EPA-650/4-74-021
            COLLABORATIVE STUDY
                        OF
    METHOD FOR THE DETERMINATION OF

    PARTICULATE MATTER EMISSIONS FROM

             STATIONARY SOURCES
  (FOSSIL FUEL-FIRED STEAM GENERATORS)

                        by
                   Henry F. Hamil
                  Richard E. Thomas

             SwRI Project No. 01-3487-01
             EPA Contract No. 68-02-0623

                   Prepared for
           Methods Standardization Branch
Quality Assurance and Environmental Monitoring Laboratory
        National Environmental Research Center
           Environmental Protection Agency
          Research Triangle  Park,  N. C. 27711
       Attn: M. Rodney Midgett, Research Chemist
     Section Chief,  Stationary Source Methods Section
                   June 30,1974
         SOUTHWEST  RESEARCH  INSTITUTE
         SAN ANTONIO      CORPUS CHRISTI      HOUSTON

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

-------
       SOUTHWEST  RESEARCH  INSTITUTE
       Post Office Drawer 28510, 8500 Culebra Road
                San Antonio, Texas 78284
             COLLABORATIVE STUDY
                         OF

    METHOD  FOR THE DETERMINATION OF

    PARTICULATE MATTER EMISSIONS FROM

             STATIONARY SOURCES
   (FOSSIL FUEL-FIRED STEAM GENERATORS)

                         by
                    Henry F. Hamil
                   Richard E. Thomas

             SwRI Project No. 01-3487-01
             EPA Contract No. 68-02-0623

                    Prepared for
            Methods Standardization Branch
Quality Assurance and Environmental Monitoring Laboratory
        National Environmental Research Center
            Environmental Protection Agency
          Research Triangle  Park,  N. C. 27711
        Attn: M. Rodney Midgett, Research Chemist
     Section Chief, Stationary  Source Methods Section

                    June 30, 1974

                         Approved:
                          John T. Goodwin
                          Director
                          Department of Chemistry
                          and Chemical Engineering

-------
                                 SUMMARY AND CONCLUSIONS
      This report presents the results obtained from a collaborative test of Method 5, a test procedure for determining
particulate emissions from stationary sources.  Method 5 specifies that particulate matter be withdrawn isokinetically
from the source and its weight determined gravimetrically after the removal of uncombined water.

      The test was conducted at a fossil fuel-fired steam generating power plant using four collaborating laboratories.
Sixteen sample runs were made over a two-week period by the collaborators for a total of 63 individual determinations.

      The reported values of one of the laboratories were not included in the analysis. Conversation with other person-
nel who participated in the test, and inspection of the laboratory's sampling train subsequent to the test, provided info-
mation which indicated that  the determinations made were not representative of Method 5 results.

      Of the remaining determinations, one was eliminated due to  failure to maintain isokinetic conditions. The remain-
ing values were subjected to statistical analysis to estimate the precision that can be expected with field usage of
Method 5. The precision estimates are expressed as standard deviations, which are shown to be proportional to the mean
determination, 5, and are summarized below.

      (a)   Within-laboratory: The estimated within-laboratory standard deviation is 31.1% of 5, with 34 degrees of
           freedom.

      (b)   The estimated between-laboratory standard deviation is 36.7% of 8, with 2 degrees of freedom.

      (c)   From the above,  we can estimate a laboratory bias standard deviation of 19.5% of 5.

     The above precision estimates reflect not only operator variability, but, to an extent, source variability which
cannot be separated from these terms.  The results summarized above were obtained from a single test using the
data from three collaborators. Further testing would, of course, be necessary to obtain conclusive results.

     Comments on the use of Method 5 from the collaborators  are included, and recommendations are made for
the improvement of the method.
                                                    in

-------
                                 TABLE OF CONTENTS

                                                                                      Page

LIST OF ILLUSTRATIONS	    vi

LIST OF TABLES	    vi

I.    INTRODUCTION	     1

II.   COLLABORATIVE TESTING OF METHODS	     2

     A.   Collaborative Test Site	     2
     B.   Collaborators	     6
     C.   Philosophy of Collaborative Testing	     7
III.  STATISTICAL DESIGN AND ANALYSIS
     A.   Statistical Terminology    	     8
     B.   Collaborative Test Design	     9
     C.   Collaborative Test Data	    10
     D.   Precision of Method 5	    10

IV.  COMMENTS AND  RECOMMENDATIONS	    14

APPENDIX A—Method 5—Determination of Particulate Emissions From Stationary Sources .   .  .    17

APPENDIX B-Statistical Methods	    21

     B.1   Preliminary Analysis of the Original Collaborative Test Data	    23
     B.2   Significance of the Port Effect	    23
     B.3   Transformations	    23
     B.4   Empirical Relationship of the Mean and Standard Deviation in the Collaborative
          Test Data	    24
     B.5   Underlying Relationship Between the Mean and Standard Deviation	    26
     B.6   Weighted Coefficient of Variation Estimates	    28
     B.7   Estimating Standard Deviation Components	    30

LIST OF REFERENCES	33

-------
                                 LIST OF ILLUSTRATIONS




Figure                                                                                   Page




    1      Allen King Power Plant, Northern States Power and Light Co.	      2




    2      Allen King Power Plant .Overall Site Configuration	      3




    3      Allen King Power Plant, Sampling Site Configuration	      4




    4      Average Velocity Profiles	     	      5




    5      Sample Platforms and Ports	      6




    6      Work Area and Sample Port	      6




    7      Control Console Operation	      7




    8      Impinger Train Operation	      7




  B.I      Interlaboratory Run Plot	     25




  B.2      Intralaboratory Collaborator-Block Plot	     26







                                  LIST OF TABLES




Table                                                                                     Page




    1      General Information—Allen King Power Plant	      2




    2      Hourly Average Coal  Burned	     10




    3      Original Particulate Collaborative Test Data	     10




    4      Particulate Collaborative Test Data Arranged By Block	     11




  B.I      Significance of Port Effect	     24




  B.2      Data Transformation  to Achieve Run Equality of Variance	     24




  B.3      Interlaboratory Run Summary	     25




  B.4      Intralaboratory Collaborator-Block Summary	     26




  B.5      Bias Correction Factors for Sample Standard Deviations	     27




  B.6      Run Beta Values and  Weights   	     31




  B.7      Collaborator-Block Beta Values and Weights	     31
                                               n

-------
                                        I. INTRODUCTION
      This report describes the work performed and results obtained on Southwest Research Institute Project 01-3487-
001, Contract No. 68-02-0623, which includes collaborative testing of Method 5 for particulate emissions as given in
"Standards of Performance for New Stationary Sources."^  '

      This report describes the collaborative testing of Method 5 in a coal-fired steam generating power plant and gives the
statistical analysis of the data from the collaborative tests, and the conclusions and recommendations based on the
analysis of data.
*Superscript numbers in parentheses refer to the List of References at the end of this report.

-------
                           II.  COLLABORATIVE TESTING CF METHOD 5


 A.   Collaborative Test Site

      Arrangements were made for collaborative testing of Method 5 at the Allen King Power Plant of Northern States
 Power Company near St. Paul, Minnesota. Through a subcontract with Thermo-Systems, Inc., access to the plant, sam-
 pling ports and platforms, and other sampling facilities were provided, and logistical support obtained. Facilities were
 installed which provide for simultaneous sampling by four collaborators, each collaborating team working on a separate
 platform with  a separate sampling port.

      The  power plant was visited in December, 1972 to inspect the facilities being constructed and to investigate several
 potential problems including length of probe required, effect of positive pressure in the duct, and statistical details of
 planning and conducting collaborative tests.

      Table 1 gives some information on the Allen King Power Plant, and Figure 1 shows a view of the power plant.  Fig-
 ure 2 shows the overall  site configuration, and the sample site configuration is in Figure 3. Velocity profiles across the
 sampling area within the duct are shown in Figure 4. The profiles were obtained by averaging the velocities of all four
 collaborators at each traverse  point from four randomly selected sampling runs. Due to the width of the duct, a wall-to-
 wall traverse could not be made with a 10-foot probe. Sampling the entire duct width would have required sampling
 from the center line of the duct out to the wall on  all  four ports for a single  sampling run. This procedure would
 have required moving all  four sampling trains through a  34-inch high crawl way underneath the duct on each run, and
 due to the time required would have precluded taking two samples per day. Since this study involved evaluation of
TABLE 1 GENERAL INFORMATION-ALLEN KING
               POWER PLANT

         Rated Capacity-550 megawatts
  (Normally Operated nearly full load at all times)
Age of facility
Stack height
Stack diameter
Coal usage at full load
Coal used
Coal sulfur content
Syr
800 ft
26 ft at bottom, 1 8 ft at top
240 tons/hr
Southern Illinois
3-1/4 percent
A small amount  of Montana  coal (sulfur content
about  1-1/2  percent)  is available  for use during
pollution alerts. Combustion chamber consists  of
twelve cyclone units exhausting into a common heat
exchanger system. The emission gas splits into two
identical gas streams shortly upstream of twin elec-
trostatic precipitators which normally  collect 98  to
99 percent of the fly ash  (by  weight). The  twin
emission gas streams meet  again at the base of the
vertical stack. Our sampling ports for the collabora-
tive tests are in the south—left horizontal duct, just
upstream of  the  vertical   stack.  Approximately
1 million cfm of emission gas passes through each  of
the twin emission gas streams.

Internal horizontal duct dimensions at    27 ft high
  sampling ports                     12 ft wide

Two  sampling ports, one on each side of the duct,
are located 6 ft above the  center line of the duct;
and two sampling ports, one on each side of the
duct, are located  6 ft below the center line of the
duct. The opposing ports are offset 6 in. vertically
to prevent interference.
                                         FIGURE 1.  ALLEN KING POWER PLANT
                                       NORTHERN STATES POWER AND LIGHT CO.

-------

                                             \
South
 Duct
Inner
Stack
North
Duct
                    to
                    O
    FIGURE 2. ALLEN KING POWER PLANT
      OVERALL SITE CONFIGURATION

-------
        Inner
        Stack
                X
                       r
                       9'
                       20'
           South Duct
   j-4	12'—-H
WU
WL
EU
                         u
                     _J.
EL
                                 ~r
                                 .1.
 FIGURE 3. ALLEN KING POWER PLANT
    SAMPLING SITE CONFIGURATION

-------
u
8

-------
Method 5 rather than characterization of the site, a modified tranversing procedure was adopted in which the samples
were obtained from a seven-foot wide section of the duct, beginning and ending two and one-half feet from the interior
duct walls. This seven-foot section could be traversed from either side of the duct, so that a sampling run consisted of
taking one-half the sample from an upper (or lower) port, followed by a port change, and taking the  other half of the
sample from the other port on the same side of the duct. A total of 24 traverse points. 12 on each port, were used on
each run. Figure 5 shows an overall view of the sampling platforms and ports on one side of the duct, while Figure 6
shows a view of an individual work area and sample port.
    FIGURE 5. SAMPLF PLATFORMS AND PORTS
B.   Collaborators
HGUKl, 6. WORK ARhA AND SAMPLK PORT
     The collaborators for the Allen King Power Plant test were Mr. Mike Taylor and Mr. Hubert Thompson of
Southwest Research Institute, Houston Laboratory, Houston, Texas; Mr. Charles Rodriguez and Mr. Ron Hawkins of
Southwest Research Institute, San Antonio Laboratory, San Antonio, Texas; Mr. Gilmore Sem, Mr. Vern Goetsch.
and Mr. Jerry Brazelli of Thermo-Systems. Inc, St. Paul, Minn.; and Mr Roger Johnson and Mr. Harry Patel of Environ-
mental Research Corporation, St. Paul, Minn.* As mentioned earlier in the report, Thermo-Systems, Inc. had the respon-
sibility for site preparation and liaison between test crews and plant personnel.

     Collaborative tests were conducted under the general supervision of Mr. Nolhe Swynnerton of Southwest Research
Institute. Mr. Swynnerton had the overall responsibility for assuring that the test was conducted in accordance with the
collaborative test plan and that all collaborators adhered to Method 5 as written in the Federal Register, December 23,
'1971.  The collaborating teams foi the test were selected by Dr. Henry Hamil of Southwest Research Institute.
*Throughout the remainder of this report, the collaborative laboratories are referenced by randomly assigned code numbers as Lab 101,
Lab 102, Lab 103, and Lab 104. These code numbers do not correspond to the above ordered listing of collaborators.

-------
      FIGURE 7. CONTROL CONSOLL OPERATION
                                                           HGURE 8. IMPINGER TRAIN OPERATION
      In Figures 7 and 8, members of the collaborative teams are shown in the operation of an impinger tram and a
control console during one of the test inns.

C.    Philosophy of Collaborative Testing

      The concept of collaborative testing followed in the tests discussed in this report involves conducting the test in
such a manner as to simulate "real world" testing as closely as possible. "Real world" testing implies that the results
obtained during the test by each collaborator would be the same results obtainable if he were sampling alone, without
outside supervision and without any additional information from outside sources, i.e., test supervisor or other
collaborators.
     The function of the test supervisor in such a testing scheme is primarily to see that the method is adhered to as
written  and that no individual innovations are incorporated into the method by any collaborator.  During the test
program, the test supervisor observed the collaborators during sampling and sample recovery  If random experimental
errors were observed, such as mismeasurement of volume of impinger solution, improper rinsing of probe, etc., no inter-
ference was made by the test supervisor. Since such random errors will occur in the everyday use of this method  in the
field, unduly restrictive supervision of the collaborative test would bias the method with respect to the held test lesults
which will be obtained when the method is put into general usage. However, if gross deviations were observed of such
magnitude as to make it clear that the collaborator was not following the method as written, the deviations would be
pointed out to the collaborator and corrected by the test supervisor.

     While most of the instructions in the Federal Register   are quite explicit, some areas are sub|ect to interpretation
Where this was the case, the individual collaborators were allowed to exercise their professional judgement as to the
interpretation of the instructions.

     The oveiall basis for this so-called "real-world" concept of collaborative testing is to evaluate the subject method
in such a manner as  to reflect the reliability and precision of the method that would be expected in peiformance  test-
ing in the field.

-------
                           III. STATISTICAL DESIGN AND ANALYSIS


A.   Statistical Terminology

      To facilitate the understanding of this report and the utilization of its findings, this section explains the statis-
tical terms used in this report.  The procedures for obtaining estimates of the pertinent values are developed and
justified in the subsequent sections.

      We say that an estimator, 6, is unbiased for a parameter 6 if the expected value of 0 is 9, or expressed in nota-
tional form, E(Q) = 0. From a  population of method determinations made at the same true concentration level, M.
let Xj... .,xn be a sample of n replicates. Then we define:

               1  "
      (1)   x = — /I x, as the sample mean, an unbiased estimate of the true determination mean, 5, the center of
              ",= 1
          the distribution of the determinations. For an accurate method, 6 is equal to p,  the true concentration.

                  1   "
      (2)   s1 =	yP (x, - x)2 as the sample variance, an unbiased estimate of the true variance, o2 . This
               n-  1^
           term gives a measure of the dispersion in the distribution of the determinations around 5.
      (3)   s = \/s  as the sample standard deviation, an alternative measure of dispersion, which estimates a, the
           true standard deviation.

      The sample standard deviation, s, however, is not unbiased foi a/ ' so a correction factor needs to be applied.
The correction factor for a sample of size n is an, and the product of an and s is unbiased for a. That is, E(ans) = a.
As n increases, the value of an decreases, going for example froma3 = 1.1284,a4 = 1.0854 to a10 = 1.0281. The for-
mula for ctn is given in Appendix B.5.

     We define
as the true coefficient of variation for a given distribution. To estimate this parameter, we use a sample coefficient
of variation, |3, defined by
where j3 is the ratio of the unbiased estimates of a and 6, respectively.  The coefficient of variation measures the
percentage scatter in the observations about the mean, and thus is a readily understandable way to express the
precision of the observations.


      The experimental plan for this test called for 16 runs.  On each run, the collaborative teams were expected to
collect simultaneous samples from the stack in accordance with Method 5, Since the actual particulate emission
concentration in the stack fluctuates, one can in general expect different true concentrations for each run.  To
permit a complete statistical analysis, the individual runs are grouped into blocks, where each block has approxi-
mately the same true emission concentration level.

-------
     We can apply the statistical terms of the preceding paragraphs both to the collaborators' values during a
given run  and  to each collaborator's values in a given block.  In this report, statistical results from the first situa-
tion are referred to as run results.  Those from the second situation are referred to as collaborator-block results.
For example, a run mean is the average of each collaborator's concentration level for the run as obtained by Method 5.
A collaborator-block coefficient of variation is the ratio of the unbiased standard deviation to the sample mean for all
the collaborator's runs grouped in the block.

     The variability associated with a Method 5 concentration determination is estimated in terms of the within-laboratory
and the between-laboratory precision components.  In addition, a laboratory bias component can be estimated. The
following definitions of these terms are given with respect to a true stack concentration, p..

     •     Within-laboratory -The within-laboratory standard deviation, a, measures the dispersion in replicate single
           determinations made using Method 5 by one laboratory team (same field operators, laboratory analyst, and
           equipment) sampling the same true concentration, ju.  The value of a is estimated from within each col-
           laborator-block combination.

     •    Between-laboratory-The between-laboratory standard deviation, a^, measures the total variability in a
           concentration determination due to simultaneous Method 5 determinations by different laboratories
           sampling the same true stack concentration, /u. The between laboratory variance, a^, may be expressed as


                                              a=a+ a2
           and consists of a within-laboratory variance plus a laboratory bias variance, a£ . The between-laboratory
           standard deviation is estimated using the run  results.

      •    Laboratory bias— The laboratory bias standard deviation, a/, = \/a6 ~~ °2 > is that portion of the total
           variability that can be ascribed to differences in the field operators, analysts and instrumentation, and
           due to different manners of performance of procedural details left unspecified in the method.  This term
           measures that part of the total variability in a determination which results from the use of the method by
           different laboratories, as well as from modifications in usage by a single laboratory over a period of time.
           The laboratory bias standard deviation is estimated from the within- and between-laboratory estimates
           previously obtained.


 B.   Collaborative Test Design

      The sampling was done through four ports, two on the east side (EU and EL)  and two on the west side (WU and
 WL). The experiment was designed so that  on each day,  each collaborator took one sample from the east side ports,
 and one from the west.  At the middle of each run, the collaborators using the upper ports shifted to the lower ones,
 and those on the lower ports began to use the upper ones. In this manner, any potential port effect was intended to
 be nullified.

      After receiving and making preliminary calculation checks on the data, an attempt was made to group the
 samples into  blocks. Considerations in setting up blocks  included time— whether each week constituted a block,
 load— whether megawatt hour load was a basis for a block, and coal burned— whether the particulate concentration
 was a function of the amount of coal burned. There is no accurate procedure for the determination of true particu-
 late concentrations, and thus it was impossible to establish blocks based on true or theoretical concentration levels.


      The plant provided its  daily logs of the hourly operating characteristics of the plant, and the pertinent informa-
 tion was extracted from these logs. It was assumed that the  amount of particulate matter which was emitted should
 depend upon how much fuel was burned. Thus, the average amount of coal burned during the course of each run
 was determined, and this was selected as the blocking criterion.  These amounts are listed in Table 2.

-------
   TABLE 2. HOURLY AVERAGE
         COAL BURNED
     Natural blocking of the sample runs appeared to be in groups of four,
from the highest fuel burn average to the lowest.  The result was four blocks
each of size four, in a randomized block design, as will be shown in Table 4.


C.   Collaborative Test Data

     The concentrations obtained by  the collaborators are presented in
Table 3.  The port sequence used to obtain the sample is also shown. Port
sequences are referred to as A (EL to EU), B (EU to EL), C (WU to WL),
and D (WL to WU).  Reported concentrations marked with a dagger are
those for which the isokmetic variation was determined to be outside the
acceptable range of 90 to 110 percent.
                                        The concentrations used in the analysis of Method 5 were those reported
                                   by Labs 102, 103 and 104 only. The observations from Lab 101 are generally
                                   lower than those from the other collaborators.  Discussions with personnel
                                   at the test site revealed that the glass joints of the filter holder were being
                                   sealed with a stopcock grease unsuitable for the high temperatures (250°F+)
                                   present in the filter oven.  Prior to each  run, the sample train was leak
checked according to the method.  This leak check is performed prior to applying heat to the filter oven. Upon heat-
ing the oven, there is a strong possibility that the melting of the low temperature grease used led to the development
of leaks during the run. Such leakage around the filter holder fitting would introduce ambient air into the sample
train as a diluent, which would lower the concentration level in the sample.
Day
8-14

8-15

8-16

8-17

8-20

8-21

8-22

8-23

Run
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Coal Burned
ton
351.0
247.2
304.1
322.0
231.4
300.9
232.6
228.8
232.1
241.3
246.0
214.4
225.5
244.5
229.2
238.3
Block
1
2
1
1
3
1
3
4
3
2
2
4
4
2
4
3
          TABLE 3. ORIGINAL PARTICULATE
             COLLABORATIVE TEST DATA
                    Ib/scfX 10-7*
Sample
(Run)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Lab
101
381 6 (A)t
51.4 (C)
237 0 (D)
85.3 (B)
139.7 (A)
50.8 (D)
114.8 (D)
63.8 (A)
87.9 (A)
72.7 (C)
96 0 (D)t
103.2 (A)
315.2 (B)
54 8 (C)
1127 (D)
56.5 (B)
102
137 3 (B)
191 2 (D)
176.8 (C)
185.3 (A)
194.9 (B)
173 6 (C)
335 7 (C)
1904 (B)
405 3 (B)
217 5 (D)
188.5 (C)
198.5 (B)
2109 (A)
205.2 (D)
138.8 (C)
167.3 (A)
103
58.4(C)
146.2 (B)
225.7 (A)
154.9 (C)
102.2 (D)
146.9 (A)
313 9 (B)
122.0 (D)
1970(0)
1309 (B)
1245 (A)
161 8 (C)
157 4 (D)
107 0 (B)
112.3 (A)
103 8 (C)
104
334.4 (D)f
151 5 (A)
375.1 (B)
103.5 (D)
102.8 (C)
163.8 (B)
132.3 (A)
125 9 (C)
161.8 (C)
151 3 (A)
351.2 (B)
111 5 (D)
119.4 (C)
- (A)
123. 8 (B)
99. 9 (D)
( ) Port sequence of sample collection in parenthesis.
*EPA policy is to express all measurements in Agency documents
in metric units. When implementing this practice will result in
undue cost or difficulty in clarity, NERC/RTP is providing con-
version factors for the particular non-metric units used in the
document. For this report, the factor is
10^7 Ib/scf = 1.6018 X 103 Mg/m3.
•j-Isokmetic variation outside acceptable range.
                          At the conclusion of the test, inspection of the
                     filter holder assembly revealed that the stopcock
                     grease melted and ran inside the filter holder, saturat-
                     ing half of the fritted glass filter support. The effect
                     of this was not immediately determinable, but it is
                     sufficient cause to suspect the validity of data from
                     Lab 101. On the basis of the above arguments, it was
                     decided to perform the data analysis without that
                     lab's values.
                                                             Sample  14 from Lab  104 is missing due to a
                                                        malfunction in the digital temperature indicator that
                                                        caused them to abort aftei beginning the run. In this
                                                        case, as was the case for the values with unacceptable
                                                        isokmetic variation factors, no attempt was made to
                                                        replace the values. The analysis was performed only
                                                        on those values which were actually taken during the
                                                        sample period. The concentrations upon which the
                                                        analysis was performed are presented in Table 4.
                                                        D.   Precision of Method 5

                                                             In a particulate matter determination, no
                                                        measurement of the accuracy of the method can be
                                                        obtained. There are no on-stream techniques for
                                                        analysis and no indicators of true concentration
                                                   10

-------
                  TABLE 4.  PARTICULATE COLLABORATIVE TEST DATA ARRANGED BY BLOCK
Method: EPA Method 5 -Determination of Participate Emissions Fiom Stationary Sources
Test Variable: X = Concentration of Participates, (Ib/scf) X 10'
Transformation: X Linear
Test Site. Allen King Power Plant
Collaborators. Lab 102, Lab 103, Lab 104.
Inter-Laboratory Run Summary
Block
1



2



3



4



Run
1
3
4
6
2
10
11
14
5
7
9
16
8
12
13
15
Lab 102
Data
137.3
176.8
185.3
173.6
191.2
217.5
188.5
205.2
194.9
335.7
405.3
167.3
190.4
198.5
210.9
138.8
Port*
(B)
(C)
(A)
(C)
(D)
(D)
(C)
(D)
(B)
(C)
(B)
(A)
(B)
(B)
(A)
(C)
Lab 103
Data
58.4
225.7
154.9
146.9
146.2
130.9
124.5
107.0
102.2
313.9
197.0
103.8
122.0
161.8
157.4
112.3
Port
(C)
(A)
(C)
(A)
(B)
(B)
(A)
(B)
(D)
(B)
(D)
(C)
(D)
(C)
(D)
(A)
Lab 104
Data
334 .4Et
375.1
103.5
163.8
151.5
151.3
351.2
O.OMt
102.8
132.3
161.8
99.9
125.9
111.5
119.4
123.8
Port
(D)
(B)
(D)
(B)
(A)
(A)
(B)
(A)
(C)
(A)
(C)
(D)
(C)
(D)
(C)
(B)
Run Summary
Mean
97.8
259.2
147.9
161.4
163.0
166.6
221.4
156.1
133.3
260.6
254.7
123.7
146.1
157.3
162.6
125.0
Std Dev
55.8
103.3
41.3
13.5
24.6
45.3
116.9
69.4
53.3
111.7
131.6
37.8
38.4
43.7
46.0
13.3
Coef of Var
0.5702
0.3986
0.2796
0.0837
0.1509
0.2718
0.5279
0.4448
0.4002
0.4285
0.5167
0.3060
0.2629
0.2777
0.2828
01063
*Port designation is the sequence of ports from which the sample was taken.
tE indicates an erroneous value due to isokmetic variation being out of acceptable range.
$M indicates no value was reported for that collaborator in that run.
levels.  Also, no type of standard sample for laboratory analysis can be prepared, either, which would give an estimate
of lab bias  and of the analysis component of the total variation.  Thus, the only technique available for evaluating
Method 5 is that of estimating the precision of the concentration estimates obtained and the degree to which the
results may be duplicated by a separate independent laboratory.


      In order to determine our variability estimates, we need to  determine what factors have a significant effect on
the variation in the reported values.  As previously stated, the possibility of a port influence on the concentration
obtained was considered, and the test was designed to minimize the effect  or the possibility of a lab-port interaction.
The hypothesis of no port effect is tested in Appendix B.2 and found to be an acceptable one. As a result, the port
factor may be eliminated from further analysis without apparent  consequence.

      In analyzing the data, two common variance stabilizing transformations, the logarithmic and square root, are
applied.  For the data under each of these transformations, and the data in its original form (linear), Bartlett's test
for homogeneity of variances is used to determine the effect of the transformation. A transformation which
satisfies the equality of variance hypothesis gives an indication of the nature of the distribution of the data  and
of any functional dependence between the  mean and the variance or standard deviation of the data. The results
of these tests are presented in Appendix B.3.
                                                     11

-------
      For the interlaboratory run data, the transformation which achieves the highest degree of equality of variance is
the logarithmic.  As was demonstrated in a previous collaborative study by Hamil and Camann/2' this is a strong indi-
cation of an underlying linear relationship between the population mean and standard deviation, 6 and a, respectively.

      As a further indication, a regression equation is fitted to both the run component and the lab-to-lab component
in Appendix B.4. Figure B.I represents a no-intercept model regression line for the sample means and sample standard
deviations from the run data. The correlation coefficient for the model is 0.939. Figure B.2 presents a similar analysis
for the collaborator-block means and standard deviations.  The correlation coefficient for this model is 0.862. Both
values indicate a significant linear relationship at the 5 percent level of significance.

      The consequence of this is to provide a model for the variance components, a2 and 0% . Let a2  be defined as
the variance associated with the replicates (samples) within a single laboratory  and ffjj be the variance associated with
the differences of means between laboratories. From the above argument, we have
and

                                                at,  = 0,5

where c and c^ are [unknown] constants. This implies that while 6, a, and a^ may vary from run to run or site to
site, the relationships a/5 and a^'6 remain constant. If we let

                                                 a
                                                 rf
and
                                               Ob
                                               I'*
then we have constant coefficients of variation, )3 and |3j, for the within-lab and between-lab components, respectively,
and a and a/, may be defined as a percentage of the mean.  In Appendix B.5, this relationship is established, and |3
and |3fc are defined.

      In Appendix B.6, the manner of obtaining estimates  using a linear combination of the beta values is obtained.
The values are given weights relative to the number of determinations used to obtain the estimate. In Appendix B.7,
the estimates of the precision components are obtained.

      The estimated within-lab coefficient of variation is (3  = (0.31 1), which  gives an estimated within-laboratory
standard deviation of
                                                  = (0.311)5.

This estimate has 34 degrees of freedom associated with it.

      The estimated between-laboratory coefficient of variation, j3^ , is (0.367), which gives an estimated between-
laboratory standard deviation of

                                              ob = (0.367)5.

This estimate has 2 degrees of freedom associated with it.
                                                      12

-------
Using these, we can estimate the laboratory bias standard deviation. From the formula in section A.,
                                          = V(0-367)262 -(0.311)282
                                          = (0.195)6.
                                                13

-------
                           IV.  COMMENTS AND RECOMMENDATIONS
     Assessments of Method 5 have been made by the collaborative test supervisor and by the collaborators them-
selves as a result of their observations and experience in conducting the field testing. These assessments have included
the following:

     (1)   In previous field experience as well as in conversations with many other persons using the method, it has
           become obvious that this method is more elaborate and time-consuming than most stack sampling methods.
           This results from mechanical  design of the equipment  plus the necessity to move heavy equipment items
           to the sampling point, especially if the sampling point is on a high stack. The necessity for mounting the
           sampling probe and sample box assembly on a rail for traversing across the stack also complicates the
           mechanical arrangement of equipment. These difficulties are inherent in the method as published, however,
           and cannot be avoided.

     (2)   The extensive use of large  amounts of glassware and ground glass connecting joints in the sampling train
           may result in leaks arising during the course of the run, which will influence the test result.

     (3)   The movement of equipment required in obtaining  a test result often leads to breakage of the glassware
           used in the sample equipment. In addition, mechanical shock placed on the equipment by raising it to
           platforms high  on the stacks, from  which sampling often must be done, can affect the calibration of the
           equipment as well as cause further glassware damage.

     (4)   The recovery of  particulate matter from the probe is a probable cause of high and low reported concen-
           tration levels.  During the extraction of the probe, the  tip may scrape against the inside of the stack, result-
           ing in an additional amount of particulate matter becoming lodged in the probe tip. This matter is then
           weighed and analyzed as part of the sample. A loss of particulate matter may occur during the probe wash,
           if care is not taken. It was noted that, from run to run and collaborator to collaborator, there was consider-
           able variation in the relative amount of particulate collected in the probe wash  as compared to the filter
           collection.

     (5)   The collaborators observed that the particulate matter collected was extremely hygroscopic. Even though
           such precautions as placing dishes of desiccant(P2 Os) inside the balance were  taken, the accuracy of the
           particulate weight determinations is doubtful.  As a result, the variation in concentration levels between
           labs is doubtlessly affected by the manner in which the filter particulate collection was handled as it was
           being weighed.

     The comments presented above  and the conclusions previously drawn provide a firm  basis for the following
     recommendations.
     (1)   Further testing at power plants is warranted in order to assess the precision of Method 5.  The relatively
           high values for the precision estimates may be representative of the true values. However, with usable
           data from  only three of the collaborators, and with only one site being tested, these results are inconclu-
           sive, and additional testing should be arranged.

     (2)   The collaborators made frequent calculation errors  in the collaborative  test data and there were differences
           among labs in the number of significant digits carried.  To prevent  these from unfairly influencing the result
           of a performance test for compliance, it is recommended that a standard Method 5 computer program be
           written to  calculate compliance test results from  raw field data.

     (3)   It is recommended that a standard  technique for  cleaning the filter apparatus be specified in detail in the
           method. As it stands now, the cleaning technique used varies somewhat from lab to lab, depends greatly
           on the  carefulness of the laboratory team, and  is  undoubtedly a major source of error.

     (4)   It is recommended that the technique for cleaning the probe be specified in greater detail in the method.
           Much of the variation in the method results from the probe cleaning, and details should be included in  the


                                                      14

-------
           method concerning the handling of the probe during sample recovery and the manner of recovering particulate
           matter from the probe.

     (5)   During sampling, many problems arise from the equipment used to obtain the samples. The design and
           reliability of much of the equipment now available for use with Method 5  do not seem adequate.
           As previously noted, the amount of glassware used in the equipment leads to unreliability, both in the
           equipment itself (from the high breakage levels) and in the performance (due to the probability of leaks
           arising during the course of the run). It is recommended that improvements be made in the equipment
           design and that efforts be made to eliminate the use of glassware and ground glass joints wherever pos-
           sible. Improvements in this area should be made at an early date, if at all feasible.

     By implementing these recommendations, the variation associated with Method 5 test results, both within- and
between-laboratory, should be able to be better separated from analytical and mechanical components.
                                                     15

-------
                       APPENDIX A

METHOD 5-DETERMINATION OF PARTICULATE EMISSIONS FROM
                  STATIONARY SOURCES

               Federal Register, Vol. 36, No. 247
                     December 23, 1971
                            17

-------
                                                   RULES  AND REGULATIONS
                                               2.1.4  Filter  Holder—.Pyrex'  glass  with
                                             heating system capable at maintaining mini-
                                             mum temperature of 325' V.
                                               2.1.5   ImplngerB / Condenser—Four impln-
                                             gers connected In series with glass ball joint
                                             fltunge. The flrat, third, and fourth  impln-
                                             gers  are of the Greenburg-Smlth  design,
                                             modified by replacing the tip with a % -Inch
                                             3D glass tube extending to one-half  Inch
                                             from the bottom of the flask. The second 1m-
                                             pinger  is  of the  Greenburg-Smith  design
                                             with the standard tip. A condenser may  be
                                             used In place of the impingere provided that
                                             the moisture content of  the stack gas can
                                             still be determined.
                                               2 1.6   Metering  system—Vacuum  gauge,
                                             leak-free pump,  thermometers  capable  of
                                             measuring temperature to within 5°  F., dry
                                             gas meter  with  2% accuracy, and related
                                             equipment, or  equivalent,  as required  to
                                             maintain an isoklnetlc sampling rate and  to
                                             determine sample volume.
                                               2.1.7   Barometer—To measure atmospheric
                                             pressure to  ±0.1 Inches Hg.
                                               2.2  Sample  recovery.
                                               3.2.1  Probe  brush—At  least as  long as
                                             probe.
                                               2.2 2  Glass wash bottles—Two.
                                               2.2.3  Olaas sample storage containers.
                                               2.2.4  Graduated cylinder—250  ml.
                                               2.3  Analysis.
                                               23.1  Glass weighing dishes.
                                               2.3.2  Desiccator.
                                               2.3.3  Analytical balance—To measure to
                                             ±0.1 ing.
                                               2.3.4  Trip balance—300 g. capacity,  to
                                             measure to ±0.05 g.
                                               3. Reagents.
                                               3.1  Samp.ling.
                                               3.1.1  Filters—Glass fiber, MSA  1106 BH>.
                                             or equivalent,  numbered  for Identification
                                             and prewelghed.
                                               3.1.2  Silica  gel—Indicating  type,  6-18
                                             mesh, dried  at 175" C. (350° F.)  for 2 hours.
                                               3.1.3  Waiter.
                                               3.1.4  Crushed ice.
                                               3.2  Sample recovery.
                                               3.2.1  Acetone—Reagent grade.
                                               3.3  Analysis.
                                               3.3.1  Water.
                                                                                              IMP4NGER TRAIN OPTIONAL MAY BE REPLACED
                                                                                                    BY AH EQUIVALENT CONDENSER
METHOD  5—DETERMINATION or PARTICTJLATI
   EMISSIONS  FEOM STATIONARY  SOURCES

   1. Principle  and applicability
   11  Principle Particulate matter is with-
drawn Ifiokinetically from the source and its
weight is determined gravlmetrically after re-
moval of uncomtoined water.
   1.2  Applicability. This method is applica-
ble for the determination of particulate emis-
sions from stationary sources only  when
specified by the test procedures for determin-
ing compliance with New Source  Perform-
ance Standards
  2  Apparatus.
  2.1   Sampling train The design specifica-
tions of the particulate sampling train used
by EPA (Figure 5—1) are described in APTD-
0501.  Commercial  models  otf  this train are
available
  2.1.1  Nozzle—Stainless  steel (316)  with
sharp, tapered leading edge
  2.12  Probe—Pyrex1 glass with a heating
Bystern capable of  maintaining a minimum
gas temperature of 250° F. at the exit end
during  sampling  to prevent  condensation
from  occurring.  When  length  limitations
(greater than about 8 ft.) are encountered at
temperatures lese than 600° F., Incoloy 825 ',
or equivalent, may be used. Probes for sam-
pling gas  streams  at  temperatures  in excess
of 600° F. must have been approved by the
Administrator.
  2.1.3  Pltot  tube—Type  S,  or  equivalent,
attached  to  probe to monitor stack  gas
velocity.
                                              REVERSE-TYPE
                                              PITOT TUBE
                                                                        HEATED AREA  FILTER HOLDER / THERMOMETER   CHECK
                                                                                                                   ^VALVE
                                                                                                                    ,VACOUM
                                                                                                                       LINE
             THERMOMETI
                        DRY TEST METER
AIR-TIGHT
  PUMP
                          Figure 5-1. particulate-sampling train.
  332  Desiccant- -Drierite,1 indicating
  4  Procedure.
  4.1  Sampling
  4 1.1  After selecting the sampling site and
the  minimum  number of sampling  points,
determine  the  stack pressure, temperature,
moisture, and range of velocity head.
  412  Preparation  of   collection   train.
Weigh to the nearest gram approximately 200
g of silica gel. Label a filter of proper diam-
eter,  desiccatea for  at least  24 hours  and
weigh to the nearest 0 5 mg in a room where
the relative humidity Is less than 50%. Place
1OO  ml. of  water  in each of  the  first  two
Impingers, leave the third impinger empty,
and place approximately 2OO g of preweighed
silica gel in the fourth  impinger Set up the
train without the probe as in Figure  5-1.
Leak check the  sampling train at  the sam-
pling site by plugging up  the inlet to the fil-
ter holder and pulling a 15 in Hg vacuum  A
leakage rate not in excess of 0.02 c f m at a
vacuum of  15 in. Hg Is  acceptable  Attach
the probe and adjust the  heater to  provide a
gas temperature of about 250° F at the probe
outlet. Turn on the filter heating system.
Place crushed Ice around  the implngers,  Add
  'Trade name.
  1 Trade name.
  'Dry using Drierite1 at 70° F.±10° F.
  more ice during the run to keep the temper-
  ature of the gases leaving the last impinger
  as low as possible  and preferably at 703 F ,
  or less Temperatures above 70° F. may result
  in damage to the dry gas meter  from either
  moisture condensation or  excessive  heat.
    413  Particulate train operation. For each
  run, record the data required on the example
  sheet shown in Figure 5-2 Take readings at
  each sampling point, at least every 5 minutes,
  and  when significant changes  in stack con-
  ditions  necessitate additional adjustments
  in flow rate. To begin sampling, position the
  nozzle at the first traverse  point  with the
  tip pointing directly into  the gas  stream
  Immediately start the pump and adjust the
  now to  isokinetic conditions. Sample for at
  least 5 minutes at each traverse point; sam-
  pling time must be the same for each point.
  Maintain Isoklnetlc sampling throughout the
  sampling  period. Nomographs are  available
  which aid In the rapid adjustment of the
  sampling rate  without other computations
  APTD-O576  details  the procedure for using
  these nomographs. Turn off the pump at the
  conclusion of each  run and record the final
  readings. Remove the probe and nozzle from
  the stack and handle In accordance with the
  sample recovery process described in section
  42
                                FEDERAL REGISTER,  VOl. 36, NO.  247—THURSDAY,  DECEMBER 53,  197]
                                                                 19

-------
                                                   RULES AND REGULATIONS
                                                                 AMBIENT TIWttATU«_

                                                                 BAHOICTUC mS5O«_

                                                                 ASSUMED MOtSTME. *__

                                                                 HEATEBMKSnTMQ	

                                                                 MQtf LENGTH.».	
                                                                 NOZZLE DIAMETER. (•.	
                                   SCHEMATIC Cf PACK CTOSS SECTION
IRA VEJtK POUR
NUMKR












TOTAL
SAHKtNO
TW
(M, •**













AVfHAGt
STATIC
PflEUUK
!fsl. (•, Hf.














STACK
TEkVUATU*
(V.*f














vciocm
HEAD
(*M.














FHUSUW
ofHKWTUL
Acwoa
Mtfict
UCTCB
UH.
!*>*>














GASSAUU
VOLUME
I*-).*1














GAS &AMFU TEITEIUTUK
AT OUT GAS UCHR
INUT
IT-fcl.'f












A*9.
OUTLET
(T» ^l. * F












A**.
A.B
SAHAfKnt
T£M>£RAtUK.
°F














rewflwrure
OF CAS
LEAVING
COHDEKER OR
LAiT MPIHGER
°F














                                                  Tn—Average dry gas meter temperature,

                                                 P,,.,—Barometric pressure at the orifice
                                                         meter, laches Hg.
                                                  AH—Average  pressure drop  across  the
                                                         orifice  meter, Inches H,O.
                                                 13.6=Specific gravity of mercury.
                                                 P.,4 = Absolute pressure at standard con-
                                                         ditions, 29.92 Inches Rg.
                                                8.3  Volume of water vapor.
                                                                                                                    0.0474
  4.2  Sample recovery. Exercise care in mov-
ing the collection train from the test site to
the  sample recovery area  to minimize  the
loss of  collected  sample  or  the  gain  of
extraneous  participate matter.  Set  aside  a
portion of the acetone  used In the sample
recovery as a blank for analysis. Measure the
volume of water from  the first three 1m-
plngers,  then discard. Place the  samples in
containers as follows:
  Container  No.  1. Remove the  alter from
Its holder, place  In this container, and seal.
  Container  No.  2. Place  loose  participate
matter   and  acetone  washings  from  all
sample-exposed  surfaces prior to the filter
In this container and seal. Use a razor blade,
brush, or rubber  policeman to lose adhering
particles.
  Container  No.  3  Transfer the  silica  gel
from the fourth Implnger to the original con-
tainer and seal.  Use  a rubber policeman as
an  aid  in  removing silica  gel from   the
implnger
  4.3  Analysis. Record the data required on
the  example  sheet shown in  Figure  5-3.
Handle each sample container as follows:
  Container No.  I. Transfer  the filter  and
any loose particulate matter from the sample
container to  a  tared glass  weighing dish,
desiccate, and dry to a constant weight. Re-
port results to the  nearest 0 5 nag.
  Container  No.   2.  Transfer  the  acetone
washings to a tared beaker and evaporate to
dryness at ambient temperature and pres-
sure. Desiccate and dry to a constant weight.
Report results to the nearest 0 5 mg.
  Container No. 3. Weigh the spent silica gel
and report  to the nearest gram.
  5. Calibration.
  Use  methods  and equipment which have
been  approved  by  th«  Administrator  to
calibrate  the orifice  meter, pitot tube,  dry
gas  meter,  and probe  heater. Recalibrate
after each test series.
  6. Calculations.
  6.1  Average dry gas meter temperature
and average orifice pressure drop.  See data
sheet  (Figure 6-2).
  6.3  Dry  gas volume.  Correct the sample
volume measured by  the dry gas meter to
standard conditions (70° *„ 29.92 Inches Hg)
by using Equation 5-1.
                                                                                           v..td
                               equation 5-*
where:
  Vw,ti= Volume of water vapor In the gas
            sample   (standard  conditions) .
            cu. ft.
     Vi0 = Total volume of liquid collected  in
            Imptngers and silica gel (see Fig-
            ure 6-3), ml.
    pa,o= Density of water, 1  gymL
  1 Ma,o— Molecular weight of  water, 18 lb./
            Ib.-mole.
      R= Ideal  gas  constant, 21.83 Inches
            Hg — cu, ft./lb.-mole-°R-
    T.ta = Absolute temperature at standard
            conditions, 630* R-
    P.n — Absolute pressure at standard con-
            ditions, 29.92 Inches Hg.
  6.4  Moisture content.
v     -
 °~

            '17.71
                    °R
                   in. Hg>
                              equation 5-3

where:
  Bwo ^Proportion by volumeof water vapor ill tliegas
         stream, dlmensionl&ss.
  ^'•m31 Volume of water in the gas sample (standaid
         conditions) , cu. ft.
  ^"•td =- Volume of gas sample through the dry gas mctr r
         (standard conditions), cu. ft.
  6.5  Total particulate  weight. Determine
the total particulate catch from  the sum of
the  weights  on  the  analysis   data  sheet
(Figure 5-3) .
  6.6  Concentration.
  6 8.1  Concentration In gr./s.c.f .
                                                     c'.
                                                                  ing.
                              equation 5-1
where:
  V».ld
        = Volume of gas sample through the
           dry gaa  meter (standard  condi-
           tions), cu. ft.
        «Volume of gas sample through the
           dry  gas  meter   (meter  condi-
           tions) , cu. ft.
        'Absolute temperature at standard
           conditions, 530° R.
                                              here:
                                                                           equation 5-4
    c'.= Concentration of particulate matter in stack
         gas, gr./s.c.f., dry basis.
   M.»Total amount of particulate matter collected,
         mg.
   "•w^Volume of gas sample through dry gas meter
         (standard conditions), cu.  ft.
                                                                     20

-------
     APPENDIX B




STATISTICAL METHODS
         21

-------
                                      STATISTICAL METHODS
      This appendix consists of various sections which contain detailed statistical procedures carried out in the analy-
sis of the particulate matter collaborative study data. Reference to these sections has been made at various junctures
in the Statistical Design and Analysis part of the body of this report. Each Appendix B section is an independent
ad hoc statistical analysis pertinent to a particular problem addressed in the body of the report.

B.1   Preliminary Analysis of the Original Collaborative Test Data

      Preliminary checks were made on the originally reported data in order to eliminate calculation errors and to
ensure that all concentrations were obtained using the specified formulas and correction factors.  In this manner,
results presented are representative of actual differences in the concentration levels found, rather than due to improper
or inaccurate use of the equations. Thus, we can discern more clearly the variability that is associated with Method 5
itself.

      The concentrations from the three collaborators, Labs  102, 103, and  104, were analyzed with two missing
observations. In Run 1, Lab 104 was only 77 percent of isokinetic, and the result was considered invalid.  On Run 14,
Lab  104 was forced to abort as a result of equipment malfunction, and no value was reported. The analysis was per-
formed on the remainder of the observations, with no attempt to replace these values.

      In each lab, there are values which appear, at first glance, to be outliers. These are of a magnitude of approxi-
mately twice that of the rest of the samples for that  lab.  However, these occurred with such regularity (2 to 3/lab)
that  it was decided not to make any adjustment for them in the analysis.  Indeed, they appear to be representative
of the type of error that occurs when Method 5 is used on a power plant.

B.2   Significance of the Port Effect

      The sampling at the Allen S. King Power Plant was done through four sample ports, two  on either side of the
duct. The ports were designated East Upper, East Lower, West Upper, and West Lower (EU, EL, WU, WL). On each
run,  a sampling team began on one side at either the upper or the lower port, shifting at the halfway point of the run
to the other port on the  same side. The result  is four port combination patterns:  EL to EU, EU to EL, WU to WL,
WL to WU, assigned the  designations A, B, C and D respectively.

      The test for port effect was made using the four sequences as different ports, rather than merely East vs West.
It is  felt that this provides a more complete evaluation of the  possible differences. The analysis is done  using
Youden's(4) rank test.

      Each sequence is assigned a rank within each run. The ranks are then summed for each sequence and the results
compared against  confidence limits tabled by Youden. The test is presented in Table B.I.

      The null hypothesis to be tested is that there is no port (port sequence) effect. This hypothesis may be
rejected, at the 0.05 level of significance, only  if the high port sum or the low port sum falls outside the limits of the
confidence interval. Since the highest sum, 44 (WL to WU), is less than the upper limit of 52, and the smallest sum,
33 (EU to EL) is greater than the lower limit of 28, the hypothesis may not be rejected.  Thus, we can assume no
differences due to ports, and further analysis is done under that assumption.

B.3   Transformations
      As a means of obtaining information about the nature of the distribution of the concentrations, the values
were examined to determine what kind of transformation best gives an equality of variance for the sample data.
                                                    23

-------
            TABLE B.I. SIGNIFICANCE
                OF PORT EFFECT

                Youden's Rank Test
                         TABLE B.2. DATA TRANSFORMATION TO ACHIEVE
                                  RUN EQUALITY OF VARIANCE
            Run
              1
              2
              3
              4
              5
              6
              7
              8
              9
             10
             11
             12
             13
             14
             15
             16
                         Port*
           Rank
           Sums  40
           2
           1
           2
           3
           4
           4
           4
           3
           2
           1
           4
           3
           3
           2
           3
           3
33   43   44
           H0:  No Port Effect

                  C/0.95<28,52)

           Conclusion  Accept Ho, No
                      Port Effect
           *Port Sequences:
              A(ELto Ell),
              B(EU to EL),
              C(WU to WL),
              D(WL to WU).
Transformation
Linear
Logarithmic
Square Root
Test
Statistic
19.071
10.902
13.753
Degrees
of Freedom
15
15
15
Significance
0.20
0.75
0.55
For each transformation used, Bartlett's test for homogeneity of
variance was used to ascertain the degree of equality obtained.

      The data were examined in their original form (linear) and
passed through two transformations, the logarithmic and the square
root.  The results are presented in Table B.2 for the run component
data.  The significance levels are taken from a  table of x2 with
15 degrees of freedom.

      The logarithmic transformation clearly attains the best results,
implying that the run data follow the lognormal distribution. In addi-
tion, this is an indication, as presented in Hamil and Camani/2\ that
there is an underlying linear relationship between the mean and the
standard  deviation  of the distribution of the run data.

B.4   Empirical Relationship of the Mean and Standard
      Deviation in the Collaborative Test Data

      In order to properly analyze the data from the collaborative
test, it is necessary to investigate the relationship between the mean
and the standard deviation for both the interlaboratory and  intra-
laboratory components. We wish to determine to what extent the
variability in the concentrations is related to the actual concentration
level.  Therefore, let
          be the concentration reported by lab i in block/ of run k.

          be the mean for run k in block; across labs,
             r-3
      sik =A / ~~'- (xijk ~~*./A;)2 > be the run standard deviation, assuming 3 collaborators.
             9';=1
      Table B.3 gives the values of the sample means and standard deviations obtained from the St. Paul test.
Asterisks denote those values taken over only two responses due to a missing or erroneous data point.

      By visual inspection, it appears that there is a linear trend between x.jk and s,-^; that is, as x.jk increases, so
does Sjk. A least squares regression line is calculated  for these points and presented in Figure B-l.  A no-intercept
model is used since a mean of zero could only logically occur when each reported concentration was identically
equal to zero, thus resulting in a zero standard deviation.

      A correlation coefficient between x ^  and s,^  is calculated to be 0,939 for the no-intercept model.  This value
is significant at a level greater than 10 percent.  The coefficient of determination is 0.881, indicating 88.1 percent of the
variation in the magnitude of the standard deviation is attributed to the variation in the magnitude of the sample mean.
                                                    24

-------
TABLE B.3. INTERLABORATORY RUN SUMMARY
Block
1



2



3



4



Run
*1
3
4
6
2
10
11
*14
5
7
9
16
8
12
13
15
Mean
Ib/scfX 10'
97.85
259.20
147.90
161.43
162.97
166.57
221.40
156.10
133.30
260.63
254.70
123.67
146.10
157.27
162.57
124.97
Std Dev
Ib/scfX 107
55.79
103.31
41.35
13.51
24.59
45.27
116.88
69.44
53.35
111.67
131.61
37.84
38.41
43.68
45.97
13.29
*Runs with only 2 determinations.
Thus, we have empirical evidence that the linear
relationship indicated by the transformation data
in Appendix B.3 is present.

      We can perform a similar analysis for the
intralaboratory  collaborator-block data. Let us
denote

      ~xij. as the mean for collaborator /', block/
                                                                   3  =
                                                                                     as the collabora-
                                                                tor-block standard deviation, assuming
                                                                four runs per block.

                                                           The values obtained for Xj/. and s,y are
                                                      presented in Table B.4.  Asterisks denote those
                                                      values based on only three runs due to missing
                                                      values. A no-intercept regression line is fitted to
               Run Standard Deviation
                    10~7 Ib/scf
           25    50    75    100   125   150   175   200   225   250   275   300

                                       Run Mean, 10~7 Ib/scf

                          FIGURE B.I. INTERLABORATORY RUN PLOT
                               X.
                                  Jk
                                              25

-------
 TABLE B.4. INTRALABORATORY COLLABORATOR-
              BLOCK SUMMARY
Block
1
2
3
4
Collaborator
Lab 102
Lab 103
Lab 104*
Lab 102
Lab 103
Lab 104*
Lab 102
Lab 103
Lab 1 04
Lab 102
Lab 103
Lab 104
Mean,
Ib/scfX 107
168.25
146.47
214.13
200.60
127.15
218.00
275.80
179.22
124.20
184.65
138.37
120.16
Std Dev
Ib/scf X 10 7
21.22
68.57
142.62
13.44
16.23
115.35
113.54
100.13
29.03
31.71
24.89
6.37
*Blocks with only three observations.
 Si;
175


150

125


100


 75


 50

 25
               Collaborator-Block
               Standard Deviation, 10~7 Ib/scf
these points and represented in Figure B.2.
As before, the linear tendency of the Sy's
relative to the Jc,y.'s is apparent.

     The value of the correlation coeffi-
cient is 0.862, which again has a significance
level greater than 10 percent. The coeffi-
cient of determination is 0.742.

B.5  Underlying Relationship Between
     the Mean and Standard Deviation

     In Appendix B.4, the empirical rela-
tionship between the mean and  standard
deviation  for the between-laboratory com-
ponents is established. This implies that
                                                                       •V =
         25    50   75    100   125  150   175   200  225  250   275

                                  Collaborator-Block Mean, 10~7 Ib/scf

               FIGURE B.2. INTRALABORATORY COLLABORATOR-BLOCK PLOT
                      X,;
                                           26

-------
TABLE B.5. BIAS CORRECTION
  FACTORS FOR SAMPLE
  STANDARD DEVIATIONS
n
2
3
4
an
1.2533
1.1284
1.0854
                                " where b is the sample coefficient of variation. The collaborator block standard
                                deviation, s/f, represents an estimate of the within-lab standard deviation, a. How-
                                ever, it has been shown(5) that the sample standard deviation is a biased estimate
                                and that a correction factor must be applied to remove the bias.  The correction
                                factor is given by
                                                                      r(-
where n is the size of the sample, and F represents the standard gamma function.  Values of an are presented in Table B.5.

Thus, E(anSjf) = a and

                                              a = E(anS(/)
If we let J5 = anb , we have
where a and 5 are the true mean and standard deviation for the distribution of the collaborator-block data.
      In Appendices B.3 and B.4, the linear relationship between the run mean and standard deviation is established,
that is
where b' represents the sample coefficient of variation for the run data. The expected value of sjk is a2 +
a within-laboratory component plus a laboratory bias component.  As before, s,^ is a biased estimator for
\/a2 + QI , and the bias may be corrected in the same manner.  Thus we have
and defining fe = aMfe', we have
                                                    27

-------
where a£ represents the true laboratory bias variation. From this relationship, it can be shown^ that
or

                                                 OL -

where
      Thus, it is established that for both the within-laboratory component and the laboratory bias component,
there is a linear relationship between the standard deviations, a and GI , and the mean 6.


B.6  Weighted Coefficient of Variation Estimates

      The technique used for obtaining estimates of the coefficients of variation of interest is to use a linear combi-
nation of the individual beta values obtained. The linear combination used will be of the form
                                                 /-!
where ft/ is the/th coefficient of variation estimate, k is the total number of estimates, and Wj is a weight applied to
the/th estimate.

     As previously discussed, the individual estimate of j3 is obtained as
for a sample of size n. This estimator is shown in B.5 to be unbiased for the true coefficient of variation. However,
since we are dealing with small samples to obtain our individual estimates, weighting is more desirable in that it pro-
vides for more contribution  from those values derived from larger samples.  There is more variability in the beta
values obtained from the smaller samples, as can be seen by inspecting the variance of the estimator.  We have that


                                       Var(|3) = Var ( °^

                                                   fr          1
                                             = a>   -(1+202)
                                                   l_2«         J
for normally distributed samples,(3)  and true coefficient of variation, /J.  Rewriting this expression, we have
                                                     28

-------
                                                       0+202)
and all terms are constant except for a2  and n. Thus, the magnitude of the variance changes with respect to the
factoi  oi^/n. Now, since an decreases as n increases, the factor Q$i/n must decrease as n increases, and the variance
is reduced.

      The weights, w/, are determined according to the technique used in weighted least squares analysis'  ', which
gives a minimum variance estimate of the parameter.  The individual weight, w,-, is computed as the inverse of the
variance of the estimate, ft, and then standardized. Weights are said to be standardized when
                                             "/=!

To standardize, the weights are divided by the average of the inverse variances for all the estimates. Thus, we can
write

                                                  = Uj
                                                    u
where
and
                                                     1
                                              M,-=-
                                                  Var(ft)
                                                1  K     1
     Now, from the above expressions, we can determine u,, u and w,- for the beta estimates.  For any estimate, ft,

                                                1
                                             Var(ft)
                                             <
for sample size n,. and
                                      u = —
                                          k  —' n2
                                          k /= 1 ««/
                                                  29

-------
 Thus, the ;'th weight, w/, is
                                      w, =
                                             *«,
      The estimated coefficient of variation is
                                          */=!
                                                      1=1
                                                      k   «c
                                                     V   '
                                                     2-i
B.7  Estimating Standard Deviation Components




     In Appendix B.3 and B.4, the relationships for the between-laboratory and within-laboratory standard deviations,

Oj, and a, are established as
and
                                                   30

-------
where fa and (3 represent the true coefficients of variation, and 6 is the true mean determination. In Appendix B.5
it is shown that for the laboratory bias standard deviation, QI , the above expressions imply
     In Appendix B.6, the technique for obtaining an unbiased estimate of a coefficient of variation as a linear
combination of the individual values is discussed. The estimator is of the form
where J3 is the estimated beta from the iih sample and w,- is a weight applied. For the between-laboratory coefficient
of variation, this becomes
                                                1
                                                     16
                                             = 77  £
              TABLE B.6  RUN BETA
              VALUES AND WEIGHTS
Run
I
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Beta Hat
0.7146
0.1703
0.4497
0.3155
0.4516
0.0944
0.4835
0.2967
0.5831
0.3067
0.5957
0.3134
0.3191
0-5575
0.1200
0.3453
Weight
0.573
1.061
1.061
1.061
1.061
1.061
1.061
1.061
1.061
1.061
1.061
1.061
1.061
0.573
1.061
1.061
TABLE B.7  COLLABORATOR-
   BLOCK BETA VALUES
       AND WEIGHTS
Block
1
2
3
4
Collaborator
Lab 102
Lab 103
Lab 104
Lab 102
Lab 103
Lab 104
Lab 102
Lab 103
Lab 104
Lab 102
Lab 103
Lab 104
Beta Hat
0.1369
0.5081
0.7516
0.0727
0.1385
0.5971
0.4468
0.6064
0.2537
0.1864
0.1952
0.0576
Weight
1.054
1.054
0.731
1.054
1.054
0.731
1.054
1.054
1.054
1.054
1.054
1.054
The individual beta values and their weights are shown in Table B.6.  Substituting these into the formula we have

                                            ft,  =(0.367)

and as a result


                                                 = (0.367)5

For p = 3 laboratories, there are 3 — 1  = 2 degrees of freedom associated with this estimate.

     Similarly,  for the within-laboratory coefficient of variation, we have

                                            .   1  I*    -
                                                    31

-------
where fy is an estimated beta value from a collaborator-block combination, and wy is the corresponding weight. The
individual values and their weights are shown in Table B.7. Substituting into the above formula, we obtain

                                              0= (0-311)

which implies that

                                                  
-------
                                       LIST OF REFERENCES
I.    Environmental Protection Agency,  "Standards of Performance for New Stationary Sources," Federal
     Register, Vol. 36,  No. 247, December 23,  1971, pp  24876-24893.

2.    Hamil, Henry F., and Camann, David E., "Collaborative  Study  of Method for  the Determination of
     Nitrogen Oxide Emissions from Stationary  Sources,"  Southwest  Research  Institute  report  for Environ-
     mental Protection  Agency,  October 5, 1973.

3.    Cramer, H., Mathematical Methods of Statistics.   Princeton University Press, New Jersey,
     1946.

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

5.    Ziegler, R. K., "Estimators of Coefficients  of Variation Using k  Samples," Technometrics,  Vol 15,
     No. 2, May, 1973, pp 409-414.

6.    Dixon, W. J. and Massey, F. J., Jr., Introduction To Statistical Analysis, 3rd Edition. McGraw-Hill,
     New York, 1969.
                                                  33

-------
                                   TECHNICS* REPORT DATA
                            [Please read [inunctions tin the r^crse hi:tore complettn/!)
 ! REPORT NO.
 EPA-650/4-74-021
                                                           3. RECIPIENT'S ACCESSION-NO.
4. TITLE ANDSUBTITLE
  Collaborative  Study  of Method for the Determination of
 Particulate Matter Emissions from Stationary Sources
 (Fossil Fuel-Fired Steam Generators)	
5. REPORT DATE
  June 30, 1974
6. PERFORMING ORGANIZATION CODE
7. AUTHOH(S)

 Henry F. Hamil  and  Richard E. Thomas
                                                           8. PERFORMING ORGANIZATION REPORT NO
 SwRI 01-3487-01
9. PERFORMING ORGANIZATION NAME AND ADDRESS

 Southwest  Research  Institute
 8500 Culebra  Road
 San Antonio,  Texas
10. PROGRAM ELEMENT NO.

 1HA327   (ROAP 26AAG)
11. CONTRACT/GRANT NO.

 68-02-0623
 12. SPONSORING AGENCY NAME AND ADDRESS
 Environmental  Protection Agency, NERC
 Quality Assurance  & Environmental Monitoring Laboratory
 Methods Standardization Branch
 Research Triangle  Park, N.  C.  27711	
13. TYPE OF REPORT AND PERIOD COVERED
 Final Report
14. SPONSORING AGENCY CODE
 15. SUPPLEMENTARY NOTES
 16.ABSTRACT ifofs  report presents the results obtained  from  a  collaborative test of Method
 5, a test procedure  for determining particulate emissions from stationary sources.
 Method 5 specifies  that particulate matter be withdrawn isokinetically from the source
 and its weight  determined gravimetrically after the  removal  of uncombined water.  The
 test was conducted  at a fossil fuel-fired steam generating  power plant using four colla
 borative laboratories.   Sixteen sample runs were made  over  a two-week period by the
 collaborators for a  total of 63 individual determinations.   The reported values of one
 of the laboratories  were not included in the analysis. Conversation with other person-
 nel who participated in the test, and inspection of  the laboratory's sampling train
 subsequent  to the test, provided information which indicated that the determinations
 made were not representative of Method 5 results. Of the  remaining determinations, one
 was eliminated  due  to failure to maintain isokinetic conditions,  the remaining values
 were subjected  to statistical analysis to estimate the precision that can be expected
 with field  usage  of  Method 5. The precision estimates  are expressed as standard devia-
 tions, which are  shown  to be proportional to the mean  determination, <5, and are sum-
 marized as  follows:  (a) Within-lab: The estimated within-lab standard deviation is 31.1
 of 
-------

-------