EPA/600/4-85/080
December 1985
EPA METHOD STUDY 32
METHOD 450.1 - TOTAL ORGANIC HALEDES (TOX)
by
Carol H. Tate, Bruce M. Chow, Robert R. Clark,
Nancy E. Grams, Lewis K. Hashimoto
James M. Montgomery, Consulting Engineers, Inc.
Pasadena, California 91101-7009
Contract Number 68-03-3163
Project Officer
Terence M. Grady
Environmental Monitoring and Support Laboratory
Cincinnati, Ohio 45268
ENVIRONMENTAL MONITORING AND SUPPORT LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
-------
TECHNICAL REPORT DATA
(I'lfBSc read fnilruvtwns on the reirrir hr/orr comi>lrtii)g)
1. REPORT NO. 2
EPA/600/4-85/080
3. RECIPIENT S ACCESSION NO
PB8 6 13 6 5 3 8 /AS
4. TITLE AND SUBTITLE
EPA Method Study 32:
Method 450.1 - Total Organic Halldes (TOX)
5- REPORT DATE
December 1985
6. PERFORMING ORGANIZATION CODE
7. AUTHORIS)
Tate, C. H.; Clark, R. R.; Chow, B. M.; Grams, N. E.;
and Hashimoto, L. K.
B. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADORESS
James M. Montgomery Consulting Engineers, Inc.
555 East Walnut Street
Pasadena, CA 91109-7009
10. PROGRAM ELEMENT NO.
5CBL6
11. contract/grant no.
68-03-3163
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Monitoring and Support Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, OH 45268
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
EPA 600/6
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This report describes the interlaboratory method study that was performed to
evaluate interim Method 450.1 for total organic halides (TOX). In the method, a
measured volume of water is passed through two columns in series, each containing 40
mg of activated charcoal. Organic halides (OX) present in the water are adsorbed on
to the charcoal which is washed to eliminate trapped inorganic halides. The contents
of the columns are then pyrolzed converting the halides to titratable species that
are measured microcoulometrically. In this study, three water matrices, reagent
water, ground water, and surface water, were spiked at six concentrations with a
solution containing a combination of four model compounds. These were lindane,
bromoform, pentachlorophenol, and tetrachloroethene. A chlorinated drinking water
diluted to four concentrations with distilled water were also analyzed.
Ten laboratories participated in the study. Data obtained were analyzed using a
standardized package known as Interlaboratory Method Validation Study (IMVS), which
is designed to implement the recommendations of ASTM Standard D-2777. The IMVS
package includes rejection of outliers; estimation of mean recovery as a measure of
bias; estimation of single-analyst and overall precision, and tests for effects of
water type on these parameters.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATi Field/Group
18. DISTRIBUTION STATEMENT
Distribute to Public
19 SECURITY CLASS (This Rcporti
Unclassified
21 NO OF PAGES
79
20 SECURITY CLASS i Tiiii patei
Unclassified
22. ooirp
EPA Form 2220-1 (R«v. 4-77) previous edition is obsolete
i
-------
DISCLAIMER
The information in this document has been funded wholly by the United States
Environmental Protection Agency under Contract Number 68-03-3163.
Mention of trade names or commercial products does not constitute
endorsement or recommendation for use. It has been subject to the Agency's
peer and administrative review, and it has been approved for publication as a
USEPA document.
ii
-------
FOREWORD
Environmental measurements are required to determine the quality of ambient
waters and the character of waste effluents. The Environmental Monitoring
and Support Laboratory-Cincinnati conducts research to:
o Develop and evaluate techniques to measure the presence and
concentration of physical, chemical, and radiological pollutants in
water, wastewater, bottom sediments, and solid wastes.
o Investigate methods for the concentration, recovery, and
identification of viruses, bacteria, and other microorganisms in
water and determine the responses of aquatic organisms to water
quality.
o Develop and operate an Agency-wide quality assurance program to
assure standardization and quality control of systems for
monitoring water and wastewater.
This publication reports the results of a study of the carbon adsorption
microcoulometric titration method for determining the concentration of
organically bound halides in water. Federal agencies, states, municipalities,
universities, private laboratories, and industry should find this evaluative study
of vital importance in their efforts in monitoring and controlling halogenated
organic pollution in the environment.
Robert L. Booth
Director, EMSL - Cincinnati
iii
-------
ABSTRACT
This report describes the interlaboratory method study that was performed to
evaluate Interim Method 450.1 for total organic halides (TOX). In the method,
a measured volume of water is passed through two columns in series, each
containing 40 rag of activated charcoal. Organic halides (OX) present in the
water are adsorbed onto the charcoal which is washed to eliminate trapped
inorganic halides. The contents of the columns are then pyrolyzed converting
the halides to titratable species that are measured raicrocoulometrically. In
this study, three water matrices; reagent water, groundwater, and surface
water, were spiked at six concentrations with a solution containing a
combination of four model compounds; lindane, bromoform, pentachlorophenol,
and tetrachloroethene. A chlorinated drinking water diluted to four
concentrations with distilled water were also analyzed.
Ten laboratories participated in the study. Data obtained were analyzed using
EPA's computerized statistical program known as Interlaboratory Method
Validation Study (IMVS), which is designed to implement the recommendations
of ASTM Standard D-2777-77. The IMVS package includes rejection of
outliers; estimation of mean recovery as a measure of bias; estimation of
single-analyst and overall precision; and tests for effects of water type on
three parameters.
This report was submitted in fulfillment of Contract No. 68-03-3163 by James
M. Montgomery, Consulting Engineers, Inc. It covers work performed from
September 1982 to June 1985.
iv
-------
CONTENTS
Foreword iii
Abstract iv
List of Tables vi
Acknowledgements vii
1. Introduction 1
2. Conclusions 2
3. Recommendations 5
4. Description of the Study 6
Test Design 6
Selection of Laboratories 7
Phase I -Performance Evaluation 8
Phase II - Interlaboratory Method Study 8
5. Statistical Treatment of Data 17
Rejection of Outliers 17
Statistical Summaries 20
Statement of Method Bias Z3
Statement of Method Precision Z4
Comparison of Bias and Precision Across
Water Types 26
6. Results and Discussion 32
Outliers 32
Statistical Summary 32
Statements of Bias and Precision 34
Effect of Water Types 37
References 38
Appendices
A. Interim Method 450.1 39
B. Instructions to Participating Laboratories 56
C. Sample Preparation Data 65
v
-------
TABLES
Number Page
1 Regression Equations for Precision and Bias 3
2 Theoretical TOX Concentration of Spiking
Solution - 10
3 Theoretical TOX Concentrations of Paired Study
Samples 11
4 Summary of Verity and Homogeneity Data 12
5 Stability Test Summary 15
6 Statistical Summary for TOX Analyses by Water Type ... 33
7 Regression Equations for Precision and Bias 35
C-l Raw Data for Reagent Water 66
C-2 Raw Data for Surface Water 67
C-3 Raw Data for Groundwater 68
C-4 Raw Data for Chlorinated Drinking Water 69
C-5 Effect of Water Type on TOX Analysis 70
vi
-------
ACKNOWLEDGEMENTS
The authors wish to thank the following persons for their help with the project:
Albert R. Trussell, Eric Crofts, John Wanek, Shine Pearson, and Harry Morrow.
vii
-------
SECTION 1
INTRODUCTION
This project determined, by interlaboratory method study, the precision and
bias of EPA Method 450.1 for total organic halide (TOX), a surrogate
parameter used to measure the amount of halogen-containing organic material
in a water sample. The method detects organically bound bromine, chlorine,
and iodine but is not sensitive to organic fluorine compounds; nor does it
provide structural information for any of the compounds comprising the TOX.
The halogenated organics measured by TOX are usually indicative of
anthropogenic contamination. Compounds which contribute to TOX include
organic cleaning solvents such as trichloroethene and 1,1,1-trichloroethane;
chlorination products such as trihalomethanes (THMs), chlorophenols, certain
pesticides and herbicides. In addition, TOX includes high molecular weight
chlorinated compounds which generally comprise a higher percentage of
organic halides (OX) than do THMs in chlorinated finished drinking water.
1
-------
SECTION 2
CONCLUSIONS
The object of the study was to characterize the performance of Method 450.1
in terms of precision, bias, and the effect of water types on precision and bias.
The regression equations shown in Table 1 are the result of statistical analyses
of 220 analytical values. Twelve points were rejected as laboratory outliers,
10 were rejected by Cochran's test and seven as individual outliers. Rejected
data totaled 12.2 percent of the 220 analytical values.
The bias of the method was estimated by comparing mean recoveries to true
TOX values at six concentration levels between 38.7 and 441.1 yjg/L. The
average recovery calculated from the regression equations was 86.5 percent,
with the actual recoveries ranging from 83.5 percent to 117.2 percent. The
highest recoveries occurred at the lowest concentration levels.
The overall standard deviation, S, was not significantly dependent on recovery,
X, as indicated by slopes of regression equations which ranged from -0.0128 to
0.0374. The intercepts ranged from 6.4 to 14.1 and closely approximated the
actual S values obtained for the low, medium and high concentration ranges:
2.9 to 14.4 ug/L, 5.7 to 14.1 pg/L and 10.4 to 15.4 pg/L, respectively. Percent
relative standard deviations for low, medium and high Youden pair samples
were 7.2 to 31.8 percent, 3.2 to 6.6 percent, and 3.0 to 4.4 percent,
respectively.
The single-analyst precision Sr, indicating the precision associated with a
single laboratory also showed little dependence on recovery, X. The slopes of
the regressions for Sr ranged from -0.0092 to 0.0033 with intercepts ranging
2
-------
TABLE 1. REGRESSION EQUATIONS FOR PRECISION AND BIAS
Water Type X S S
x_
Reagent X = 0.807C + 14.1 S =-0.0128 X + 14.2 Sr = -0.0092 X + 12.7
Surface X = 0.894C + 7.14 S = 0.0374 X + 2.68 Sr = -0.0109 X + 6.14
Ground X = 0.896C + 6.38 S = 0.0280 X + 3.40 Sr = 0.0033 X + 5.48
Chlorinated
Drinking
Water — S = 0.0946 X - 9.22 S, = 0.1037 X - 0.1014
X = Mean recovery (bias) as !g/L
S = Overall precision as !g/L
Sr = Single-analyst precision as !g/L
C = True value as !g/L
3
-------
from 5.48 to 12.7. Single-analyst precision values actually obtained for low,
middle and high concentrations ranged from 5.7 to 12.3 Ug/L, 4.5 to 9.3 vig/L,
and 9.4 to 12.0 yg/L, respectively. Single-analyst relative standard deviations
for low, middle and upper concentrations were 11.8 to 23.7 percent, 2.2 to
3.9 percent and 2.5 to 3.4 percent, respectively.
No regression equation for TOX recovery from chlorinated drinking water was
calculated due to the absence of a true concentration value for that sample
type. Regressions calculated for overall S and Sr against mean analyzed value,
X, yielded an equation for S with a strongly negative intercept. The equation
generated did not accurately predict the S values obtained from the study data
and was considered invalid. The most probable cause for this was considered
to be the use of four rather than six concentration levels for calculation of the
regression. Individual S values for the four water samples ranged from 3.1 to
7.9 Ug/L as for TOX concentrations between 63.8 ug/L and 83.6 yg/L. For TOX
concentrations in the range of 137.8 to 178.5 yg/L, S values ranged from 12.7
to 29.6]jg/L. Single-analyst precision ranged from 4.5 at the low
concentration range to 22.8 pg/L for the higher concentration range.
Statistical comparisons of the effect of water type were performed. No
significant effect of water type on bias or precision of Method 450.1 was
observed.
4
-------
SECTION 3
RECOMMENDATIONS
Method 450.1 is recommended for the analysis of Total Organic Halide (TOX)
in drinking, ground, and surface waters. The method bias and precision are
acceptable and there are no significant matrix effects with the waters listed
above. The "Interim" designation should be removed from the current title of
the method.
o To ensure more consistent overall performance of the method,
several ambiguous points that became apparent during Phase I of
the study, should be clarified in future versions of the method.
o Additional research should be conducted on performance of the
method when analyzing chlorinated drinking water supplies.
o In order to avoid TOX carry over from one sample to the next, the
sample reservoir should be rinsed with two 100 ml volumes of
reagent water before adding another sample.
o Users of this method must take precautions to avoid contamination
of samples and the analytical system, especially when analyzing
samples expected to have low TOX concentrations. The potential
for contamination from contact with the fingers can be greatly
reduced by following the recommendations found in Section 5.4.2
of the method.
5
-------
SECTION 4
DESCRIPTION OF THE STUDY
TEST DESIGN
The overall experimental design was governed by Youden's original non-
replicate design for collaborative evaluation of precision and bias for
analytical methods (1). The design is recommended by ASTM in Standard
Practice D2777-77 "Determination of Precision and Bias of Methods of
Committee D-19 on Water" (2). According to Youden's plan, paired samples
containing analytes at similar but distinct concentrations are analyzed
collaboratively by a group of laboratories. In this study, sample pairs were
prepared at low, medium, and high TOX concentrations within the analytical
range of the method. Samples were prepared as full volume aqueous solutions
for shipment to 10 laboratories. No sample preparation was required of the
analysts at the participating laboratories.
A summary of the test design according to Youden's design is given below:
1. Three Youden pairs were used for the analyses in reagent, ground and
surface waters. A chlorinated drinking water was diluted to four
concentrations constituting two Youden pairs.
2. The three Youden pairs were spread across the working range of the
method.
3. Analyses for TOX were performed by 10 laboratories according to
Method 450.1.
6
-------
4. Each sample was analyzed in duplicate as required by the method. The
means of duplicate results were analyzed statistically using the IMVS
statistical package.
Selection of Laboratories
The initial contacts with laboratories were made using an instrument
placement list obtained from the manufacturer. Willing laboratories with
equipment to perform TOX analyses according to Method 450.1 were asked to
submit bids. Performance evaluation samples were sent to 16 laboratories
including two EPA laboratories. Each of the 16 laboratories was evaluated on
accuracy, ability to strictly adhere to the method, timeliness, and ability to
follow reporting procedures. Based on these criteria, the eight paid and two
unpaid EPA laboratories listed below were selected for participation.
Aquatec Environmental Services
75 Green Mountain Drive
South Burlington, Vermont 05401
Harmon Engineering & Testing
1550 Pumphrey Avenue
Auburn, Alabama 36830
MMTL Analytical Services
206 South Keene Street
Columbia, Missouri 65201
Spectrix Corporation
3911 Fondren, Suite 100
Houston, Texas 77063-5821
U.S. EPA
Office of Drinking Water
Technical Support Division
26 W. St. Clair Street
Cincinnati, Ohio 45268
Gascoyne Laboratories
27 South Gay Street
Baltimore, Maryland 21202
McKesson Environmentail Services
6363 Clark Avenue
Dublin, California 94568
Radian Corporation
8501 MoPac Boulevard
Austin, Texas 78766-0948
Timber Products
884 Blacklawn Road
Conyers, Georgia 30207
U.S. EPA
Water Engineering Research
Laboratory
26 W. St. Clair Street
Cincinnati, Ohio 45268
7
-------
Phase I - Performance Evaluation
Sixteen laboratories were provided with performance evaluation samples. The
Phase I samples were prepared by spiking reagent water with sufficient lindane
to give a TOX concentration of 220.9 lJg/L as chloride. In order to present an
analytical challenge, the sample also contained 25.6 ug/L of inorganic chloride
as NaCl. Each laboratory received a single sample, a copy of Interim
Method 450.1 (Appendix A), and a statement of conditions (Appendix B).
Laboratories were required to submit the final results, blank values, standard
and recovery results, raw data forms, and the signed statement of conditions
within 10 days of sample receipt. The results were collected and evaluated for
accuracy and completeness as described above. Laboratories with analytical
problems were contacted to discuss and clarify analytical procedures.
Phase II - Interlaboratory Method Study
Water types used in this study were: (1) reagent water from the James M.
Montgomery laboratory, (2) surface water from Azusa, (3) groundwater from
Rubio Canyon (collected prior to chlorination), and (4) chlorinated drinking
water from the Garvey Reservoir, operated by the Metropolitan Water District
of Southern California. The waters were collected in clean five-gallon glass
carboys. Nitric acid was added to the carboys prior to sampling as an inhibitor
of possible biological degradation. The final pH of the water in each carboy
was 1.0. Sodium sulfite was added to the distilled, ground and surface waters
as a precaution against residual chlorine.
Triplicate subsamples were taken from each carboy of ground and surface
water and analyzed for TOX background contamination. If TOX in excess of
5 ug/L was detected, the entire sample was heated, purged with nitrogen for 24
hours and cooled to room temperature. The procedure was repeated until re-
analysis indicated no detectable TOX concentration.
8
-------
The reagent, ground and surface waters were then spiked with a mixture of
four halogenated organic compounds in methanol. Table 2 gives the
composition of the spiking solution and the corresponding theoretical TOX
concentration contributed by each component. The four compounds were
chosen to represent differing properties including stability, volatility, polarity
and type of halogenation. All four compounds are priority pollutants and
indicators of industrial contamination. The water in each carboy was mixed by
magnetic stirrer for one minute after addition of the spiking solution. Using a
sampling tap, thirty-two 250 mL amber glass bottles were filled headspace-
free. The TOX concentrations of the chlorinated drinking water were adjusted
by dilution with reagent water. The target concentrations for these samples
were near those of the low and medium Youden pairs of the spiked samples.
The final TOX concentrations of the paired study samples are shown in
Table 3.
Each bottle was labeled with a three letter code which identified the filling
order, water type, and TOX concentration. Samples were stored in the dark at
4°C.
To establish verity, homogeneity and stability of the study samples, three
bottles were randomly chosen from the beginning and end of the filling
sequence and analyzed for TOX by Method 450.1. Verity was established if the
mean result for each sample was 90 to 110 percent of the known TOX
concentration. Table 4 indicates that all samples met the verity criterion.
The samples were considered stable if "time zero" analyses were 90 to
110 percent of the later "time one" analyses. For this study, stability was
established for periods of up to 140 days (Table 5).
Following verification of the study samples by two independent referee
laboratories, packages were prepared for shipment to the participating
laboratories. Sets of 18 spiked and four unspiked (chlorinated drinking water)
9
-------
TABLE 2. THEORETICAL TOX CONCENTRATION OF SPIKING SOLUTION
Compound
Lindane
Pentachlorophenol
Bromoform
Tetrachloroethene
Concentration
kgAjL)
2.736
3.026
4.184
2.294
TOX
(yg/uL as CI)
2.001
2.014
1.761
1.962
10
-------
TABLE 3. THEORETICAL TOX CONCENTRATIONS OF
PAIRED STUDY SAMPLES
Pair
Low Level
Mid Level
High Level
Concentration
Member (lJg/L)
A1 38.69
B1 193.4
CI 386.9
Concentration
Member (pg/L)
A2 54.17
B2 243.7
C2 441.1
11
-------
TABLE 4. SUMMARY OF VERITY AND HOMOGENEITY DATA
Verity
Homogeneity
Sample
ID
You den
Pair
Mean
True Analyzed Calculated
Value Value Percent F-
Vig/L as Cl~ yg/L as Cl~ Recovery Value
Chlorinated
Drinking Water
B-2
N/A 175.9 N/A 0.58
B
Surface
Water
A-2
54.17 52.65 97.2 0.13
Surface
Water
C-2
441.1 408.9 92.7 0.39
D
Reagent
Water
B-l
193.4 176.0 91.0 3.1
Chlorinated
Drinking Water
A-2
N/A 84.9 N/A 0.71
Ground
Water
B-2
243.7 231.0 94.8 1.37
Reagent
Water
C-2
441.1 396.8 90.0 0.13
H
Surface
Water
C-l
386.9 360.5 93.2 0
12
-------
TABLE 4 (Cont'd)
Verity
Homogeneity
Sample
ID
Youden
Pair
Mean
True Analyzed Calculated
Value Value Percent F-
Vlg/L as Cl~ vig/L as CI" Recovery Value
Ground
Water
A-2
54.17 51.56 95.2 0.46
Surface
Water
A-l
38.69
41.83 108.1 5.28
K
Surface
Water
B-2
243.7 232.0
95.2 0.22
Reagent
Water
A-2
54.17 51.35 94.8 0
M
Chlorinated
Drinking Water
B-l
N/A
150.4
N/A
1.66
N
Reagent
Water
C-l
386.9 352.7
91.2 2.05
Chlorinated
Drinking Water
A-l
N/A
69.59 N/A 0.25
Ground
Water
B-l
193.4 179.3
92.7 2.67
Reagent
Water
A-l
38.69 38.3
99.0 0.29
13
-------
TABLE 4 (Cont'd)
Verity
Homogeneity
Sample
ID
Y ouden
Pair
Mean
True Analyzed Calculated
Value Value Percent F-
)jg/L as CI" pg/L as Cl~ Recovery Value
R
Ground
Water
C-l
386.9
359.6
92.9 14.08
Ground
Water
A-l
38.69
41.2
106.5 0.54
Surface
Water
B-l
193.4
177.0
91.5 0.07
U
Reagent
Water
B-2
243.7
222.3
91.2 0.48
V
Ground
Water
C-l
441.1
405.5
91.9 0.01
S = Significant difference at 0.01 level
N/A = Not applicable
NS = Difference is not significant at 0.01 level; degrees of
freedom are (1,4) Critical F = 21.2
True Value = Theoretical TOX concentration
Mean Value = Measured TOX concentration
14
-------
TABLE 5. STABILITY TEST SUMMARY
Sample
Code
Time
Zero
Mean
yg/L as CI"
Time
One
Mean
pg/L as CI"
Distilled Water
Q
L
D
U
N
G
33.12
48.50
195.83
250.78
381.84
435.07
30.95
48.99
181.15
234.80
360.96
414.82
Surface Water
J
B
T
R
H
C
35.36
49.64
194.82
252.10
386.12
441.63
38.35
52.27
207.70
239.24
378.13
430.54
Groundwater
S
I
P
F
K
V
33.17
46.46
197.40
255.23
394.73
447.67
34.04
47.19
197.60
253.95
379.38
420.83
Chlorinated
Drinking Water
M
117.68
110.88
15
-------
samples were packed with refreezable gel packs into coolers. Each of the 10
shipments also contained a cover letter, chain-of-custody forms, TOX report
forms, and a set of clarifications to Method 450.1. The pacakges were shipped
early in the week by a major overnight courier. The laboratories were
contacted to confirm delivery of intact samples.
16
-------
SECTION 5
STATISTICAL TREATMENT OF DATA
This interlaboratory study was conducted to obtain information about the bias
and precision associated with measurements by Interim Method 450.1 "Total
Organic Halides". The statistical techniques employed in the data reduction
process are similar to the techniques recommended in the ASTM Standard
Practice D2777-77.
The algorithms required to perform the statistical analyses have been
integrated by USEPA into a system of computer programs referred to as
Interlaboratory Method Validation Study (IMVS) (3). The analyses performed
by IMVS include:
o several tests for rejection of laboratory and individual outliers;
o summary statistics for mean recovery (accuracy), overall and
single-analyst standard deviations;
o determination of the linear relationship between mean recovery
and concentration level;
o determination of the linear relationship between the precision
statistics and mean recovery, and
o testing for the effect of water type on bias and precision.
17
-------
REJECTION OF OUTLIERS
Outlying data points will occur in any set of data collected during an
interlaboratory test program. It is important to identify and remove these
data points because they can lead to summary statistics which are not
representative of the general behavior of the method. However, some erratic
behavior in the data may be directly related to some facet of the method
under study. Therefore, seemingly unreliable data points should not be
removed indiscriminantly, and any points that awe removed should be clearly
identified since further investigation of the analytical conditions related to
the outliers might be of value. Data rejected as outliers for this study as a
result of Cochran's test, Youden's laboratory ranking procedure, or the test for
individual outliers have been identified by the symbol in the raw data
tables (Appendix C).
Cochran's Test
Traditionally, only single determinations are required by analytical methods
under study. Because, however, duplicate measurements are required by
Method 450.1 and the IMVS software package does not allow entry of mean of
duplicate values, it was necessary to calculate the mean of duplicate
determinations. Prior to screening for outlying laboratories, an additional
level of outlier testing was imposed in the form of Cochran's Test.
According to Cochran, if a standard deviation of one pair of duplicates is
significantly different from the other standard deviations in that
concentration group, then that pair belongs to a separate population and can
be rejected subject to the significance level criteria below. The critiera for
the rejection was 0.01 significance level for Cochran's "C" given by the
formula shown below:
18
-------
c _ largest
Sj = standard deviation of the ith pair of duplicates
Data rejected using Cochran's test are denoted by "CO" following the results
in raw data tables (Appendix C).
Youden's Laboratory Ranking Procedure
Youden's (1) ranking test for outlying laboratories was applied separately to
data from each water type used in this study. Each laboratory ranking test
was performed at the five percent level of significance.
The Youden laboratory ranking procedure requires a complete set of data from
every laboratory within a given water type. Missing data from laboratory i for
sample type j were replaced by the following procedure. Letting denote
the reported measurement from laboratory i for water type j and
concentration level Cfc, it is assumed that
where and yj are fixed parameters which determine the effect of water type
j, Lj is the systematic error due to laboratory i and e is the random within
laboratory error. Taking natural logarithms, it follows that
which is a linear regression model with dependent variable &n Xjj^ and inde-
pendent variable Jin C^.
Hn Xjjk = &n3j + Yj£nCk+JlnLi+Jln
19
-------
The natural logarithms of the individual laboratory's data were regressed
against the natural logarithms of the true concentration levels for each water
type. The predicted values S,n were obtained from the regression
equation, and any missing values for Xjjfc were estimated by Xjj^ = exp(£n
Xijk)- (For complete details of this procedure see Reference 4).
If the ranking test rejected a laboratory for a specific water type, then all of
the laboratory's data for that water type were rejected as outliers. The
rejected values were excluded from all the remaining statistical analyses. In
values created to fill in the missing data were excluded from further
statistical analyses.
Tests for Individual Outliers
The data remaining after the laboratory ranking procedure were grouped by
water type. For each sample type, the data were divided into subsets defined
by the concentration levels used in the study. Next, the test for individual
outliers constructed by Thompson (5), and suggested in the ASTM Standard
Practice D2777-77, was applied to the data using a five percent significance
level. If an individual data point was rejected based on this test, it was
removed from the subset, and the test was repeated using the remaining data
in the subset. This process was continued until no additional data could be
rejected.
STATISTICAL SUMMARIES
Several summary statistics were calculated using the data retained after the
outlier rejection tests were performed. These summary statistics include: the
number of retained data points (n), the mean recovery (X), bias as a percent
relative error, the absolute overall standard deviation (S), the overall relative
standard deviation (% RSD), the absolute single-analyst standard
20
-------
deviation (Sr), and the single-analyst relative standard deviation (% RSD-SA).
The formulas used to calculate these statistics are presented below where Xj,
X£. Xn denote the values of the n retained data points for each
concentration level.
Mean recovery (X):
- 1 n
X = - £ Xj
n . ,
Accuracy as a % Relative Error:
%RE = X - True Value x 100
True Value
Overall Standard Deviation:
s * ;xi - *2
v 1=1
and
Percent Relative Overall Standard Deviation:
%RSD = — x 100
X
The overall standard deviation, S, estimates the precision of measurements
generated by a group of laboratories in the interlaboratory study. However, a
measure of how well an individual analyst can expect to perform in his/her
own laboratory is another important measure of precision. This single-analyst
precision, denoted by Sr, was estimated for each Youden pair by
m
Sr "/ 2(m - l)£jDi '
21
-------
where
m = the number of complete Youden pair observations remaining
after outliers have been removed,
Dj = the difference between the observations in the Youden pair
and
Dj = average of the Dj values
The single-analyst relative standard deviation was calculated by
%RSD-SA=^5 x 100
X*
where X* is the average of the two mean recovery statistics corresponding to
the two concentration levels defining the particular Youden pair.
These summary statistics provide detailed information on the bias and
precision of the data obtained for each concentration level. One objective of
the statistical analysis of the data is to summarize the information about bias
and precision which is contained in the statistics.
Frequently a systematic relationship exists between the mean recovery (X) and
the true concentration level (C) of the analyte in the sample. In addition,
there are often systematic relationships between the precision statistics (S and
Sr) and the mean recovery (X). Usually these systematic relationships can be
adequately approximated by a linear relationship (i.e., by a straight line).
Once these straight lines are established, they can be used to conveniently
summarize the behavior of the method within a water type and can be used to
obtain estimates of the bias and precision at any concentration level within
the concentration range studied.
22
-------
STATEMENT OF METHOD BIAS
The bias of the method is characterized by the relationship of the mean
recovery (X) to the theoretical TOX concentration (C) in the water sample. In
order to obtain a mathematical expression for this relationship, a regression
line of the form
X = a + b • C (1)
was fitted to the data by regression techniques.
Often the true concentration values in a collaborative study cover a wide
range. In such cases, the mean recovery statistics associated with the larger
concentration values tend to dominate the fitted regression line producing
relatively larger errors in the estimates of mean recovery at the lower
concentration values. To reduce the overriding effects of high concentration,
a weighted least-squares technique was used to fit the mean recovery data to
the true concentration values. The weighted least-squares technique was
performed by dividing both sides of Equation (1) by C resulting in Equation (2)
— • = a • i + b (2)
C U
The (X/C) values were regressed against the (1/C) values using ordinary least
squares to obtain estimates for the values of a and b. (This is equivalent to
performing a weighted least-squares with weights w = l/C^; see reference (3),
page 108 for details). Equation (2) can easily be converted to the desired
relationship given by Equation (1). The intercept (b) from Equation (2)
becomes the slope (b) for Equation (1) and slope (a) from Equation (2) becomes
the intercept (a) for Equation (1). Equation (1) can be used to calculate the
percent recovery over the applicable range of concentrations used in the
study.
23
-------
The percent recovery is given by
Percent Recovery =
a + b • C
x 100 =
itb
100
(3)
If the absolute value of the ratio (a/C) is small relative to the slope (b) for
concentration in the low end of the range of concentration levels used in the
study, then the percent recovery can be approximate by b x 100. For example,
suppose the true concentration values range from 25pg/L to 515yg/L, the
fitted line is given by X = 0.20 + (0.85 • C). The percent recovery would be
approximated by (0.85) x 100 = 85 percent over the specified range of 25 yg/L
to 515 pg/L'
If the absolute value of the ratio (a/C) is not small relative to the slope (b),
then the percent recovery depends upon the true concentration (C), and it
must be evaluated at each concentration value within the specified range.
STATEMENT OF METHOD PRECISION
The precision of the method is characterized by the relationships between
precision statistics (S and Sr) and mean recovery (X). In order to obtain a
mathematical expression for these relationships, regression lines of the form
S = d + e • X (4)
and
Sr = f + g • X* (5)
were fitted to the data.
As discussed previously with respect to bias, the values of X and X* often vary
over a wide range. In such cases the standard deviation statistics associated
with the larger mean recovery values will dominate the regression lines. This
24
-------
will produce relatively larger errors in the estimates of S and Sr at lower mean
recovery values. Therefore, a weighted least squares technique was used to
establish the values of the parameters d, e, f, and g in Equations (4) and (5).
The weighted least squares technique was performed by dividing both sides of
Equation (4) by X resulting in Equation (6).
S , 1 ,/>
— = d • — + e (6)
X X
and dividing both sides of Equation (5) by X* resulting in Equation (7)
S_i = f • 4" + 8 (7)
X*
X*
The (S/X) values were regressed against the (1/X) values and the (Sr/X*)
values were regressed against the (1/X*) values using ordinary least squares to
obtain estimates for the parameters d, e, f, and g.
Equations (4) and (5) were obtained from Equations (6) and (7) in a manner
similar to that discussed for mean recovery. The slope (d) for Equation (6) is
the intercept (d) for Equation (4), and the intercept (e) for Equation (6) is the
slope (e) for Equation (4). Similarly, the slope (f) for Equation (7) is the
intercept for Equation (5), and the intercept (g) for Equation (7) is the slope (g)
for Equation (5).
Given Equations (4) and (5), the percent relative overall standard deviation and
the percent relative single-analyst standard deviation are
% RSD =
_d_
X
+ e
x 100
(8)
and
% RSD-SA =
— + g
X
100
(9)
25
-------
respectively. If absolute value of the ratio (d/X) is small relative to the slope
(e), then the percent relative overall standard deviation can be approximated
by (e x 100) over the applicable range of mean recovery values. Similarly if
the absolute value of the ratio (f/X*) is small relative to the slope (g), then
the percent relative single-analyst standard deviation can be approximated by
(g x 100) over the applicable range of mean recovery values.
If the ratios (d/X) and (f/X*) are not small relative to the slopes (e) and (f),
then the percent relative standard deviations depend upon the values of the
mean recovery statistics X and X*, and they should be evaluated separately
for each value of X and X*.
COMPARISON OF BIAS AND PRECISION ACROSS WATER TYPES
It is possible that the bias and precision of the Interim Method 450.1 depend
upon the water being analyzed. The summary statistics X, S and Sr are
calculated separately for each concentration level within each water type.
They can be compared across water types in order to obtain information about
the effects of water type on bias and precision. However, the use of these
summary statistics in this manner has several disadvantages. First, it is
cumbersome since there are six mean recovery statistics (X) (six
concentrations) six precision statistics (S) and six precision statistics (Sr)
calculated for each compound. Comparison of these statistics across
concentration levels and across water types becomes unwieldly. Second, the
statistical properties of this type of comparison procedure are difficult to
determine.
An alternative approach, described in detail in Reference (6), has been
developed to test for the effects of water type. This alternative approach is
based on the concept of summarizing the average effect of water type across
concentration levels rather than studying the local effects at each
26
-------
concentration level. If significant differences are established by this
alternative technique, then the summary statistics can be used for further
local analysis.
The test for the effect of water type is based on the following statistical
model. If denotes the measurement reported by laboratory i, for water
type j, and concentration level k, then
^ijk = " Ck J • Lj • ejjk i = n (10)
j = 1
k = 1,2
The model components and Yj are fixed parameters which determine the
effect of water type j on the behavior of the observed measurements (Xjjk).
The parameter Ck is the true value associated with concentration level k. The
model component Lj is a random factor which accounts for the systematic
error associated with laboratory i. The model component ejjk is the random
factor which accounts for the within-laboratory error.
The model is designed to approximate the globed behavior of the data. The
multiplicative structure was chosen because of two important properties.
First, it allows for a possible curvilinear relationship between the data (Xjjk)
and the true concentration level Ck through the use of the exponent Yj on Ck.
This makes the model more flexible in comparison to straight line models.
Second, as will be seen below, there is an inherent increasing relationship
between the variability in the data and the concentration level Ck in this
model. This property is important because it is typical of interlaboratory data
collected under conditions where the true concentration levels vary widely.
Bias is related directly to the mean recovery or expected value of the
measurements (Xjjk). The expected value for the data modeled by
Equation (10) is
E(Xjjk) = nj • CkYj • E(Lj • ejjk) (11)
27
-------
Precision is related to the variability in the measurements (Xjjk). The
variance of the data modeled by Equation (10) is
Var(Xijk) = [gj CjJj
which is an increasing function of Ck.
Var (Li • eijk) (12)
The bias and precision of the method for TOX analysis depend upon water
types through Equations (11) and (12) and the parameters (gj) and (yj). If the
(3j) and (yj) vary with j (i.e., vary across water type), then the bias and
precision of the method also vary across water type.
In order to determine if these parameters do vary across water type and to
compare their values, they must be estimated from the laboratory data using
regression techniques. Equation (10) represents the basic model. However,
taking natural logarithms of both sides of Equation (10), the following straight
line regression model is obtained
&n — J,n 3j + yj Ck + &n Lj + 5,n £jjk (13)
which can be analyzed using standard linear model analysis techniques. The
parameter 5,n is the intercept and yj the slope of the regression line
associated with water type j. It is assumed that &n Lj is normally distributed
2
with mean 0 and variance0l> an^ that J,n is normally distributed with
2
mean 0 and varianceae, and that the (5,n Lj) and (in £ijk) terms are inde-
pendent.
Based on Equation (13) the comparison of water types reduces to the
comparison of straight lines. The reagent water is viewed as a control, and
the remaining lines (for surface or groundwater) are compared directly to the
line for the reagent water.
28
-------
Using the data on the log-log scale and regression techniques, the parameters
£n8j (and hence Bj) andyj can be estimated. The estimates are then used to
test the null hypothesis that there is no effect due to water type. The formal
null and alternative statistical hypotheses Hq and are given by
Hq: Jlngj - Jlngi = 0 andyj - yj = 0 for j = 2,3
versus
H^: £,n [3j - Jlng j = 0 and/orYj - Yj = 0 for j = 2,3
The test of null hypothesis is Hq against the alternative hypothesis is based
on F-statistic derived from standard linear model theory. The probability of
obtaining a value of an F-statistic as large as the value which was actually
observed, (F OBS), denoted by P(F>F OBS), is calculated under the assumption
that Hq is true. The null hypothesis Ho is rejected in favor of if P(F>F
OBS) is less than 0.05.
If Hq is rejected, then some linear combination of the differences fl,n£j - in
3^ and Yj-Yj is statistically different from zero. However, this does not
guarantee there will be a statistically significant direct effect attributable to
any specific water type since the overall F test can be overly sensitive to
minor systematic effects common to water types. The effect due to a specific
water type is judged to be statistically significant only if one of the
differences Otng^ - inftj) and/or (y^ *s statistically different from zero.
This is determined by checking the simultaneous 95 percent confidence
intervals which are constructed for each of these differences. Each true
difference can be stated to lie within its respective confidence interval with
95 percent confidence. If zero is contained within the confidence interval,
then there is no evidence that the corresponding difference is signficantly
different from zero.
29
-------
If at least one of the confidence intervals for the differences (&n 3j ~ $^) or
(y. - Yj) fails to include zero, then the statistical signifiance of the effect due
to water type has been established. However, establishment of a statistically
significant effect due to water type does not necessarily mean that the effect
is of practical importance. Practical importance is related to the size and
interpretation of the difference.
The interpretation of the differences involves comparing the mean recovery
and standard deviation of the (Xyk) data for each water type to the mean
recovery and standard deviation obtained for reagent water. These
comparisons are made on a relative basis. The mean recovery for water type j
is given by Equation (11). The mean recovery for water type j is compared to
that for reagent water (j=l) on a relative basis by
(yj - yi) (14)
E(Xijk) = Bj E(Lj • ejjk) _ 3j
E(^ilk) Yl g
01 Ck E(Lj • £iik) P1
(The ratio of the standard deviations would be equivalent to Equation (14), and
therefore the interpretation of the effect on precision is the same as that for
the effect on mean recovery).
The ratio in Equation (14) is a measure of the relative difference in mean
recovery between water type j and reagent water. It is comprised of two parts
(a) 8i/01 > which is independent of the true concentration level (i.e., the
(y. _ Y.)
constant bias), and (b) J which depends upon the true concentration
level (i.e. the concentration dependent bias). If (yj - Yl) *s zero, then the
relative difference in mean recovery is just Bj/Bl which is independent of
concentration level Ck- It can then be stated that the mean recovery from
water type j is (0j/31) x 100 percent of the mean recovery from the reagent
water. If (yj - yi) is not zero, then the mean recovery from water type j is
((Bj/3l) Ck ^ ^ x 100 percent of that from reagent water, and therefore
depends upon the true concentration level Ck»
30
-------
In order to illustrate these points, consider the following example which
compares at least five water types. Suppose that a significant F-value has
been obtained, and the confidence intervals for all the differences contain
zero except for water type 5. For water type 5, the point estimate for
(^n - ^n 3^) is -0.38 and the confidence interval for (&n 8^ -&n 3^) is (-0.69,
-0.07). The points estimate for (Y5 - Yj) is 0.07, and the confidence interval
for (Y5 -Yj) is (-0.04, 0.18). In this case a statistically signficant effect due to
water type has been established which involves only water type 5. The
practical significance of this effect is judged by considering Equation (14).
The ratio of mean recoveries from water type 5 and reagent water is given by
E(Xi5k) = B5 ^ (Y5 - Yj) (15)
E(xnk)
and the ratio of the standard deviations is given by
/Var(Xiik) = ck ^5"^)
V Var(Xjik)
(16)
Since the confidence interval for (Y5 - Yj) contains zero this difference is
assumed to be insignificant and is set to zero. Therefore Equations (15) and
(16) reduce to 3 5/6 The point estimate for (&n 3 5 - &n 3 j) was -0.38.
Therefore, the point estimate for Bg/3j is 0.68, and the mean recovery from
water type 5 is estimated to be 68 percent of the mean recovery from reagent
water. Similarly the standard deviation for the data for water type 5 is
estimated to be 68 percent of the standard deviation for the reagent water.
Since the 95 percent confidence interval for (&n3g -^n3^) was (-0.69, -0.07),
any value in the interval (0.50, 0.93) is a reasonable estimate for 3 5/6 2., and
the mean recovery (standard deviation) for water type 5 can be claimed to be
from 50 percent to 93 percent of the mean recovery (standard deviation) for
reagent water. The practical significance of the effect due to water type 5
would depend upon the importance of a mean recovery (standard deviation)
observed for reagent water.
31
-------
SECTION 6
RESULTS AND DISCUSSION
OUTLIERS
Three levels of outlier testing were performed on the data collected in this
study. The Cochran's test for homogeneity of variance was conducted in
addition to the tests that are part of the IMVS package. By performing
Cochran's test both before and after the standard IMVS package, it was
determined that Cochran's testing prior to the IMVS run did not affect the
results of the laboratory ranking or individual outlier tests. In the raw data
Tables C-l through C-4, data rejected by Cochran's test are indicated by
"*CO" in the data columns. Data rejected as individual outliers are marked
in the data columns. Laboratory outliers are designated by am asterisk in
the "lab rejected" column. The three tests used in this study rejected 28 of
220 data points, or 12.7 percent of the data.
STATISTICAL SUMMARY
After outliers were rejected using the tests above, retained data were
statistically analyzed. A summary of those analyses are presented in Table 6.
The statistical parameters included are:
a. n, Number of data points: the number of laboratories that
submitted data which were not outliers.
b. T, True value, yg/L: theoretical TOX concentration of the sample
based on weighed amounts of compounds added.
32
-------
TABLE 6. STATISTICAL SUMMARY FOR TOX ANALYSES BY WATER TYPE
CHLORINATED
REAGENT SURFACE GROUND DRINKING
WATER WATER WATER WATER
LOW YOUDEN PAIR
1
2
1
2
1
2
1
2
NUMBER OF DATA POINTS (n)
9
10
9
8
8
9
8
10
TRUE CONC (T) UG/L
38.7
54 . 2
38.7
54 . 2
38.7
54.2
-
-
MEAN RECOVERY (X)
45.3
58 . 3
40.2
58.7
40.7
55.6
63.8
8 3.6
BIAS (%REL ERROR)
17.17
7.57
3 .79
8.41
5.26
2.64
-
-
OVERALL STD DEV (S)
14.4
12.3
2 . 9
8 . 0
2 . 9
8 . 1
3 . 1
7 . 9
OVERALL REL STD DEV, %
31.85
21.18
7.24
13.56
7.18
14.61
4 . 9
9 . 5
SINGLE STD DEV, (Sr)
12
. 3
6 . 7
5 . 7
4
. 5
ANALYST REL DEV, %
2 3 .
67
13.52
11.80
6
. 1
MEDIUM YOUDEN PAIR
3
4
3
4
3
4
3
4
NUMBER OF DATA POINTS (n)
9
10
8
9
8
8
9
10
TRUE CONC (T) UG/L
193.4
2 4 3 .7
193.4
2 4 3 .7
193.4
2 4 3.7
-
-
MEAN RECOVERY (X)
161.6
211.9
178.8
229 .8
178.9
223 .2
137.8
178.5
BIAS (%REL ERROR)
-16.45
-13.04
-7.57
-5.70
-7.48
-8.42
-
-
OVERALL STD DEV (S)
7 . 1
14.1
5 . 7
12.8
8 . 8
8 . 1
12.7
29.6
OVERALL REL STD DEV, "s
4.38
6.66
3.21
5.59
4.93
3.62
9 . 2
16.6
SINGLE STD DEV, (Sr)
9
. 3
7 . 9
4 . 5
22
. 8
ANALYST REL DEV, %
4 .
98
3.87
2.24
14
. 4
HIGH YOUDEN PAIR
5
6
5
6
5
6
NUMBER OF DATA POINTS (n)
10
8
7
8
9
8
TRUE CONC (T) UG/L
386.7
441.1
386.7
441.1
386.7
441.1
MEAN RECOVERY (X)
33 2.0
378 .2
349 .0
392 .2
352.0
404.2
BIAS (%REL ERROR)
-14.16
-14.2 7
-9 .76
-11.09
-8.97
-8.37
OVERALL STD DEV (S)
12.0
14.3
15.4
14.9
10.4
12.8
OVERALL REL STD DEV, %
3.61
3.79
4.41
3.81
2.96
3.16
SINGLE STD DEV, (Sr)
12
. 0
10.9
9 . 4
ANALYST REL DEV, %
3 .
39
2.93
2.49
-------
c. X, Mean recovery, Jjg/L: overall mean of the retained data.
d. Bias as percent relative error: difference between mean recovery
and true value as a percentage of the true value.
e. S, Overall standard deviation, gg/L: standard deviation of Xj values.
f. Overall relative standard deviation, percent: standard deviation of
Xj values as a percentage of X .
g. Sr, single-analyst standard deviation, |Jg/L: standard deviation for a
given Youden pair (as described in the Statistical Treatment
Section).
h. Single-analyst relative standard deviation, percentage: Sr as a
percentage of X .
STATEMENTS OF BIAS AND PRECISION
The regression equations found in Table 7 indicate the performance that can
be expected from routine use of Method 450.1. The bias of the method is
estimated from the recovery regression equations in the first column of
Table 7. In most studies of this kind, the slope of the equation is used to
estimate the percent recovery for each concentration level and water type.
The validity of using the slope to estimate recovery depends upon the
magnitude of the intercept. If the intercept is considered negligible or at
least insignificant, the slope may be considered a reliable estimator of
recovery. If the intercept is large compared to the slope, the intercept itself
contributes the most to estimates of recovery.
Examination of the statistical summary, Table 6, for reagent water indicates
that the recovery for the low Youden pair was 107.5 to 117.2 percent, while
34
-------
TABLE 7. REGRESSION EQUATIONS FOR PRECISION AND BIAS
Water Type X S Sj.
Reagent X = 0.807C + 14.1 S =-0.0128 X + 14.2 Sr = -0.0092 X + 12.7
Surface X = 0.894C + 7.14 S = 0.0374 X + 2.68 Sr = -0.0109 X + 6.14
Ground X = 0.896C + 6.38 S = 0.0280 X + 3.40 Sr = 0.0033 X + 5.48
Chlorinated
Drinking
Water - S = 0.0946 X - 9-22 Sr = 0.1037 X - 0.1014
X = Mean recovery (bias) as yg/L
S = Overall precision as yg/L
Sr = Single-analyst precision as yg/L
C = True value as yg/L
35
-------
those of the medium and high Youden pairs ranged from 83.5 to 92.4 percent.
Inspection of the recovery regression equation for reagent water clearly
indicates that the intercept is large compared to the slope because the actual
recoveries obtained for all Youden pairs far exceed the 80.7 percent suggested
by the slope of the regression equation.
In contrast, the slope of the recovery regressions for ground and surface
waters predict recoveries of about 89 percent, while the recoveries actually
observed for the medium and high Youden pairs ranged from 88.9 to
94.3 percent. The low Youden pairs for all water types gave actual recoveries
in excess of 100 percent. The most probable causes for the positive bias at
low concentrations are contamination of the columns during preparation and
contamination of the analytical system in general when the samples axe
transferred from the adsorption column to the pyrolysis oven.
The algebraic relationship that guides the interpretation of the recovery
regressions also applies to those for overall precision S and single-analyst
precision Sr. These equations suggest that neither S nor Sr depend
significantly on concentration as evidenced by the unusual appearance of very
small or negative slopes coupled with relatively large intercepts. The
regressions for S for ground and surface waters, however, indicate slightly
more dependence of S on concentration. The extraordinarily large intercepts
associated with the S and Sr regressions for distilled water are not easily
explained. Examination of the raw data, (Appendix C), indicate that for all
concentration levels the large values of S were not caused by erratic data
from one or two laboratories. Moreover, the absolute S and Sr values in the
statistical summary, Table 6, for reagent water correspond closely to the
intercepts of the regression for S and Sr, it is difficult to dismiss the equations
as invalid. The comparative sizes and slopes and intercepts of the equations
for all waters suggest that the most significant contributors to the overall
precision of TOX analyses are related not to characteristics of the sample, but
to aspects of the method, such as sample and column manipulation and
variable contribution of blanks.
36
-------
Performance of the TOX method with chlorinated drinking water is discussed
separately here because of the nature of the sample itself and the slightly
different treatment of the study data. First, because the samples of this
water type were not spiked, but prepared by dilution of a previously
chlorinated drinking water, no true concentrations are available with which to
calculate a regression for mean recovery (X) against true value (T). Secondly,
two rather than three Youden pairs were prepared, approximating the low and
middle concentrations of the reagent, ground and surface waters discussed
above. Table 7 gives the regressions for S and Sr. The regression for overall S
takes an unusual form with its strongly negative intercept of -9.22. Because
the regression was calculated from only four data points rather than the
traditional six, the contribution from the highest S value of 29.6 for sample
four is magnified at the low concentrations.
Performance of the method for analyzing this water type is best discussed not
in terms of the regression equations but in terms of the actual S, and Sr values
obtained in the study (Table 6). The individual S values ranged linearly from
3.1 to 29.6 over the range of concentrations tested. Single-analyst precision
also appeared to be concentration dependent, ranging from 4.5 for the low
Youden pair to 22.8 for the middle pair.
EFFECT OF WATER TYPES
A summary of the statistical analyses to determine the effect of water type
on precision and bias is presented in Appendix C, Table C-5. Using the
multiplicative model described by Outler and McCreary (3), surface water and
groundwater were compared to reagent water which is viewed as a control.
Statistical significance was not demonstrated for either comparison using the
F-test.
37
-------
REFERENCES
1. Youden, W.J., and Steines, E.M., Statistical Manual of the AOAC,
AOAC, Arlington, Virginia, 1975, p. 88.
2. Annual Book of ASTM Standards, Part 31, Water, AS&M, Philadelphia,
Pennsylvania, ASTM D2777-77, Standard Recommended Practice for
Determination of Precision and Bias of Methods of Committee D-19 on
Water.
3. Outler, E.C., and McCreary, J.H., Interlaboratory Method Validation
Study: Program Documentation, Battelle Columbus Laboratories, 1982.
4. Draper, N.R., and Smith, H., Applied Regression Analysis, 2nd Edition,
John Wiley and Sons, New York, 1981.
5. Thompson, W.R., Annals of Mathematical Statistics, Vol. 6, 1935. p. 214.
6. Bishop, T.A., et al., "Development of Appropriate Statistical Techniques
to Compare Analytical Methods Across Wastewaters," Report to the U.S.
Environmental Protection Agency, June 1983.
38
-------
APPENDIX A
INTERIM METHOD 450.1
39
-------
EPA 600/4-81-056
United States
Environmental Protection
Agency
>EPA Research and
Development
Total Organic Halide
Method 450.1 - Interim
Prepared for
Joseph A. Cotruvo
Di rector
Criteria and Standards Division
Office of Drinking Water
Prepared by
Stephen Billets, Ph.D.
James J. Lichtenberg
Physical and Chemical Methods Branch
Environmental Monitoring and Support Laboratory
Cincinnati, Ohio 45268
40
-------
TOTAL ORGANIC HAL IDE
Method 450.1
Interim
U. S. Environmental Protection Agency
Office of Research and Development
Environmental Monitoring and Support Laboratory
Physical and Chemical Methods Branch
Cincinnati, Ohio 45268
November 1980
41
-------
TOTAL ORGANIC HAL IDE
Method 450.1
1. Scope and Application
1.1 This method is to be used for the determination of Total Organic
Hal ides as CI" by carbon adsorption, and requires that all
samples be run in duplicate. Under conditions of duplicate
analysis, the reliable limit of sensitivity is 5 yg/L. Organic
halides as used in this method are defined as all organic species
containing chlorine, bromine and iodine that are adsorbed by
granular activated carbon under the conditions of the method.
Fluorine containing species are not determined by this method.
1.2 This is a microcoulometric-titration detection method applicable to
the determination of the compound class listed above in drinking
and ground waters, as provided under 40 CFR 265.92.
1.3 Any modification of this method, beyond those expressly permitted,
shall be considered as major modifications subject to application
and approval of alternate test procedures under 40 CFR 260.21.
1.4 This method is restricted to use by, or under the supervision of,
analysts experienced in the operation of a pyrolysis/microcolumeter
and in the interpretation of the results.
2. Summary of Method
2.1 A sample of water that has been protected against the loss of
volatiles by the elimination of headspace in the sampling
container, and is free of undissolved solids, is passed through a
column containing 40 ng of activated carbon. The column is washed
-------
to remove any trapped inorganic halides, and is then pyrolyzed to
convert the adsorbed organohalides to a titratable species that can
be measured by a microcoulometric detector.
3. Interferences
3.1 Method interferences may be caused by contaminants, reagents,
glassware, and other sample processing hardware. All of these
materials must be routinely demonstrated to be free from
interferences under the conditions of the analysis by running
method blanks.
3.1.1 Glassware must be scrupulously cleaned. Clean all glassware
as soon as possible after use by treating with chromate
cleaning solution. This should be followed by detergent
washing in hot water. Rinse with tap water and distilled
water, drain dry, and heat in a muffle furnace at 400°C
for 15 to 30 minutes. Volumetric ware should not be heated
in a muffle furnace. Glassware should be sealed and stored
in a clean environment after drying and cooling, to prevent
any accumulation of dust or other contaminants.
3.1.2 The use of high purity reagents and gases help to minimize
interference problems.
3.2 Purity of the activated carbon must be verified before use. Only
carbon samples which register less than 1000 ng/40 mg should be
used. The stock of activated carbon should be stored in its
granular form in a glass container with a Teflon seal. Exposure to
the air must be minimized, especially during and after milling and
sieving the activated carbon. No more than a two-week supply
43
-------
should be prepared in advance. Protect carbon at all times from
all sources of halogenated organic vapors. Store prepared carbon
and packed columns in glass containers with Teflon seals.
3.3 This method is applicable to samples whose inorganic-halide
concentration does not exceed the organic-halide concentration by
more than 20,000 times.
Safety
The toxicity or carcinogenicity of each reagent in this method has not
been precisely defined; however, each chemical compound should be
treated as a potential health hazard. From this viewpoint, exposure to
these chemicals must be reduced to the lowest possible level by whatever
means available. The laboratory is responsible for maintaining a
current-awareness file of OSHA regulations regarding the safe handling
of the chemicals specified in this method. A reference file of
material-handling data sheets should also be made available to all
personnel involved in the chemical analysis.
Apparatus and Materials (All specifications are suggested. Catalog
numbers are Included for illustration only).
5.1 Sampling equipment, for discrete or composite sampling
5.1.1 Grab-sample bottle - Amber glass, 250-mL, fitted with
Teflon-lined caps. Foil may be substituted for Teflon if
the sample is not corrosive. If amber bottles are not
available, protect samples from light. The container must
be washed and muffled at 400°C before use, to minimize
contamination.
-------
5.2 Adsorption System
5.2.1 Oohrmann Adsorption Module (AD-2), or equivalent,
pressurized, sample and nitrate-wash reservoirs.
5.2.2 Adsorption columns - pyrex, 5 cm long X 6-mm OD X 2-mm ID.
5.2.3 Granular Activated Carbon (GAC) - Filtrasorb-400,
Calgon-APC, or equivalent, ground or milled, and screened to
a 100/200 mesh range. Upon combustion of 40 mg of GAC, the
apparent-halide background should be 1000-ng CI"
equivalent or less.
5.2.4 Cerafelt (available from Johns-Manville), or equivalent -
Form this material into plugs using a 2-mm ID
stainless-steel borer with ejection rod (available from
Dohrmann) to hold 40 mg of GAC in the adsorption columns.
CAUTION: Do not touch this material with your fingers.
5.2.5 Column holders (available from Dohrman).
5.2.6 Volumetric flasks - 100-mL, 50-mL.
A general schematic of the adsorption system is shown in
Figure 1.
5.3 Dohrmann microcoulometric-titration system (MCTS-20 or DX-20), or
equivalent, containing the following components:
5.3.1 Boat sampler.
5.3.2 Pyrolysis furnace.
5.3.3 Microcoulometer with integrator.
5.3.4 Titration cell.
A general description of the analytical system is shown in
Figure 2.
5.4 Strip-Chart Recorder.
45
-------
6. Reagents
6.1 Sodium sulfite - 0.1 M, ACS reagent grade (12.6 g/L).
6.2 Nitric acid - concentrated.
6.3 Nitrate-Wash Solution (5000 mg NO^/L) - Prepare a nitrate-wash
solution by transferring approximately 8.2 gm of potassium nitrate
into a 1-litre volumetric flask and diluting to volume with reagent
water.
6.4 Carbon dioxide - gas, 99.9% purity.
6.5 Oxygen - 99.9% purity.
6.6 Nitrogen - prepurified.
6.7 70% Acetic acid in water - Dilute 7 volumes of acetic acid with 3
volumes of water.
6.8 Trichlorophenol solution, stock (1 yL = 10 ug CI") - Prepare a
stock solution by weighing accurately 1.356 gm of trichlorophenol
into a 100-mL volumetric flask. Dilute to volume with methanol.
6.9 Trichlorophenol solution, calibration (1 yL = 500 ng CI") -
Dilute 5 mL of the trichlorophenol stock solution to 100 ml with
methanol.
6.10 Trichlorophenol standard, instrument-calibration - First, nitrate
wash a single column packed with 40 mg of activated carbon as
instructed for sample analysis, and then inject the column with
10 pL of the calibration solution.
6.11 Trichlorophenol standard, adsorption-efficiency (100 ug Cl~/L) -
Prepare a adsorption-efficiency standard by injecting 10 uL of
stock solution into 1 liter of reagent water.
6.12 Reagent water - Reagent water is defined as a water in which an
46
-------
interferent is not observed at the method detection limit of each
parameter of interest.
6.13 Blank standard - The reagent water used to prepare the calibration
standard should be used as the blank standard.
7. Calibration
7.1 Check the adsorption efficiency of each newly-prepared batch of
carbon by analyzing 100 mL of the adsorption-efficiency standard,
in duplicate, along with duplicates of the blank standard. The net
recovery should be within 5% of the standard value.
7.2 Nitrate-wash blanks (Method Blanks) - Establish the repeatability
of the method background each day by first analyzing several
nitrate-wash blanks. Monitor this background by spacing nitrate-
wash blanks between each group of eight pyrolysis determinations.
7.2.1 The nitrate-wash blank values are obtained on single columns
packed with 40 mg of activated carbon. Wash with the
nitrate solution as instructed for sample analysis, and then
pyrolyze the carbon.
7.3 Pyrolyze duplicate instrument-calibration standards and the blank
standard each day before beginning sample analysis. The net
response to the calibration-standard should be within 3% of the
calibration-standard value. Repeat analysis of the
instrument-calibration standard after each group of eight pyrolysis
determinations, and before resuming sample analysis after cleaning
or reconditioning the titration cell or pyrolysis system.
8. Sample Preparation
8.1 Special care should be taken in the handling of the sample to
'47
-------
minimize the loss of volatile organohalides. The adsorption
procedure should be performed simultaneously on duplicates.
3.2 Reduce residual chlorine by the addition of sulfite (1 mL of 0.1 M
per liter of sample). Addition of sulfite should be done at the
time of sampling if the analysis is meant to determine the TOX
concentration at the time of sampling. It should be recognized
that TOX may increase on storage of the sample. Samples should be
stored at 4°C without headspace.
8.3 Adjust pH of the sample to approximately 2 with concentrated HNO^
just prior to adding the sample to the reservoir.
9. Adsorption Procedure
9.1 Connect two columns in series, each containing 40 mg of
100/200-mesh activated carbon.
9.2 Fill the sample reservoir, and pass a metered amount of sample
through the activated-carbon columns at a rate of approximately
3 mL/min. NOTE: 100 mL of sample is the preferred volume for
concentrations of TOX between 5 and 500 ug/L; 50 ml for 501 to 1000
yg/L, and 25 ml for 1001 to 2000 ug/L.
9.3 Wash the columns-in-series with 2 mL of the 5000-mg/L nitrate
solution at a rate of approximately 2 mL/min to displace inorganic
chloride ions.
10. Pyrolysis Procedure
10.1 The contents of each column is pyrolyzed separately. After rinsing
with the nitrate solution, the columns should be protected from the
atmosphere and other sources of contamination until ready for
further analysis.
-------
10.2 Pyrolysis of the sample is accomplished in two stages. The
volatile components are pyrolyzed in a COg-rich atmosphere at a
low temperature to assure the conversion of brominated
trihalomethanes to a titratable species. The less volatile
components are then pyrolyzed at a high temperature in an 0?-rich
atmosphere.
NOTE: The quartz sampling boat should have been previously muffled
at 300°C for at least 2 to 4 minutes as in a previous analysis,
and should be cleaned of any residue by vacuuming.
10.3 Transfer the contents of each column to the quartz boat for
individual analysis.
10.4 If the Dohrmann MC-1 is used for pyrolysis, manual instructions are
followed for gas flow regulation. If the MCT-20 is used, the
information on the diagram in Figure 3 is used for gas flow
regulat ion.
10.5 Position the sample for 2 minutes in the 200°C zone of the
pyrolysis tube. For the MCTS-20, the boat is positioned just
outside the furnace entrance.
10.6 After 2 minutes, advance the boat into the 300°C zone (center) of
the pyrolysis furnace. This second and final stage of pyrolysis
may require from 6 to 10 minutes to complete.
11. Detection
The effluent gases are directly analyzed in the microcoulometric-titra-
tion cell. Carefully follow manual instructions for optimizing cell
performance.
49
-------
12. Breakthrough
Because the background bias can be of such an unpredictable nature, it
can be especially difficult to recognize the extent of breakthrough of
organohalides from one column to another. All second-column
measurements for a properly operating system should not exceed
10-percent of the two-column total measurement. If the 10-percent
figure is exceeded, one of three events can have happened. Either the
first column was overloaded and a legitimate measure of breakthrough was
obtained - in which case taking a smaller sample may be necessary; or
channeling or some other failure occurred - in which case the sample may
need to be rerun; or a high, random, bias occurred and the result should
be rejected and the sample rerun. Because knowing which event has
occurred may not be possible, a sample analysis should be repeated often
enough to gain confidence in results. As a general rule, any analyses
that is rejected should be repeated whenever sample is available. In
the event that the second-column measurement is equal to or less than
the nitrate-wash blank value, the second-column value should be
disregarded.
13. Quality Control
13.1 Before performing any analyses, the analyst must demonstrate the
ability to generate acceptable accuracy and precision with this
procedure by the analysis of appropriate quality-control check
samples.
13.2 The laboratory must develop and maintain a statement of method
accuracy for their laboratory. The laboratory should update the
accuracy statement regularly as new recovery measurements are made.
50
-------
13.3 It is recommended that the laboratory adopt additional
quality-assurance practices for use with this method. The specific
practices that would be most productive will depend upon the needs
of the laboratory and the nature of the samples. Field duplicates
may be analyzed to monitor the precision of the sampling
technique. Whenever possible, the laboratory should perform
analysis of standard reference materials and participate in
relevant performance-evaluation studies.
14. Calculations
OX as CI" is calculated using the following formula:
(Cr C3) + (C2 - C3 ) = yq/L T(Jtal 0rqanic Hal ide
V
where:
C.| = ug CI" on the first column in series
C2 = yg CI" on the second column in series
C^ = predetermined, daily, average, method-blank value
(nitrate-wash blank for a 40-mg carbon column)
V = the sample volume in L
15. Accuracy and Precision
These procedures have been applied to a large number of drinking-water
samples. The results of these analysis are summarized in Tables I and
II.
16. Reference
Dressman, R., Najar, G., Redzikowski, R., paper presented at the
Proceedings of the American Water Works Association Water Quality
Technology Conference, Philadelphia, Dec. 1979.
-------
TABLE I
PRECISION AND ACCURACY DATA FOR MODEL COMPOUNDS
Model
Compound
Dose
uq/L
Dose
as uq/L CI
Average
% Recovery
Standard
Deviation
No. Of
Replicates
chci3
98
88
89
14
10
CHBrCl2
160
106
98
9
11
CHBr^Cl
155
79
86
11
13
CHBr3
160
67
111
8
11
PentachloroDhenol
120
80
93
9
7
TABLE II
PRECISION DATA ON
TAP WATER ANALYSIS
Sample
Avg
uq
. halide
Cl/L
Standard
Deviation
No. of
Replicates
A
71
4.3
8
B
94
7.0
6
C
191
6.1
4
52
-------
Ln
U)
l\l 2
f
~
SAMPLE
RESERVOIR
(1 of 4)
i
GAC COLUMN 1
GAC COLUMN 2
Figure
±
NITRATE WASH
RESERVOIR
Adsorption Schematic
-------
Figure 2. CAOX Analysis System Schematic
-------
PYHOLYSIS FURNACE
Figure 3. Rear view plumbing schematic for MCTS-20 system.
Valve A is set for first-stage combustion. Oj venting
(puslt/pull valve out). Port B enters inner combustion
tube; Port C enters outer combustion tube.
-------
APPENDIX B
INSTRUCTIONS TO PARTICIPATING LABORATORIES
56
-------
January 7, 1985
Dear Participating Laboratory;
Please analyze the enclosed samples for Total Organic Halogen in strict
accordance with United States Environmental Protection Agency Method 450.1
A copy of this method has been sent to your laboratory previously.
Enclosed with this package is a note clarifying several points in the
method. Each person receiving or relinquishing custody of the enclosed
samples must complete the appropriate section of the enclosed sample
custody record. Please include the date and time of arrival of this sample
shipment on the custody record. Please store the samples in the dark and
in a refrigerator until analysis.
Please retain the shipment container until after completion of the sample
analysis. When analysis has been completed, return the sample bottle and
blue ice packet to the container, seal the container, and address the
container to our laboratory.
After we recieve your report of results for the sample analysis, we will
contact United Parcel Service for round-trip shipment completion. UPS will
pick up our shipment container and return it to our facility at no charge
to you.
Also enclosed are two Statement of Conditions for Compliance forms. One
form is to be signed by the chemist performing the TOX analysis. The other
should be signed by the supervising chemist or laboratory director. These
forms should be sent back to us along with the sample results.
A report of results, to the nearest 0.1 micrograms organic chloride per
liter, should be sent to myself at Montgomery Laboratories.
THE REPORT OF RESULTS MUST BE RECEIVED BY DR. CLARK WITHIN FOURTEEN (14)
DAYS OF SAMPLE RECEIPT.
Please enclose the two signed statement of condition forms; the chain of
custody record; and all associated raw data information, including
instrument readings, all blank, standard and sample measurements; any
instrument malfunction or repair and loss of sample.
57
-------
If you have any questions, please contact
Sincerely,
Robert R. Clark, PhD
Senior Chemist
Montgomery Laboratories
555 East Walnut
Pasadena, CA 91010
(818) 796-9141
58
-------
MONTGOMERY LABORATORIES
a Division of James M. Montgomery, Consulting Engineers, Inc.
555 East Walnut Street, Pasadena, California 91101
(818) 796-9141/(213) 681-4255 Telex: 67-5420
CHAIN OF CUSTODY RECORD
Client: Environmental Protection Agency
Contract Number: #68-03-3163
Preparer's Name: Eric Crofts
Sample Description:
Commments:
SAMPLE CUSTODY TRANSFER
1. Relinquished By (Signature):
Date: Time:
To: Organization's Name:
Received By (Signature):
Date: Time:
Comments:
2. Relinquished By (Signature):
Date: Time:
To: Organization's Name:
Received By (Signature):
Date: Time:
Comments:
3. Relinquished By (Signature):
Date: Time:
To: Organization's Name:
Received By (Signature):
Date: Time:
Comments:
59
-------
MONTGOMERY LABORATORIES
a Division of James M. Montgomery, Consulting Engineers, Inc.
555 East Walnut Street, Pasadena, California 91101
(818) 796-9141/(213) 681-4255 Telex: 67-5420
SAMPLE CUSTODY TRANSFER
(Continued)
4. Relinquished By (Signature):
Date:
To:
Organization's Name:
Received By (Signature):
Date:
Comments:
5. Relinquished By (Signature):
Date:
To:
Organization's Name:
Received By (Signature):
Date:
Comments:
6. Relinquished By (Signature):
Date:
To:
Organization's Name:
Received By (Signature):
Date:
Comments:
Time:
Time:
Time:
Time:
Time:
Time:
7. Relinquished By (Signature):
Date:
To:
Organization's Name:
Received By (Signature):
Date:
Comments:
Time:
Time:
60
-------
STATEMENT OF CONDITIONS FOR COMPLIANCE WITH USEPA CONTRACT NO. 68-03-3163
Laboratories subcontracting with J.M. Montgomery, Consulting Engineers, Inc.
(JMM) to perform analyses for Total Organic Halogen (TOX) analyses in the
evaluation of E.P.A. Interim Method 450.1, pursuant to United States
Environmental Protection Agency (USEPA) Contract Number 68-03-3163, are
required to comply with the following conditions:
1. The USEPA interim Method 450.1 must be strictly adhered to in the
analysis of all samples provided pursuant to this contract. A copy of
the method has previously been provided. Several clarifications to the
method are enclosed.
2. The performance evaluation sample will be analyzed at no charge to JMM or
the USEPA.
3. The individual chemist performing
used in the analysis of the
subsequently analyze all samples
contract.
the analysis and the instrumentation
performance evaluation sample will
providied by JMM pursuant to this
4. All raw data associated with the analysis of samples for this contract
will be submitted to Dr. Robert R. Clark, 555 E. Walnut Street, Pasadena,
CA, 91101. Raw data should include all blank, standard, and sample
measurements; reports of any instrument malfunctions, maintenance and
repair; and loss of sample.
5. A report of results to the nearest 0.1 microgram of organic chloride per
liter, signed by the chemist performing the analyses and by the chemist's
supervisor, will be submitted to Dr. Robert R. Clark within fourteen (14)
days of sample receipt.
6. Complete written documentation of the raw data, quality control
information, and maintenance records for the Xertex-Dohrmann Total
Organic Halogen instrumentation shall be maintained throughout the time
frame of this contract by the subcontracting laboratory, and will be
submitted to JMM or the USEPA upon request.
7. A sample chain of custody form shall be signed and completed by each
person taking or relinquishing custody of each sample, and upon
completion of analysis the custody form will be submitted with the report
of results to Dr. Clark.
I, ,
(Print Name) (Title)
for on this day, , 1984,
(Company Name) (Date)
having read the above conditions, do agree to comply with these conditions in
the analysis of all samples provided to me pursuant to U.S. Environmental
Protection Agency Contract Number 68-03-3163.
(Signature)
61
-------
CLARIFICATIONS TO METHOD
During phase 1 of this study, several questions arose concerning the
wording of EPA Interim Methods 450.1. These notes are intended to clarify
those points, to ensure that all participating laboratories use the same
procedures.
Section 7.2. Nitrate Vash Blanks (Method Blanks)
The method states that the repeatability of the method background must
be established each day prior to sample analysis by analyzing "several"
nitrate vash blanks. This has been further defined as at^ least three such
blanks. Repeatability of later nitrate wash blanks is satisfactory if each
measurement is within 20% of the mean of the previous blanks.
Spacing an additional nitrate wash blank between each group of eight
pyrolyses is required so that the analyst can continually update the days'
average nitrate wash value.
Section 7.3. Blank Standards (Water Blanks)
The net results of the analyses of water blanks required in this section
are intended as an indicator of the cleanliness of the system and to ensure
lack of interferences. They are not used in any of the calculations for
standards or samples.
Section 7.2.1 and 9.3. Nitrate Vashes on Single Columns
Some confusion was noted between these two sections of the method.
Section 7.2.1 indicates that the nitrate wash blanks are obtained on single
charcoal columns, washed as instructed for sample analysis section 9.3.
Please note that section 7.2.1 is referring not to column geometry in
section 9.3 but to nitrate concentration and flow rate.
Section 12. Breakthrough Calculations
This section states that "all second column measurements for a properly
operating system not exceed 10% of the two column measurement." Note that
the net second column measurement should not exceed 10% of the net two-column
total measurement.
Since the sample size is only 250 mL and the analyst is required to
perform two 100 mL filtrations, no sample would remain for analysis if
breakthrough were indicated. Montgomery Laboratories can supply a limited
number of replacement samples if problems are encountered.
62
-------
Section 14. Calculations using Nitrate Vash Blank (Hethod Blank)
Phase 1 results indicated that the average nitrate wash blank value was
not being subtracted from adsorption efficiency standard values. It should
also be subtracted from the blank standards (water blanks) to determine that
the reagent water does not contain interferences or TOX concentrations that
exceed the method detection limit. When calculations are required during the
day to determine that analyses are within control limits, the average of the
nitrate wash blanks obtained thus far for the day must be used in those
calculations.
Additional Instructions
1. The analyst must take precautions not to touch the charcoal or column
plugs with the fingers. This can lead to serious contamination of the
system. This problem can be avoided largely by using 1) the charcoal
measuring scoop (available from Dohrman), and 2) a 2 mm ID stainless
steel borer and ejection rod for cutting column plugs (also available
from Dohrman). Clean hands (or even gloves) are essential when preparing
the columns.
2. The sample reservoir should be rinsed with two 100 mL volumes of reagent
water before adding another sample.
63
-------
MONTGOMERY LABORATORIES
a Division of James M. Montgomery, Consulting Engineers, Inc.
555 East Walnut Street, Pasadena, California 91101
(818) 796-9141/(213) 681-4255 Telex: 67-5420
Date:
Water Analysis for
TOTAL ORGANIC HALOGEN (TOX)
Report of Instrument Data
Instrument Model No:
Value
Sample
Description
Top
Column
Bottom
Column
Replicate 1
Value
Top
Column
Bot torn
Column
Replicate 2
Instrument Calibration Standard
Standard Blanks
Method Blanks
Method Blanks
***
Instr. Calib. Std. or Method Blk*
***
Instr. Calib. Std. or Method Blk*
***
***
Instr. Calib. Std. or Method Blk*
***
~ ~~
Instr. Calib. Std. or Method Blk*
***
Adsorption Efficiency Standard Amount spiked
Value recovered
***
* Circle one.
*** Second adsorption column in series not required.
Submitted By
Checked By
64
-------
APPENDIX C
RAW DATA
65
-------
TABLE C-l. RAW DATA FOR REAGENT WATER
Low Youden Pair
CT>
CT*
True Concentration
in ug/L as CI
Laboratory
1
2
3
4
5
6
7
8
9
10
Lab
Rejected
AMPUL 1
38.7
33.0
38 . 2
43.8
64.4
52 . 1
70 . 9
4 1.9
28.4
50.2 *CO
35.3
AMPUL 2
54.2
68.9
51.5
77 . 7
70 . 5
55.3
50 . 7
68.8
38.8
50 . 2
50 . 3
Medium Youden Pair
AMPUL 3 AMPUL 4
193.4
164.7
158.1
155.4
164.9
160.4
159.8
174.7
150.1
190.5*
166.1
243 .7
223 .9
199.7
199.9
203 .7
239 .4
196.4
227.4
207.3
206.9
214.7
High Youden Pair
AMPUL 5 AMPUL 6
386 .7
329 .7
319.9
314.6
336 .2
355.3
3 37 .6
336.8
336 .3
318.0
335 .1
441.1
363 .7
386 .5
376.5
376.1
394 . 6 *CO
360 .2
404 .4
370.7
246.6*CO
387 .3
= Rejected, CO = Rejected by Cochran's Test
Current Significance Levels: 1. Lab Ranking Data Rejection Tests at 0.05 Significance Level
2. Individual Outlier Tests Using Thompson's Rule at 0.05 Significance Level
-------
TABLE C-2. RAW DATA FOR SURFACE WATER
Low Youden Pair
True Concentration
in ug/L in CI
Laboratory
1
2
3
4
5
6
7
8
9
10
Lab
Re j ected
AMPUL 1
38.7
39.0
39.1
3 5.1
39.9
44.2
41.6
42.6
31.6'
37.2
4 2.7
AMPUL 2
54.2
52.6
69.5
58.9
50 . 7
97.2*
55.5
56 . 1
46 . 4 *
72.4
54 . 1
Medium Youden Pair
AMPUL 3 AMPUL 4
193.4
177.1
177.8
167.4
180.6
184.7
186.2
179.2
136.6*
97.3 *CO
177.0
243 .7
209.8
222 .4
218.7
23 2 .9
253.1
231.7
243.0
200.3'
2 29.9
226 .7
High Youden Pair
AMPUL 5 AMPUL 6
386.7
327.3
362 .3
327 .5
360.9
361.3
3 50 .8
217.3*
346.0*
182.4 * CO
352.7
441.1
399.5
390.1
375.0
411.0
389.1
402 .4
456.4'
373.3'
367.2
40 3.1
Rejected, CO = Rejected by Cochran's Test
Current Significance Levels: 1. Lab Ranking Data Rejection Tests at 0.05 Significance Level
2. Individual Outlier Tests Using Thompson's Rule at 0.05 Significance Level
-------
TABLE C-3. RAW DATA FOR GROUNDWATER
Low Youden Pair
Medium Youden Fair
High Youden Pair
True Concentration
in ug/L as Cl
cr\
CD
Labo rato ry
1
2
3
4
5
6
7
8
9
10
Lab
Re j e ct e d
AMPUL 1
3 8.7
42.3
43.4
4 1.7
39.1
55.5*CO
38.3
4 3.9
39.7*
35.3
41.8
AMPUL 2
54.2
52.5
52.7
47.4
66.9
48.4
57 . 8
70 . 2
40.3*
48.9
55.6
AMPUL 3
193.4
182.7
177.8
16 8.8
185.4
13 5.7*
184.5
179.5
189.2 *CO
163.3
189.4
AMPUL 4
243 .7
214.3
216.5
211.1
231.0
230 .8
225 .4
230 .7
202.9*
270.5*
225 .7
AMPUL 5
386 .7
363 .0
354.3
343.7
3 54.0
371.7
341.4
3 53.0
332.7'
340 .3
346.7
AMPUL 6
441.1
405.3
387 .0
388 .2
418.3
419.1
414.3
396 .0
383.5*
2 3 5.9 *CO
405 .1
Rejected, CO = Rejocted by Cochran's Test
Current Significance Levels: 1. Lab Ranking Data Rejection Tests at 0.05 Significance Level
2. Individual Outlier Tests Using Thompson's Rule at 0.05 Significance Level
-------
TABLE C-4. RAW DATA FOR CHLORINATED DRINKING WATER
Low Youden Pair
CT>
Reference Value
as ug/L as CI
Laboratory
1
2
3
4
5
6
7
8
9
10
Lab
Rejected
AMPUL 1
6 3.8
60.9
69.5
60.7
64.3
64.0
77.4 *CO
66.3
77.3*
60.6
64.0
AMPUL 2
83.7
81.0
90.8
77 . 1
74.7
78.0
93.2
97.9
76.4
83.7
8 2.9
Medium Youden Pair
AMPUL 3 AMPUL 4
139.8
123.8
150.1
134.2
132.2
148.7
158.1*CO
161.2
130.4
124.5
13 5.1
178.5
147.6
210.5
147.0
156.0
176.2
192.1
172.5
235 .2
195.2
152.3
Rejected, CO = Rejected by Cochran's Test
Current Significance Levels: 1. Lab Ranking Data Rejection Tests at 0.05 Significance Level
2. Individual Outlier Tests Using Thompson's Rule at 0.05 Significance Level
-------
TABLE C-5. EFFECT OF WATER TYPE ON TOX ANALYSIS
** POINT ESTIMATES **
REAGENT WATER SLOPE:GAMMA( 1 ) = 0 . 88406
WATER INTERCEPT(WATER-REAGENT) SLOPE(WATER-REAGENT)
-0.2035
-0.2587
0.0439
0.0537
SOURCE
*~ ANALYSIS OF VARIANCE »*
DF SUM OF SQUARES MEAN SQUARE
-J
O
REG(REAGENT) 1 112.13997 112.13997
REG(WATER/REAGENT) 4 0.08406 0.02102
ERROR 140 1.51648 0.01083
1.94 0.1071
TOTAL
145
11 3 . 74052
TABLE OF 95% CONFIDENCE INTERVALS FOR THE DIFFERENCES BETWEEN INTERCEPTS AND THE DIFFERENCES BETWEEN SLOPES
INTERCEPT(WATER-REAGENT) SLOPE(WATER-REAGENT)
WATER ESTIMATE INTERVAL ESTIMATE INTERVAL
-0.2035
-0.2587
( -0.4853
( -0.540 5
0.0783)
0.0231)
0.0439 ( -0.0109
0.0537 ( -0.0010
0.0988)
0.1083)
NOTE :
IF ZERO IS CONTAINED WITHIN A GIVEN CONFIDENCE INTERVAL THEN THERE IS NO STATISTICAL SIGNIFICANCE BETWEEN
REAGENT WATER AND THE CORRESPONDING WATER FOR THE ASSOCIATED PARAMETER(INTERCEPT/SLOPE).
THE SLOPE AND INTERCEPT ESTIMATES FROM THIS ANALYSIS ARE NOT THE SAME AS THOSE OBTAINED FROM THE PRECISION
AND ACCURACY REGRESSIONS PERFORMED EARLIER.
------- |