ~
.81
.81
- MICROCRAMS
.51
.00'
.08
7.87
.81
.81
.81
.81
.81
.31
.81
.81
.81
.81
.92
.32
.32
.82
.82
.32
.32
. 82
?cr> MILLIL
C.T. (UPPER
C. I. (LOWER
SKEW NESS
AC CU RA CY
.82
.82
.82
.83
.83
.83
.83
.₯3
.83
.83
.83
.81
Is!
ill
.35
.35
ITER
) .34
) .81
3.01
.85
.85
.86
.87
.87
.87
.33
.83
.90
.91
.92
1.11
1 .17 *
1 .20
1.67 * ~
1.6* *
fl.-STTF
-------
INTER-LABORATORY STUDY
POLLUTANT - M02
SAMPLE NUMBER - 5
N 3D
iRUL VALUl 1.L.2
MEAN 1.C1
wroiAN ' i.di
CRTA IN ALL'LNUINB CKU!
.1*
.ZH
.51 *
.51 *
.75 *
.77
.80
.85 *
.85
.87
.34
.3H
.91
.9G
.97
.37
.38
.ys
.38
.38
UNI
RANCE
VAUlANCu
STH. DEV.
COEF. VAR.
R
.98
.99
.39
,9g
.99
.99
.99
.99
.99
.93
1.00
1.00
l.CO
l.DO
l.CO
1.00
l.OC
T.trrr
l.CC
l.OC
TS - MICRO GRAMS
1.2G
.OZ
.15
15.08
1.00
l.CO
1.01
1.01
1.01
1.01
1.C1
1.01
1.C1
i.ri
1.02
1.02
1.C2
1.02
1.C2
i.P2
1.02
1.C2
1.C2
r. sz
PER MILLILITE
C.T.. tUPPER J
C.I« (LOWER >
SK EU NE rS
ACCURACY
1.C2
1.03
1.03
1.03
1.03
1.03
1.03
1.03
1.01
l.Otf
1.0«l
1 .C*i
i.rif
l.Ot
1.0 t:
l.CE
l.Cr
"" i . c^r
1 .OC
1.D7
R
1 .C4
.98
-3.1?
-.88
1 .08
1.09
1 .08
1.10
1 .10
1.11
1.11
1 .IS
1 .17
1 .13
1 .20
1.25
1.38
1 .33 *
1.40
1 .97 *
7.3r *
*re,rsT *
-------
INTCR-LABORATORY STUDY
176
POLLUTANT - CO
SAMPLE NUMBER - 1
N 216
TRUE-VALUE 3.13
MEAN 3.71
"MEDIAN 3.70
UNITS -
--
RANGE
VARIANCE
STD. DEV
COEF. VAR.
PARTS PER MILLION
5.11
.15
.67
17.36
C.
C.
SH
AC
.1.
, I.
LEW
:cu
.(UPPER.)
(L OH ER )
!NESS
IRACY
3.83
3.65
7.87
* DATA TN ASCENDING ORDER
. C9 3
1.C9 3
1.55 3
z. or 3
) 2.cr; 3
Z. 26 7
2.61* 3
\ Z.bb 5
?.7C 3
J v. -y
7. rr ?
) l.CC 3
7 « c r 3
) I':!! I
-- ic 3
7. 13 7
) _; :£ ;
, rir- r
7*. 2"
> -;5r ;
'-> *- T
* "*
^. *^ -r~i *tr
< * ~*
--7 7. 1C
.10
.10
.1
.1
.1
.1
.1
.1
.1
«b
.5
.b
.5
,b
.5
.b
r
fr
.
.b
.5
«-'
.5
* ~'
.5
.r,
-
r
» -
*
.5
r
~^-
* -
"« ,
D
C
3
1
g
*
3
u
C
U
c
c
c
c
r
c
c
L
r-
u
C
C
L
r
r
C
C
-
i 7 .r n
- ^ .EC
O
3
»
*
J.
3
3
-»
-^
(5
B
U
C
C
C
c
0
u
0
u
C
u
c
r.
n
c
0
r
o
c
c
c
L.
O
c
c
' *
r-
V
r*
c i . ? * *
C tj . "*
^':"
- 4.2F
1
4
t
4
1
1
1
1
1
H
4
1
u
1
4
1
4
I
5
r
f-
f-
i-
-
-
r
~->
"T
a
.27
.29
. ib *
r o
. - '-
.5C
.5C
.61
. r-'
7C
.75 *
.73 *
.73 *
.7°!
.3P *
.sr
.en
.31 *
.y.
. cn *
.cr *
o .
r~ **
pr
.:- *
. 2"" *
'.7-
.IS
*
. '!' *"
". CU *
,5H *"
-------
_. INTCR-LA BOP ATC RY S TUDY. 4.76
POLLUTANT - CO
SAMFLF NUMBER - 2 . .
_.N 217 RANGF
TRUE-VALUE 20.15 VARIANC
_MEAN _ 2C.57 . ._ STD* DE
MEDIAN 20.50 COEF. V
DATA IN ASCENDING ORDER
b.bt, * 13.8" 'rD«J~
15.57 * 19.80 2r.31
17. 5S 13. 8C 2C.32
i '.83 1*3.81 2C.35
18. IP 12.82
13. IP 1*3 «2C
13.50 1°.90
13.51 13.30
13.7C 2r .rc
1 3 . 7 1 / 1 . L" L.
18. 8C 2r.CC
18.37 2'"'.CJL'
1 i r<~ T r r
-L ^> i^ Lv £.- * L L
1 "3 L*"1 Zr . '" C
i^.c0 2r .rc
is.u': zr.ur
i ° . cr T r *~
- j.. » _ L,
^ ~ 1* ., / ' » » i.
I'ri.L; j.: .LL
^ r» r "* ~ r >~' ^
X v- *,* £»l_r
i-.r.- I-.IL
lrt.2? zr .ic
1 °» 2 * - -«. ** .1 2
ir»4r, ^ . i r
^ » *-> .. ^. i.
-i*«^-. t«_i
j. .~ *_ w. i
i r- ~ * ' T~ -* -
1. * _
:-.:- -.7- :-.::
* a. ^
- i " . B r -
UNITS _-__P ARTS PEB_M ILLI ON
5 qa
C 1.13 C.T. (LOWER) 20.42
V. 1.C6 SKEUNCSS .06
AR. 5.ia ACCURACY 1.74
2C.35 21. CO "1 .00 M..81* ?1.80
20.40,21.00 21.00 21. 70. 21. 8?
^Ju.4L ^l.l'U 21. OL1 21. MJ 21. yO
2o.4C 21. OC 21.00 ?1.73 ?1 . 9H
Z0.4D 21. 00 21. CO 21.75 ?1.9C
Zu.4L 21.00 22. CC
2C.^C 21.00 22.00
ZO.^C Zl.DU 'l'l»V
2C.5C 21.00 22. IE
L Li . M; ,:'± . U V /i! . Z' '
2 G. ^L 21.00 22.2H
20.^^ 21.00 22. 'ir
IL.5C 71.00 ??.20
I'C.FL n.cir: 2^.3r
2C.rC 71. CC ??.4T
L:L. --L n.uu z::.tn
C^r "1 r ^ '>"> c; P
L. i L I j_ .^.31
7 f cr "iii T> r n
^-i-**-' ^.x«j.> *M(..^)L
t. L. ' L ;i . '. i' ;*.' . L-"
o ,- . trr ^ i -» c; -7^ on
^^-. i-< ^^.., ...O-
^ c. rc 21 . i rv '" . 3r
2 L. 5C ^l.^P 2T . C"
i *j . ^L ^ j. . ^ j_ ., L
^! L.* JC " 71 . 5 n ?""" . Cr *
, T --i r r -» r »i
i_L-»t t^..-- ...i_"*
i'..:i n.- -.^. *
*. *- ^" G ^j-s^r n^»u*
- 1 -»." 1 r ^ Ofi -» r» -^
ji_ -- i, < -. » 4 <_ j
^ L . u .'. . r ^M . i A
^ ... ~L n ." r. ""4 . 3" *
» -7 -ir~'~^ "^r*(~io^.
i. L * «.*. »«.«
..:.-»L ri.-- -a.-, *
*-X»'«- *. JL r~ . -
14 1 . *
T.-T nr> *
-------
INTER-LABORATCRY STUDY 476
POLLUTANT - CO
SAMPLE NUMBER -
M
TRUE-VALUE 4C
.MEAN . . 11
MEDIAN HI
UNITS -
3
217 RANGE 1
.60 VARIANCE
28 STD.. DEV.
.CO COEF. VAR.
PARTS PER M
3.31
3.51
1.51
ILLION
C T tUPPrR> ftl ^T
C.T. (LOWER) 11.03
SK FU Mr SS 17
ACCURACY .99
DATA IN ASCENDING ORDER
2 T . 9C *
21.ZP *
7T.G1 *
?4.ei *
35. SC
7 1 . 5 C *
37. CP
T?.CC
3^.39
37. 8r
TS.CC
7S.C2
7 S . 44
i- *> t-
~* P 3r
?n.cr
"» o pr
T °» 1C
-7 ^ - -i
-:*'-
" ".."."
V ' <>
*
« t
-3. -£±. ?>*.'- - - -
-
40. CC 40.C2 4C.75 4ij.
4U.UL' 4F.C? 4t'.!iC 4C.
4C.OC 4P.1 5 4C .8C 4 U.
40. CC 4L.ia 4C.80 4 U.
HH ,^C 4 L.
4P.22 41.
4f'.25 41.
4n.3C 41.
m ..si: 41.
nr .3C 4 i.
4D.4T 41.
4T' .4C 41.
4i".rr 41.
4 r . r r 41.
fc n C C 41.
4r .5 C 4 ^.
4r.r. r. 41.
4> .5 1: 41.
u1" .rr 41.
t: r . E r. 4 ^ .
4^.5" 41.
'f .r. :, 41.
f . z r m.
47' .0 C 41.
s - ' C '» _ .
4 '" . " i ;+ 1 .
', *" 4 > ' 1 ^ .
; ' . *. 4 j..
'- * ~* ^
s . .
8C 1 » l.TC
RC 41.70
BI 4 l.TC
81 41.70
^ 41.7C
rc
CO
CC
ru
CC
CC
CC
CC
1 . U
re
P3
, i_
"C
^4
"'SI
"C
;?"" ~"
3
*:>- - -
---
T " %? .Bl 4~ .0"
41. 8 c 42.83 «n.cn
41. 8 C 13. CC 43. lr
41. 8 C 43.CO 43.2?
*J1.8L' 43.00 1-'.2V
11. 8C 43. 4P
41. 9C 47. 5C
41. 9 C 43. 5 P
4i. nc 43. sr
42 c r ^4 . cr
42 .r 0 44 . LG
nz.nr 44. c^
HZ .r C 44 .C"
42. cc 44. rr
t2.rp 44.cn
"2 .CC 44 .C"'
42. PC 44. 2r
*t ^ . i ' ? ^^.r*1
42. T- 4^.L4
^2.2- ir.c- *
^Sr -!4^-
*?.*c 1r..'-
'2.4" £»".2^ *
t* "" ^ *^ J^O'~*^'jL
'2.r^ "1.7°*
._ . r ' . *
^ ff . 2 ? *
j r o r* O JL.
o?V ^r~ ^f"
-------
INTER-LABORATORY S
POLLUTANT - SOI
SAMPLE
K
TRUE-V
MEAN
MEDIAN
NUMBER - 0
38
ALUE 21.53
21.13
21.08
DATA IN ASCENDING
.11 *
7.1% * "
11.60 *
IrTsTO1-
17.76
10*15 *
18.72
20.00
20.00
IiiPJL___--JLl£_______.
UNITS - MICR06RAMS PER CUBIC METER
RANGE
VARIANCE
STO. DFV-
COEF. VAR.
ORDER
20.03
20.10
20.32
20.10
20.10
ZD.5Z
20.69
20.71
20.79
7.30
3.10
1.7K
8.22
2O. 91
'20.93
20.93
.23
21.26
2 1. 52
21.70
2 1. 71
22.00
22.12
C.I. (UPPER 1 Z^.flS
C.I. ILOWER) 2C.87
ACCURACY -2.09
22.20 21.78
ZI> .UU
22 .88 27.90 *
23 wl 0 12 . UtI *
23 -12 312.00 *
23 >! '»
23 .20
. 2M JOG
21 ^C
-------
INTER-LABORATORY STUDY
176
POLLUTANT - S04
UNITS
- MICROGRAMS PER CUBIC
METER
SAMPLE NUMBER - 2
N
TRUE- VALUE
MEAN
MEDIAN
77
39.29
38.67
38.52
DATA IN ASCENDING
\ 13.
'IT
X J*
18.
26.
ZT.
) 28.
26.
! 29.
30.
28 *
Z8 *
20 *
00 *
63
*
14 *
ll *
40
45
30. ox
1 30.78
31.
31.
us
30
i -5Z.1U
33.00
ORDER
33
y
34
34
34
34
35
35
36
36
35
36
36
36
36
37
37
RANGE
43.62
VARIANCE
STD. DEV.
COEF.
.12
.08
ZO
.32
.80
.00
6
.00
DO
.06
18
.40
.54
.55
bl
.80
.20
.50
VAR.
37
3 7
37
r
37
37
37
3 b
38
15.61
6.76
17.47
»
»
9
*
3 8.
38.
3 8
33
39
39
33
39
*
9
9
9
9
9
50
50
50
53
63
70
78
20
52
5Z
64
83
00
If
21
60
62
80
C,
. T. (UPPER) 10.
C.I. (LOWER) 37.
SK EU NT SS 2
AC CU RA CY
39
40
4D
4D
40
ID
*HJ
4D
41
41
41
41
41
41
41
42
42
.90
Ou
.00
0 0
.04
.05
.10
11
.49
.00
DO
jD 0
.10
.70
70
JO
.DC
.55
-1.
42
45
45
48
50
57
18
16
07
96
.70
.23
.bO
.27
.U6
.30
.su
.60
*
*
67.68
69.12
7s
78
534
624
UU
.60
.UU
.00
ft
*
x
*
-------
INTER-LABORATORY STUDY
476
POLLUTANT - SOI
UNITS - MICR06RAHS PER CUBIC METER
SAMPLE NUMBER - 3
N
TRUE-
MEAN
-VALUE
MEDIAN
1
1
.
.
.
27
CO
18
01
RANGE
VARIANCE
STD. DEV.
COEF. VAR.
3.59
.90
.95
80.34
C.
C.
SK
AC
I.
I.
EW
cu
(UPPER
(LOWER
NESS
RACY
>
)
1.54
.82
.75
.00
DATA IN ASCENDING ORDER
.01
.30 *
,75
1.32
2.50
. U£.
.02
51
1.00*
1.01
1 .
3.60
36.CO *
.20
.62
-.6-5-
1.20
2 .28
2.40
2.47
.30
.72
1.20
-------
INTER-LABORATORY STUDY
176
POLLUTANT - S04
SAMPLE NUMBER - 9
N 75
TRUE-VALUE 4.47
MEAN 4.74
MEDIAN 4.60
^ ' n«T« Tit *ferun-rut*
* ' .03 *
x .60 *
en *
. uu *
.89 *
. ' - - .. qn *
1.60
ZD *
^ 2.80
2.90 *
2.99
1 T ntt
\ o.D4
3.21
.3. 1b
) 3.60
3.60
3.70
) 3./D
UNITS - MICROGR
RANGE
VARIANCE
STD. DEV.
COEF. VAR.
nnnF~R
3.75
T3
3.90
4.03
4.08
* .AU
4 .1 6
4.25
4.27
4 .34
4.34
4.36
4.37
4.37
5.84
1.05
1.03
21.63
4.39
4. 4 A
4.47
50
4.54
4.55
4.57
4! 59
4.60
4.61
4.64
4.79
4.80
4.80
1m BO
4.90
4.90
AMS PER CUBIC METER
C.I. CUPPER J
C.I. (LOWER)
SK EU ME SS
AC CU RA CY
4 .90
4^7
3 .0 2
5 .04
lu
5 .18
& »zu
5 .25
5 .ZS
5.30
5 .40
5 .43
5.48
5 .5D
5.50
5.64
4.98
4.51
2.91
5.70
b .UU
6.00 *
6 .24
6.27
6.27
6.48
b.48
6.60 *
6.60
7.. 00
7.UU
7.44
9.20 *
9.60 *
135. OD *
208.80 *
-------
INTER-LABORATORY
POLLUTANT
- NO 3
SAMPLE NUMBER -
N
TD HC* M ftl HP"
MEAN
DATA TN A«:
2.
2.
3.
7.
10.
1 t
11.
CFNHTN
08 *
07
60
62 *
80
40
STUDY
0
23
O"T
01
0 ORDER
10
10
10
10
10
10
476
RANGE
STD.
.43
.58
.59
.60
.60
.62
UNITS
DEV.
10
- MICRO GRAMS
18.90
9m
3.00
.77
10.79
1 0.80
10
1C
11
.87
.90
.10
PER CUBIC
C.I. (UPPER
GT fl fil/T~n
SKEW NESS
11 JL'3
11 .38
11 J4O
11 .56 *
12 Ji q
12 .23
METER
1 12.
i 1
.
24
7 n ,,
85
-1-T
12.50
21
4fi
84
.50
.En *
.10 *
-------
INTER-LABORATORY STUDY
476
POLLUTANT - N03
UNITS - MICRDGRAMS PER CUBIC METER
SAMPLE NUMBER - 2
N
TRUE" j AI UF
MEAN
47
6m->r>
6.32
*1U
RANGE
' VAlTTflMPF1
STD. DEV.
CUEr. V AR .
11. CD
T nr
O* LIU
1.75
I * D J
C.I. CUPPER 1
e_i T , /I />!_! f"D * M .
* ! IL UP LR 1
SKEUNESS
AC CU HA CY
6.82
r* o *^
.31
-A .3 J
DATA IN ASCENDING ORDER
.05 *
.05 *
.54*
1.23 *
1.40
1.40
2.10 *
2.66 *
5.30
5.30
5.52
5.71
5.75
5.76
5.88
5.90
5.30
5.90
5.90
5.90
5.93
5.96
5.96
5.97
5.97
6.02
6.04
6.07
6.09
6.10
6.10
6.12
6.15
6.18
6.21
6.30
E.30
E .33
E .34
E .39
E .44
E .58
E -B2
E.70
E.70
E .78
E .BO
E .88
7.40
8.11
12.30
13.50 *
13. 7f>*
27.60 *
29.40*
-------
INTER-LABORATORY STUDY
176
POLLUTANT - N03
UNITS - MICROGRAMS PER CUBIC METER
SAMPLE NUMBER - 3
N
TDfjr VAI UF'
MEAN
ui-n-r«tt
22
nn
45.69
RANGE
VI RT AMP r
STD. DEV.
r»nvf %» «»
999.95
It r ft "T ^ -1 -7
213.15
teC-C Ct9
C.T. CUPPER)
GT f 1 nil T~O 1
. !> IU UN C.K I
SKEyNESS
134.76
43 . 30
4.07
DATA TN ASPFNnTNG ORDER
.05
.05
.06
.06
.08
.08
.09
.10
.10
.10
.11
.13
.20
.22
.25
.39
.41
.52
.57
.7n *
.75
.84
i.ao *
1000.00
-------
INTER-LABORATORY STUDY
476
POLLUTANT - N03
UNITS - MICROGRAMS PER CUBIC METER
SAMPLE NUMBER - 9
N
TO MIT \l fit fir-
MEAN
flE'DlAN »-
WiTA TM Arrr
47
7 flft
7.67
'fin TNG nnnrn
RANGE
STO. DEV.
COCF . VAR.
13.50
2.12
ZT.oo
C.I. CUPPER)
r T 1 1 nun? i
SKEW NESS
8.28
.73
.06 *
.06 £
1.49 *
1.55 *
1.60
1.60
2.72 *
2.83 *
5.20
6.20
6.30
6.80
6.93
7.00
7.04
7.11
7.15
7.20
7.20
7.31
7.31
7.40
7.45
7.53
7.56
7.60
7.60
7.62
7.71
7.73
7.76
7.77
7.80
7.86
7.91
7.94
7.98
B .03
B .84
B .04
B .20
B.28
B .42
B.SQ
9.00
9.03
14.90
15. in
15.40
33.00
34. an
*
*
*
*
6.40
7.40
7.80
B .90
-------
TECHNICAL REPORT DATA
(Please read Inunctions on the reverse before completing)
1 REPORT NO.
2.
3. RECIPIENT'S ACCESSION-NO.
4 TITLE AND SUBTITLE
SUMMARY OF AUDIT PERFORMANCE MEASUREMENT OF S09, NO,
CO, SULFATE, NITRATE - 1976 ^ '
5. REPORT DATE
November 1977
6. PERFORMING ORGANIZATION CODE
7 AUTHOR(S)
Steven M. Bromberg, Robert L. Lampe, Berne I. Bennett
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
U.S. Environmental Protection Agency
Environmental Monitoring and Support Laboratory
Quality Assurance Branch, MD-77
Research Triangle Park, North Carolina 27711
10. PROGRAM ELEMENT NO.
1HD621
11. CONTRACT/GRANT NO,
12,. SPONSORING AGENCY NAME A.ND ADDRESS
Environmental Monitoring and Support Laboratory
Office of Research and Development
U. S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
EPA/ORD
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The report summarizes the results of the quality assurance audits for 1976.
Pollutants for which audits were conducted and results reported are S0?, N02» CO,
sulfate, and nitrate. The operation of the EPA audit program is also Described.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
COSATI Field/Group
Air Pollution
Quality Control
1 DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
75
20 SECURITY CLASS (This page)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
------- |