Umt>
•
Research and Development
&EPA
Data Analysis of
Drinking Water
Asbestos Fiber
Size
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL HEALTH EFFECTS RE-
SEARCH series. This series describes projects and studies relating to the toler-
ances of man for unhealthful substances or conditions. This work is generally
assessed from a medical viewpoint, including physiological or psychological
studies. In addition to toxicology and other medical specialities, study areas in-
clude biomedical instrumentation and health research techniques utilizing ani-
mals — but always with intended application to human health measures.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/1-79-020
May 1979
DATA ANALYSIS OF DRINKING WATER
ASBESTOS FIBER SIZE
by
Michael E. Tarter
Department of Biomedical and Environmental Health Sciences
University of California
Berkeley, California 94720
Order No. CA-7-3036-J-I
Project Officer
James R. Millette
Exposure Evaluation Branch
Health Effects Research Laboratory
Cincinnati, Ohio 45268
HEALTH EFFECTS RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
-------
DISCLAIMER
This report has been reviewed by the Health Effects Research Laboratory,
U.S. Environmental Protection Agency, and approved for publication. Approval
does not signify that the contents necessarily reflect the views and policies
of the U.S. Environmental Protection Agency, nor does mention of trade names
or commercial products constitute endorsement or recommendation for use.
ii
-------
FOREWORD
The U.S. Environmental Protection Agency was created in response to
increasing public concern about the dangers of pollution to the health and
welfare of the American people and their environment. The complexities of
environmental problems originate in the deep interdependent relationships
between the various physical and biological segments of man's natural and
social world. Solutions to these environmental problems require an inte-
grated program of research and development using input from a number of
disciplines.
The Health Effects Research Laboratory was established to provide sound
health effects data in support of the regulatory activities of the EPA.
Evaluating man's exposure to environmental health hazards is a key segment
in developing such a data bank. Studies of exposure require identification,
characterization, and quantification of physical, chemical, and biological
agents found in the environment. In addition, exposure assessment involves
the determination of conditions that cause agents to be released into the
environment, the study of the routes and pathways to man, and research into
the body's ability to prevent the entrance of environmental hazards.
This report presents the results of statistical comparisons of the
sizes of asbestos fibers found in various types of drinking water supplies.
Because the body may handle asbestos fibers of various sizes in different
ways, it is important to study the sizes of^fj.bers to which man is exposed
from drinking water.
R. J. Garner
Director
Health Effects Research Laboratory
111
-------
ABSTRACT
A statistical study of asbestos fiber size characteristics was conducted
using data obtained from a variety of San Francisco Bay Area water systems.
Particular emphasis was placed on comparison of fiber length distributions for
samples collected from pre and post asbestos cement (AC) pipe systems. Sig-
nificant differences were detected between the fiber size distributions in
samples of raw water and water collected after a length of AC pipe. Little
difference was detected between the fiber size distributions of a raw water
sample and a treated water sample. It was also shown that before and after AC
pipe, fibers in the water differed most significantly in the length distribu-
tions of narrow fibers.
This report was submitted in fulfillment of order CA-7-3036-J-I by
Michael E. Tarter of the University of California, Berkeley. This report
covers a period from August 16, 1977, to June 30, 1978.
iv
-------
CONTENTS
Foreword iii
Abstract iv
Figures vi
Acknowledgments x
1. Introduction 1
2. San Francisco Water District (SFWD)
Raw Water and Treated Water Com-
parisons 3
3. East Bay Municipal Utility District
(EBMUD), Before and After Flow
Through Asbestos Cement (A/C) Pipe 9
4. SFWD, Hetch Hetchy Supply, Before
and After Flow Through A/C Pipe 20
5. Combined Region, Before and After
A/C Pipe Samples 25
6. Comparison of Narrow and Thicker
Fibers for Combined Before and
After A/C Pipe Samples 33
7. Comparison of Before and After
A/C Pipe Samples for Narrow
Fiber Group 37
8. Comparison of Before and After
A/C Pipe Samples for the wider
Fiber Group 44
9. Raw Water and Combined After A/C
Pipe Fiber Length Comparisons for
Narrow Fibers 47
10. Raw Water and Combined After A/C
Pipe Fiber Length Comparisons for
Thick Fibers 52
11. Discussion 54
References 57
-------
FIGURES
Number Page
1 Comparison of Cumulative Estimated Fiber
Length Data for Raw and Treated Water (X
axis scaled in microns) .................. 4
Comparison of Estimated Length Density
Function Data for Raw and Treated Water
(X axis scaled in microns) ...... . ......... 5
Histogram of San Andreas Treated Water
Fiber Length Data (micron X axis scale) .......... 7
Histogram of Crystal Springs Raw Water
Fiber Length Data (micron X axis scale) .......... 8
Estimated Cumulative Plot of EBMUD
Fiber Length Data (X axis scaled in log
micron units) ....................... 10
Estimated Population Density Plots of
EBMUD Fiber Length Data (X axis scaled
in log micron units) .................... H
7 Histogram of Fiber Length Data after
A/C Pipe (X axis log micron scale) ............. 12
R Histogram of Fiber Length Data before
A/C Pipe (X axis log micron scale) ............. 13
9 Frequency Diagram for Sample Number
Plotted against Fiber Length (Y axis
micron scale) ....................... 14
10 Frequency Diagram after Elimination of
Possible Outlier (Y axis micron scale) ........... 16
11 Estimated Population Density Plots of
EBMUD Fiber Length Data (X axis in
micron scale) ....................... 17
12 Histogram of EBMUD Fiber Length Data
after A/C Pipe (X axis micron scale) ............ 18
vi
-------
Number Page
13 Histogram of EBMUD Fiber Length Data
Before A/C Pipe (X axis micron scale) 19
14 Estimated Cumulative Plots of Fiber
Length Data from the Hetch-Hetchy
System (X axis micron scale) 21
15 Estimated Population Density Plots of
Fiber Length Data from the Hetch-
Hetchy System (X axis micron scale) 22
16 Histogram of Fiber Length Data after
A/C Pipe from Hetch-Hetchy System
(X axis in micron scale) 23
17 Histogram of Fiber Length Data Before
A/C Pipe from Hetch-Hetchy System
(X axis in micron scale) 24
18 Estimated Cumulative Plots of Pooled
Fiber Length Data (X axis in micron
scale, T = before, X = after) 26
19 Estimated Population Density Plots
of Pooled Fiber Length Data (X axis
in micron scale) 27
20 Histogram of Pooled Fiber Length Data
after A/C Pipe (X axis in micron
scale) 28
21 Histogram of Pooled Fiber Length
Data before A/C Pipe (X axis in
micron scale) 29
22 Comparison of Estimated Cumulative
Plots of Pooled Fiber Length Data
from Raw Water and Water after A/C
Pipe (X axis in micron scale, T «
raw, x = after) 30
23 Histogram of Pooled Fiber Length
Data after A/C Pipe (X axis in
micron scale) 31
24 Histogram of Raw Water Fiber Length
Data (X axis in micron scale) 32
vii
-------
Number Pagi
25 Frequency Diagram showing Fiber
Diameter against Fiber Length 34
26 Position of Frequency Diagram dur-
ing the Separation Process 35
27 Comparison of Cumulative Fiber
Length Data for Narrow and Wide
Fibers (T - narrow; X - wide) 36
28 Comparison of Cumulative Fiber
Length Data before and after A/C
Pipe for the Narrowest Fiber Group
(X axis scaled in microns) 38
29 Comparison of Cumulative Fiber
Length Data before and after A/C
Pipe for the Narrowest Fiber
Group using a Smoothing Technique
(X axis scaled in microns) . 39
30 Comparison of Cumulative Log Fiber
Length Data before and after A/C
Pipe for the Narrowest Fiber Group
(X axis in log micron scale) 40
31 Comparison of Estimated Fiber Size
Population Density Curves of Fiber
Length before and after A/C Pipe
for Narrowest Fiber Group (X axis
in log micron scale) 41
32 Histogram of Fiber Length Data for
Narrowest Fiber Group after A/C
Pipe (X axis in log micron scale) 42
33 Histogram of Fiber Length Data for
Narrowest Fiber Group before A/C
Pipe (X axis in log micron scale) 43
34 Comparison of Cumulative Fiber Length
Data before and after A/C Pipe for
the Wider Fiber Group (X axis in
micron scale; T = before; X = after) 45
35 Comparison of Cumulative Fiber
Length Data before and after A/C
Pipe for the Wider Fiber Group (X
axis in log micron scale; T =
before; X = after) 46
viii
-------
Number Page
36 Comparison of Pooled Fiber Length
Data after A/C Pipe and Raw Water
Fiber Length Data for the Narrow
Fiber Ranges (X axis in log micron
scale). 48
37 Comparison of Estimated Population
Density Curves for Pooled Fiber
Length Data after A/C Pipe and
Raw Water Fiber Length Data for the
Narrow Fiber Ranges (X axis in log
micron scale) 49
38 Histogram of Narrow Fiber Length
Data after A/C Pipe (X axis in
log micron scale) 50
39 Histogram of Narrow Fiber Length
Data for Raw Water (X axis in log
micron scale) 51
40 Comparison of Wide Fiber Length
Data after A/C Pipe and Wide Fiber
Length Data from Raw Water (X axis
in micron scale) 53
41 Frequency Diagram of Diameter and
Length for Combined Samples before
A/C Pipe (maximum sample size that
can be analyzed using GRAFSTAT) 55
42 Plot of Linear Regression for Data
Shown in Figure 40 (a = .05 confi-
dence band) . 56
ix
-------
ACKNOWLEDGMENTS
The cooperation of Jack C. Murchio and Lorene A. Jackson, who provided
considerable insight into the asbestos problem, is gratefully acknowledged.
GRAFSTAT program options were implemented by Jansin T. Wong-Lee. Technical
assistance was provided by Susan B. Lum.
-------
SECTION 1
INTRODUCTION
In this paper, with very few exceptions, information is presented in one
of the following three forms: 1) histograms, where estimated probability of
occurrence, i.e., the estimated distribution density of the variable under
consideration, is presented on the y-axis and plotted against the value of
the variable (See [2] Section 2), 2) nonparametric density estimates which
are similar to histograms but tend to be both smoother and more data-effi-
cient (See [2] Sections 3 through 5), and 3) cumulative distribution estimates
which are essentially indefinite integrals of the nonparametric density esti-
mates. Specifically, both the histogram and the nonparametric density esti-
mate associate the relative odds or chance y=f(x) On the y-axis, to the
observed length x of the fiber on the x-axis. The cumulative distribution
estimator associates the chance y=£(x) of observing an asbestos fiber of a
given length x, or a smaller length, to x.
For each set of comparisons the above three estimates will be presented
in reverse order with F(x) shown first, optimized nonparametric ?(x) second
and histogram ?(x) last. This choice of presentation was motivated by the
fact that significant differences between cumulatives F(x) are easier to con-
firm than differences between optimized nonparametric t(x). On the other
hand, once the difference between two f(x) has been confirmed, optimized non-
parametric f(x) tend to show where or for what values of x, differences occur.
Conventional histograms are presented last because, as described in [2] Sec-
tion 2, in a more detailed analysis, one may wish to view the sample rather
than to simply utilize the sample to infer properties of a population or,
differences or lack of differences between populations.
Information about the various sites is presented in order of increased
statistical significance or decreased p-value of estimated difference. For
example, despite large sample sizes, n = 318 and n = 658, Crystal Springs raw
water and San Andreas treated water samples yielded almost identically shaped
estimated fiber length distributions. Roughly speaking, this suggests that
the effect of treatment Is constant or non-existent for all fiber lengths for
these groups. In the next section this finding will be described in detail.
Understandably, the more complex the combination of variables used, the more
likely it is for one to be able to discern differences. Specifically, when
the lengths of the narrowest and the thicker fibers are considered separately,
it was found that one could more clearly visualize the difference between pre-
and post- A/C pipe samples of the narrowest fibers. Examples of the above
finding are presented towards the end of this paper in keeping with the
low-to-high level of significance order of presentation.
-------
It should also be noted that detailed information about the nature and
location of data sources can be found in Reference 1.
-------
SECTION 2
SAN FRANCISCO WATER DISTRICT RAW WATER
AND TREATED WATER COMPARISONS
Although the graph shown in Figure 1 may appear to be a single curve, it
is in actuality two estimated cumulatives of fiber length distributions. One
curve is constructed using data from samples of Crystal Springs raw water and
the other using samples of San Andreas treated water. The Crystal Springs
water flows into the San Andreas Reservoir and then through the San Andreas
treatment plant before distribution to the consumers. The Crystal Springs raw
water and San Andreas treated water asbestos counts are representative of wat-
er before and after treatment. The near identity of these two estimated dis-
tributions is implied by the bottom two lines of Figure 1. The maximum dif-
ference between the two estimated cumulatives is 0.14488E-01 = 0.14488E times
10 = 0.014488. Using the conventional Kolmogorov-Smirnov two sample-two
sided test (See [3] pp. 127-36) a maximum difference of .92769E-01 would be
necessary to assert, with an o = .05, that there is a difference between
the two groups. Naturally, an absence of evidence of population differ-
ence does not necessarily imply that the populations are statistically
identical. However, in light of the large sample sizes involved, Figure 1
strongly suggests that the San Andreas water treatment, which consists of
coagulation, filtration flocculation, caustic soda, fluoridation and chlo-
rination, is not affecting asbestos fiber length distribution.
*
As an exercise in interpreting Figure 1, consider a line segment drawn
parallel to the x-axis which begins at the 0.5000 value midway up the y-axis,
intercepts the curve y=F(x) and then changes direction and descends vertic-
ally until it intercepts the x-axis at a point close to x = 1. This value,
x = 1 is identical to the estimated median of both the sample size n = 318
and n = 658 group. In general, the numbers printed below each graph pro-
vide reference points, e.g., the 50th percentile =1, to be used for com-
parison purposes.
In all but one set of figures in this paper the post-treatment sample
will be described in the leftmost column headed by EXCL GRP (x) and the pre-
treatment samples will be described in the column headed by INCL GRP(T).
Usually the sample sizes will be large for the post-treatment samples, e.g.,
across from the word SAMPLE, under the symbols EXCL GRP(x), N = 658 while n
= 318 for the pre-treated sample.
Figure 2 corresponds to Figure 1 with the exception that the estimated
population probability densities, £(x), are displayed in place of the esti-
mated population cumulatives, F(x). The wavy shape of the estimated proba-
bility density in the tail or extreme value regions is either due to samp-
-------
UNIVARIATE
CTRL
1 GET NEW DATA
0 756011
2 VIEW/EDIT DATA
3
4
5
6
7
8
9
A
B
C
D
E
F
G
HSTGRAM
CHOOSE AB
RESET AB
PROB. DENSITY o.snooo
•H-CUM. CURVE
R(X) FCTN
SVIVAL CURVE
LOG XFORM
LKH SVIVAL ° 25I"°
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
0 e
QUIT & EXIT g 200
TiHyHjKkHMHHWO
rlM*'
(***'
/""*"
¥
f
*
X
,
1
j
J
I
<*
ix
JX
/
00 I. ison ? ipoo 3 nson
PARAMETERS EXCL GRP(X) INCL ORP(T)
SAMPLE 663 318
EST.MU 0.84600 0.84600
EST.SCMA 0.42025 0.44943
EST.MAX 0.91251 0.85515
MEDIAN 1.0000 1.0000
MAX DIF - 0.1448RE-01
SIC. PT (0.05) - .92769E-01
:W*,WftWWftW*
Figure 1. Comparison of cumulative estimated fiber length data for raw and
treated water (X axis scaled in microns).
-------
F f&ER
W-tf 1*31 .
UNIVARIATE
CTRL
1
2
3
i
S
fr
7
8
9
A
B
C
D
E
P
C
GET HEW DATA 0.6M38.
VIEW/EDIT DATA
HSTCRAM
CHOOSE AB
RESET AB
0.45625,
44PROB. DENSITY
CUM. CURVE
R(X) FCTN
SVIVAL CURVE
0.22813
LKH SVIVAL
HEUR. SVIVAL :
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR 0-0
» K
X x
y \
t
1 i
x x
T t
* T
I *
J
1 *
* T
V
I "
X '
» t
T* *f
x xt
i ,
i" XTT
x "'T
X T
X
0.20000 1 ISOO
«UIT * nlT PARAMETERS
SAMPLE
EST.MU
EST.SCHA
EST.MAX
MEDIAN
\Mitatm
***** *****&
%
*^^!J!!!^
7 1000 3.0SOO 4.0000
EXCL CRP(X) INCL GRP(T)
663 31R
0.84600 0.84600
0.42025 0.44943
0.91251 0.85515
1.0000 1.0000
VALUE OP T
0.0
Figure 2. Comparison of estimated length density function data for
raw and treated water (X axis scaled in microns).
-------
variation or the discrete nature of the asbestos fiber length measure-
ment process.
Unlike the estimated probability densities shown in Figure 2, the histo-
grams of San Andreas and Crystal Springs fiber length distributions are un-
corrected for sampling variation. However, due to the large sample sizes
available the two histograms are remarkably similar and lend a great deal of
support to the assertion that the fiber length populations under considera-
tion are statistically identical.
-------
11NIVAR1ATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 -H-HSTGRAM
4 CHOOSE AB
5 RFSF,T AB
6 PROB. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVF.
A LOG XFORM
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
F EXIT SVL TO UNIVAR n
G QUIT & EXIT
0 295
0
(3X
67X
73X
«
,3X
8X 9X
-gfn-^ V^r^T-g, ».
I 06751. 1 «S5 1435 815 2.195 2575 2955 3333 3715
«850 0 865« 1 215 1 625 2 005 2 3S5 2 765 3 145 3 525 3 90S
PARAMETERS KXCL GRP(X) INCL GRP(T)
SAMPLE 663 31*
EST m 0.84600 0.84600
EST.SCMA 0-42025 0.44943
EST.MAX 0.91251 0.85515
MEDIAN 1-0000 1.0000
Figure 3. Histogram of San Andreas treated water fiber length data (micron X axis scale)
-------
UNIVARIATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 -H-HSTGRAM
4 CHOOSE AH
5 RESET AB
6 PROB. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SV1VAL CURVE
A I,OC XFORM
B IJCH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
F EXIT SVL TO UNIVAR
C QUIT d EXIT
IT IT II
'SO I 055 I
0 8650 1 245
PARAMETERS
SAMPLE
EST.MU
EST.SGMA
EST.MAX
MEDIAN
1 815 2 195 2 S7S 2.955
3 335
I 625 2 005 2 3«5
EXCL GRP(X)
663
0.84600
0.42025
0.91251
1.0000
2 765 3 M5 3
INC!, GRP(T)
318
0.84600
0.44943
0.85515
1.0000
Figure 4. Histogram of Crystal Springs raw water fiber length data (micron X axis scale)
-------
SECTION 3
EAST BAY MUNICIPAL UTILITY DISTRICT (EBMUD)
COMPARISON BEFORE AND AFTER FLOW THROUGH AC PIPE
For reasons to be described later in this section, the natural logarithm
of fiber length was used as the x - axis variable in Figures 5 through 8. To
convert any of the numbers on the x - axis to the usual arithmetic scale one
can use a table of exp(x) or, somewhat less conveniently a table of loga-
rithms; e.g., the median for before AC pipe, 0, is equivalent to a fiber
length of exp(O) =1. In this one instance the sample size, n = 123, availa-
ble for the before AC pipe group is larger than that, n = 24, for the after
AC pipe group.
The after AC pipe distribution appears to be shifted to the right rela-
tive to the before AC pipe distribution. However, since the observed maxi-
mum difference equals 0.22444, which is smaller than the Kolmogorov-Smirnov
critical value of 0.30369, the difference between the populations cannot be
said to be significant for the sample sizes available.
One advantage of using a logarithmic transformation of fiber length data
is illustrated by the population density estimates shown in Figure 6. The
symmetry, homoscedasticity and bell-shape of these curves suggests that a
distribution*dependent-test for group difference might be preferable to the
Kolmogorov-Smirnov test which, although asymptotically distribution-free,
tends to have lower power than the latter test. However, in light of more
clear-cut results to be described in later sections of this paper, further
statistical elaboration for EBMUD data was not considered.
The two histograms shown in Figures 7 and 8 again tend to illustrate the
tendency towards discreteness or clumping of asbestos fiber length data.
Also note the possible "outlier" which occurs at a considerable distance
from the next largest value in Figure 8. Since this outlier was brought
closer to the main body of data by the logarithmic transformation its appa-
rent separation from the other data values would have been considerably
greater in a plot constructed by using an arithmetic scale as is shown in
Figure 9.
The numerical value of the possible outlier was determined by plotting
the sample code number against fiber length using the GRAFSTAT light pen
sensing option. (This and other GRAFSTAT data editing and identification
options are described in [4] and [5]). The fiber length determined by this
process was 58.721 microns as shown in Figure 9.
-------
CTRL
1
2
3
4
5
6
UNIVARIATE
GET NEW DATA « TSOHO
VIEW/EDIT DATA
HSTCRAM
CHOOSE AB
RESET AH
PROS. DENSITY 0 500(10
7 -t-t-CUM. CURVE
8
9
A
B
C
D
E
F
G
R(X) FCTN
SVIVAL CURVE
LOG XFORM
0 2SKOO
LKH SVIVAL
HEUR. SVIVAL
CUM. HA7.ARD
SVL DENSITY
EXIT SVt. TO UNIVAR
n ft
^^.m^fmmrwwmniuimOTfB*
T X
T X
' t
I
T X
I X
T
X
T
1 "
! X
T X
T X
T
X
T
X
I
I
t "
T *
I X
T X
T X
T X
I X
T X
t X
I X
T X
I X
t X
T'' J?
xW»*ilxxxx«*x'C
« •• — ^ - • -- T- 1 1 1
QUIT & EXIT "' 6094 - 19769 1 2340 "3 ^358 4 n/7S
PARAMETERS EXCL GRP(X) INCI. GRP(T)
SAMPLE 24 123
EST.MU 0.38100 -.13083
EST.SGMA 0.61R01 0.68323
EST.MAX 0.64274 0.58000
MEDIAN 0.22234 0.0
MAX RIF - 0.22444
SIG. FT (0.05) - .30349
Figure 5. Estimated cumulative plot of EBMUD fiber length data (X axis scaled in log micron units).
-------
UNIVARIATE
CTRL
1 GET NEW DATA , 4831)6
2 VIEW/EDIT DATA
3 HSTGRAH
4 CHOOSE AB
5 RESET AB
6 -WTROB. DENSITY 032137.
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
F EXIT SVL TO UNIVAR
G QUIT & EXIT
0 16069
X T
T X
JL
r&~-
- 18769
PARAMETERS
SAMPLE
EST.Ml)
EST.SCMA
EST.HAX
MEDIAN
T *
T x
T
T '
t
KXCI. CRP(X)
24
0.38100
0.61801
0.64274
0.22234
[NCL CRP(T)
123
-.13083
0.6R323
0.58000
0.0
VALUE OF T - 3.4065
Figure 6. Estimated population density plots of EBMUD fiber length data
(X axis scaled in log micron units).
-------
UHIVARIATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 -HflSTCRAM
4 CHOOSE AB
5 RESET AB
6 PROS. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B UCH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
F EXIT SVL TO UNIVAR
G QUIT & EXIT
NO. OF INTERVAL - 20
IX
IX
2X
3X 3X
SX
3X
1 «7 -8986 -3299 02388 0 80TS 1376 194S 2 5M 3082 3651
-1183 -6142 - <552E-0» 5232 1B92 1661 2229 2798 3367 3935
PARAMETERS EXCL CRP(X) INCL GRP(T)
SAMPLE 24 123
EST.HU 0.38100 -.13083
EST.SGMA 0.61801 0.68323
EST.MAX 0.6427* 0.58000
MEDIAN 0.22234 0.0
Figure 7. Histogram of fiber length data after A/C pipe (X axis log micron scale),
-------
UNIVARIATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 ++HSTCRAM
* CHOOSE AB
5 RESET AB
6 PROB. DENSITY
^ CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
' EXIT SVL TO UNIVAR
G QUIT & EXIT
167
2T
..
6T
12T
I7T
.211
101
gr
5T
31
It IT It 11
~~i n
OTH6 - 3299 « 2388 0 «075 1 376 1 915 ? 511 3 982 3 E.S1
-1 183 - 611? - 1S57E-OJ ^232 I S12 1 661 2 229 7 TP* 1 ~!Kf 3 9T5
NO. OF INTERVAL - 20
PARAMETERS EXCL CRP(X) INCL CRP(T)
SA1IPLE 24 123
FST.MU 0.38100 -.13083
EST SCMA 0.61801 0.68323
FST MAX 0.64274 0.58000
MEDIAN 0.222J4 0.0
Figure R. Histogram of fiber length data before A/C pipe (X axis log micron scale)
-------
SMMPLE NUMBEK
CTRL
1
VIEW/EDIT DATA
GET NEW DATAi-
2 -H-VIEW/EDIT DATA
3
4
5
6
7
8
9
A
B
C
D
E
F
G
COR(X
S.D.
S.D.
MODIFY WEIGHTS
CHANGE LAMBDA'
LINEAR REGRESSION
PLOT FOE CONTOURS
TRANSFORMATION
EXIT TO UNIVAR
X & Y RANGE FACTORS
PLOT FULLSCREF.N
EXPAND PLOT SCALE
RESET PLOT SCALE
EDIT DATA POINTS
PLOT HSTGRM(X)
1.7844
UNIVAR OF Y
QUIT & EXIT
-7. ISOO-
J
».'
y
a
1
I
,
' *
:3
:,
•~'t
•
I \ 8 1
'"
199.75 300 37
WtS • 147 OF 147
,Y) = 0.100
(X) - 0.751
(Y) - 4.93
r:
i
B
P
L.
r.i
G
T
H
1 J07 K
\
X • 701.04
T - S8.72I
Figure 9. Frequency diagram for sample number plotted against fiber length (Y axis micron scale)
-------
Each number or letter of Figure 10 indicates the number of data points
at a specific sample index-number, length in microns, coordinate. For exam-
ple the letter C, at the mean fiber length y coordinate 1.3925 of leftmost
group 200 indicates, since C is the third letter of the alphabet, that 9+3
= 12 (the number of non-zero digits + the position of C in the alphabet)
points takes on the value 1.3925 within the group indexed by the number 200
(which corresponds to the before treatment subgroup).
Note that after the grouping option has been executed, there are 123-1 =
122 points in the second group (as indicated on the third line from the
bottom) due to the removal of the possible outlier.
With the exception of the one possible outlier, whose removal was pre-
viously described, Figure 11 is the arithmetic scale equivalent of Figure 6.
Note that the median of 0.0 appearing in the second column of Figure 4 cor-
responds to the value exp(O) = 1 of Figure 11.
The slight, but not necessarily statistically significant, shift to
larger fiber lengths after flow through A/C pipe is illustrated in Figures
11, 12 and 13.
15
-------
VIEW/EDIT DATA
CTRL
1 GET NEW DATA
2 -H-VIEW/EDIT DATA
3 MODIFY WEIGHTS
4 CHANGE LAMBDA
5 LINEAR REGRESSION
6 PLOT PDE CONTOURS
7 TRANSFORMATION
8 EXIT TO UNIVAR
9 X 4 Y RANGE FACTORS
A PLOT FULLSCREEN
B EXPAND PLOT SCALE
C RESET PLOT SCALE
D EDIT DATA POINTS
E PLOT HSTGRM(X)
F UNIVAR OF Y
G QUIT & EXIT
SAMPLE MUMBER
199 7S
2«0
wre - i
37
Figure 10. Frequency diagram after elimination of possible outlier (Y axis micron scale)
-------
UNIVARIATE
CTRL
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F
G
GET NEW DATA o 5491*
VIEW/EDIT DATA
HSTGRAH
CHOOSE AB
RESET AB
++PROB. DENSITY * Xe'°'3
CUM. CURVE
R(X) FCTN
SVIVAL CURVE
LOG XFORM
e 18305
LKH SVIVAL
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVARo ^^
T
T
T
T
T
T
1
1
XX
I ** ' *X
* X
* X
* I X
X X
* X
* T X
X X
X IX
X
1 X
T X
T X
1
T X
T
I X
T X
'T V"TT
T TXT
f I* \^ T fTTi jTrTTTTTfTTT
QUIT & EXIT '»"""> 2<"" "•**«?" ' "^?«r- '""
PARAMETERS EXCL CRP(X) INCL GRP(T)
SAMPLE 24 122
EST.MU 1.4320 0.81600
EST.SGMA 0.92951 0.52532
EST.MAX 0.42851 0.73218
MEDIAN 1.2500 1.0000
VALUE OF T - 4.5358
Figure 11. Estimated population density plots of EBMUD fiber length data (X axis in micron scale)
-------
00
UNIVARIATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 ++HSTCRAM
4 CHOOSE AB
S RESET AB
6 PROB. DENSITY
7 CUM. CURVE
8 R(X) FCTH
9 SVIVAL CURVE
A LOG XFORM
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
F EXIT SVL TO UNIVAR
G QUIT t EXIT •
281
28T
9t
ST
3T
21
II It 11 II 11
1 1 II II 1 1
420» 1 300 2 180 3 «60 3940 4.820 S 700 6580 7460 8340
08610 1.740 2.620 3500 4 3BO S 261 6140 7020 7900 8781)
PARAMETERS EXCL GRP(X) INCL GRP(T)
SAMPLE 24 122
EST.MU 1.A320 0.81600
EST.SGHA 0.92951 0.52532
NO. OF INTERVAL - 20 EST.WX 0.42851 0.73218
MEDIAN 1.2500 1.0000
Figure 12. Histogram of EBMUD fiber length data after A/C pipe (X axis micron scale)
-------
8X
UNIVARIATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 4+HSTGRAM
4 CHOOSE AB
5 RESET AB
6 PROB. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORH
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
T EXIT SVL TO UNIVAR
G QUIT & EXIT 1
NO. OF INTERVAL - 20
2X
3X 3X
5X
2X
IX
4211 1300 2189 3 060 3941 4 820 5700 6580 7468 8340
• 8600 1741 2.620 3500 4381 S 269 6141 7020 7900 8780
PARAMETERS EXCL GRP(X) INCL CRP(T)
SAMPLE 24 122
EST.MU 1.4320 0.81600
EST.SGMA 0.92951 0.52532
EST.MAX 0.42851 0.73218
MEDIAN 1.2500 1.0000
Figure 13. Histogram of EBMUD fiber length data before A/C pipe (X axis micron scale)
-------
SECTION 4
SAN FRANCISCO WATER DEPARTMENT HETCH-HETCHY (HH)
BEFORE AND AFTER FLOW THROUGH ASBESTOS CEMENT PIPE
Before proceeding to data analyses which produced positive findings, it
seems appropriate to present an analysis which, possibly due to sample size
considerations, seems to contradict the trends suggested in the previous
section. In this instance, unlike the previous data, a larger sample was
available for after than for before AC piped water.
The cumulatives shown in Figure 14 differ substantially from their Fig-
ure 5 counterparts. Unlike those of Figure 5, Figure 14 cumulatives cross
each other. This finding would not be affected by use of a logarithmic or
any other conventional data transformation.
Two features which differentiate HH and EBMUD fiber length distribu-
tions are indicated by Figure 15. Despite the small sample size of
n = 15, one can probably infer that the before AC pipe fiber lengths are
associated with a two component distribution. It is also apparent that at
least before logarithmic transformation, both before and after AC pipe data
is highly clumped.
The skewed shape of the after AC pipe and the possibly two component
structure of the before AC pipe fiber length distributions is discernible
from the two histograms shown in Figures 16 and 17.
20
-------
CTRL
1
2
3
4
5
6
UNIVARIATE
GET NEW DATA f ^^
VIEW/EDIT DATA
HSTGRAM
CHOOSE AB
RESET AB
PROB. DENSITY » 58(00
7 ++CUM. CURVE
8
9
A
B
C
D
E
F
G
R(X) FCTN
SVIVAL CURVE
LOG XFORM
0 750110
LKH SVIVAL
HEIffi. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
0 0
x**1"""""* ,T'
/T
T
r
x" '
•*)&' T
x^
,'T'
x
/i TtT,intnimT"
" /'''
X |-
x r
X 1
r
x .
T
X
I
7
f
T
t*
1
*
T X
I
T X
X
X
X
X
r i •>
QUIT 6 EXIT ° ?r t<)e 2 4000 4 6000 6 pooo
PARAMETERS EXCL CRP(X) INCL CRP(T)
SAMPLE 263 15
EST.MJ 1.0800 1.1680
EST.SflMA 0.56163 1.3217
EST.MAX 0.64533 0.28742
MEDIAN 1.4000 1.2000
:»> <,•.»*
MAX DIF - 0.23763
Stn.PT(0.05) - .36103
Figure 14. Estimated cumulative plots of fiber length data from the
Hetch-Hetchy System (X axis micron scale).
-------
N>
N)
CTRL
1
2
3
*
5
864533.
UNIVARIATE
GET NEW DATA
0 48401
VIEW/EDIT DATA
HSTCRAH
CHOOSE AB
RESET AB
6 -H-PROB. DENSITY
7
8
9
A
B
C
D
E
F
G
0 32267
CUM. CURVE
R(X) FCTN
SVIVAL CURVE
LOG XFORM
LKH SVIVAL » 16133
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
OUIT S EXIT
x : '•••'"•'• !'' '-'-'"•'-
X
X
X
X
X X
X
X
X
X
X X
x *
X
x
x
/tT ''''I
TT TT x
T T
Tx r
T
X T
X T7 T
/ '\
T x x T 'T
x T x TT *
X T X T T
XX T ^QtXX'^x T T
TT TT'X yX"" IT X»
T'tTITfTT Xw^^X^X1*^ ^jOtXXXx, xfXXX)StIT y** Xxx'! Tj
»0 2 4(00 4 60CO ^"feoilO 'T'TTITTTITT^ "
PARAMETERS EXCI. GRP(X) INCL GRP(T)
SAMPLE 263 15
EST.MU 1.0800 1.1680
EST.SGHA 0.56163 1.3217
EST.MAX 0.64533 0.28742
MEDIAN 1.4OOO 1.2000
VALUE OF T - -.53217
Figure 15. Estimated population density plots of fiber length data
from the Hetch-Hetchy System (X axis micron scale).
-------
UNIVARIATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 -H-HSTGRAM
4 CHOOSE AB
5 RESET AB
6 PROB. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
F EXIT SVL TO UNIVAR
G QUIT & EXIT
NO. OF INTERVAL - 20
S9X
5X
2X 2X 2X
> I I I
1 7-40 2620
PARAMETERS
SAMPLE
EST.MU
EST.SGMA
EST.MAX
MEDIAN
3 500
43*0 5260
EXCL CRP(X)
263
1.0800
0.56163
0.64533
1.4000
6 MO 7 020 7900
INCL GRP(T)
15
1.1680
1.3217
0.28742
1.2000
Figure 16. Histogram of fiber length data after A/C pipe from Hetch-
Hetchy System (X axis in micron scale).
-------
ST
UNIVARIATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 -M-HSTGRAM
4 CHOOSE AB
5 RESET AB
6 PROS. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
F EXIT SVL TO UNIVAR
G QUIT & EXIT *
27
2T
IT IT IT IT IT IT
•1206 300 2180 3(60 3940 4820 S 700 6580 7460 8340
0.86(0 1.740 2620 3500 4380 S 260 6140 7020 7.9(10 8780
PARAMETERS EXCL CRP(X) INCL CRP(T)
SAMPLE 263 15
™ „* THTFBVAI . in EST.MU 1.0800 1.1680
HO. OF INTERVAL 20 EST.SGMA 0.56163 1.3217
EST.MAX 0.64533 0.28742
MEDIAN 1.4000 1.2000
Figure 17. Histogram of fiber length data before A/C pipe from Hetch-
Hetchy System (X axis in micron scale).
-------
SECTION 5
COMBINED BEFORE AND AFTER AC PIPE SAMPLES
The complex distributional forms illustrated in the previous two sec-
tions suggested that larger sized samples were required to produce defini-
tive statistical conclusions. In this section, EBMUD and HH data were
pooled in order to provide the sample size needed for detailed analysis.
After pooling EBMUD and HH data, the estimated cumulatives of Figure 18
were obtained which are almost identical in shape to those shown in Figure
5. Again the smaller fibers are more common for samples before AC pipe than
after AC pipe. Unlike the statistical analysis described in Section 3,
however, here the maximum cumulative difference substantially exceeds the
critical point and hence the difference is significant.
Figure 19 indicated a need for still more elaborate analysis of this
data as do the histograms shown in Figures 20 and 21.
In order to check on the previous finding and assess its statistical
significance the Crystal Springs raw water sample was substituted for the
before AC pipe pooled sample and a comparable analysis performed. Figure 22
is similar in form to Figure 18 and the difference of cumulatives is again
highly significant.
This comparison would indicate that there is a significant difference
between the fiber lengths of natural fibers and those found after AC pipe.
The fibers found after AC pipe have a significantly larger portion of long
fibers than the raw water fibers.
Because of the high slope of the cumulative, no density estimates were
associated with the cumulatives of Figure 22. It should also be noted that
the scale change needed to compare pooled after AC pipe data to Crystal
Springs raw water data caused the difference between the histograms of
Figure 23 and 20.
From the study reported in this section, it would appear that Crystal
Springs raw water and before AC piped HH and EBMUD water distributions show
similar differences with post AC piped HH and EBMUD water.
25
-------
UNIVARIATE
CTRL
1
2
3
4
5
6
7
8
9
A
R
C
D
E
F
G
GET NEW DATA „ 7500n
VIEW/EDIT DATA
HSTGRAM
CHOOSE AD
RESET AB
PROD. DENSITY ° S0001'
•f+CUM. CURVF.
R(X) FCTN
SVIVAL CURVE
LOG XFORM
0 2500B
LKH SVIVAL
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR „ „
^^n=~'^""m"!;
/^'T
/ x*
T X
' X
T
X
T
X
T X
X
T X
X
T
1 H
1 *
X
T
X
T
X
T
X
t
X
1 X
I »
X
1 X
X
x
- — • 1 .1
QUIT & EXIT " 7°"00 2"'""> •|F'ot"> 6SI"1"
PARAMETERS EXCL CRF(X) INCL GRP(T)
SAMPLE 287 137
EST.MU 1.0800 1.81600
EST.SCMA 0.54862 0.53110
F.ST.MAX 0.66302 0.71369
MEDIAN 1.4000 1.0000
MAX DIF - 0.22170
SIC. PT(0.05) - .14123
Figure 18. Estimated cumulative plots of pooled fiber length data (X
axis in micron scale, T = before, X = after).
-------
UNIVARIATE
CTRL
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F
C
GET NEW DATA
«.S3527
VIEW/EDIT DATA
HSTCRAM
CHOOSE AB
RESET AB
-H-PROB. DENSITY , 35685
CUM. CURVE
R(X) FCTN
SVIVAL CURVE
LOG XFORH
LKH SVIVAL • »*«•
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
0 1192 IE
'f
T
TX
xx
T T
X
X
T T
X
T
T X
X
Xx
X I
T X x
X
X T
X
x
X
X T
X
X
T
X
T
x x xx
X
T X
T
T X X
Ti r*T tn
XX1, TTXXx**1fjv |T
QUIT & EXIT I 26000 2-1000 V.fe'otO 6*80 1» "**" ''W g „„„„
PARAMETERS EXCL GRP(X) INCL GRP(T)
SAMPLE 287 137
EST.MU 1.0800 1.81600
EST.SGMA 0.54862 0.53110
EST.MAX 0.66302 0.71369
MEDIAN 1.4000 1.0000
VALUE OF T - 4.6816
Figure 19. Estimated population density plots of pooled fiber length
data (X axis in micron scale).
-------
UNIVARIATE
CTRL
00
1 GET NEW DATA
2 VIEW/EDIT DATA
3 -H-HSTGRAM
4 CHOOSE AB
5 RESET AB
6 PROB. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
37X
67X
S9X
28X
49X
F6X
8X
BX 5X
"llXW * ™ IK IX 2X 2X ^
F EXIT SVL TO UNIVAR
I 4200 1300 2.180 3060 3 940 4820 5700 6580 7160 8340
C QUIT & EXIT 0 86»0 I 74« 3 620 3500 4 380 5260 6 I40 7 020 7900 8780
PARAMETERS
EXCL GRP(X)
INCL GRP(T)
NO. OF INTERVAL - 20
SAMPLE
EST.MU
EST.SGMA
EST.MAX
MEDIAN
287
1.0800
0.54862
0.66302
1.4000
137
1.81600
0.53110
0.71369
1.0000
Figure 20. Histogram of pooled fiber length data after A/C pipe (X axis in micron scale),
-------
to
V0
UNIVAKIATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 -H-HSTGRAM
* CHOOSE AB
5 RESET AB
6 PROB. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
f SVL DENSITY
T EXIT SVL TO UNIVAR
G QUIT I EXIT
33T
291
9T
ST
4T
2T 2T 21 2T
IT II IT 1 III I IT
1 1 1 1 1 1 1 1
• 4700 1300 2180 3060 3940 4820 S 700 6 S80 7460 8340
» 8600 1741 1 620 3500 4 38« S 260 6 MO 7020 7900 8780
PARAMETERS EXCL GRP(X) INCL GRP(T)
NO. OP INTERVAL - 20 SAMPLE 287 137
EST.MU 1.0800 1.81600
EST.SCHA 0.54862 0.53110
EST.MAX 0.66302 0.71369
MEDIAN 1.4000 1.0000
Figure 21. Histogram of pooled fiber length data before A/C pipe (X axis in micron scale),
-------
FIBER LEN&TH
DNIVARIATE
CTRL
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F
G
GET NEW DATA
1 75000
VIEW/EDIT DATA
HSTGRAM
CHOOSE AB
RESET AB
PROS. DENSITY , Sg0tg
•H-CUM. CURVE
R(X) FCTN
SVIVAL CURVE
LOG XFORM
LKH SVIVAL 0 ZSOOO
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
00
QUIT & EXIT «.2W
yTTTTTT^ .^Yy^iOCXXJOOOOOOQOQWXW^
yi ^flOQQO^
/ /**
T X
T X
t' **
X
T
X
T
X
X
I
X
X
J x
X
I
X
X
IX
X
T
X
I*
X
01 3 4000 6 6900 9 8000 13 000
PARAMETERS EXCL GRP(X) INCL CRP(T)
SAMPLE 287 318
EST.MU 1.0960 1.84000
EST.SCMA 0.55532 0.44797
EST.MAX 0.66304 0.85480
MEDIAN 1.4000 1.0000
MAX DIP - 0.25901
SIG. PT (0.05) - .11073
Figure 22. Comparison of estimated cumulative plots of pooled fiber length data from raw water
and after A/C pipe (X axis in micron scale, T = raw, X = after).
-------
UNIVARIATE
UJ
96X
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 -H-HSTCRAM
4 CHOOSE AB
5 RESET AB
6 PROB. DENSin
7 COM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B IJCH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
F EXIT SVL TO UHIVAR •*
0
G QUIT t EXIT
NO. OF INTERVAL • 20
S3X
76X
25x*"i
9X
"- "•rjT-!«^a. l~l-^
S20» 1800 3180 < 36« S M( 6920 8200 9-180 1076 12 IM
1 160 2 MO 3720 -B 000 6280 7560 8 8<0 10 12 1 1 «0 12 6*
PARAMETERS EXCL CRP(X) INCL CRP(T)
SAMPLE 287 318
EST.MU 1.0960 1.84000
EST.SGMA 0.55532 0.44797
EST.MAX 0.66304 0.85480
MEDIAN 1.4000 1.0000
Figure 23.
Histogram of pooled fiber length data after A/C pipe (X axis in micron scale)
-------
OJ
N>
UNIVARIATE
CTRL
1
2
3
4
5
6
7
8
9
A
B
C
D
e
F
C
GET MEW DATA
VIEW/EDIT DATA
•H-HSTCRAM
CHOOSE AB
RESET AB
PROB. DENSITY
CUM. CURVE
R(X) FCTN
SVIVAL CURVE
LOG XFORM
LKH SVIVAL
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
QUIT & EXIT
•
22T
481
I6T
31 31
1 1 IT IT IT IT II
5200 800 3 080 4 360 5 (H« B 920 8 200 9 481 10 76 1? 04
1 16t 2.4«« 3 726 5 000 6.280 7 56« 8 84« 10 12 II 40 12 68
PARAMETERS F.XCL CRP(X) INCL GRP(T)
NO. OF INTERVAL - 20 SAMPLE 287 31g
EST.MU 1.0960 1.84000
EST.SCMA 0.55532 0.44797
EST.MAX 0.66304 0.85480
MEDIAN 1.4000 1.0000
Figure 24. Histogram of raw water fiber length data (X axis in micron scale).
-------
SECTION 6
COMPARISON OF NARROW AND THICKER FIBERS FOR COMBINED
BEFORE AND AFTER AC PIPE SAMPLES
Each number, letter or symbol within the frequency diagram of Figure 25
indicates the number of fibers (of combined before and after AC pipe samples)
with a specific fiber diameter and length. Since an asterisk represents the
position of thirty-five or more fibers, Figure 25 indicates that many of the
narrowest fibers also seem to be fairly short. The possible existence of two
distribution subgroups, (as shown in Figure 15) led to the comparison of
narrow with the wide fibers with respect to their fiber length distributions.
As indicated in Figure 26, the bulk of pooled before and after AC pipe data
seemed to consist of narrow fibers (n = 337) 0.03 microns in width.
As shown in Figure 27, there seems to be a considerable difference
between the fiber lengths of the narrowest in comparison to wider fibers.
This comparison is made with the combined fiber data from before and after
AC pipe. These studies, which considered the narrowest and wider fibers
separately, tended to yield comparisons with high levels of statistical
significance.
33
-------
VIEW/EDIT DATA
CTRL
1 GET NEW DATA
2 ++VIEV/EDIT DATA
3 MODIFY WEIGHTS
4 CHANGGR LAMBDA
5 LINEAR REGRESSION
6 PLOT PDE CONTOURS
7 TRANSFORMATION
B EXIT TO UNIVAR
9 X & Y RANGE FACTORS
A PLOT FULLSCREEN
B EXPAND PLOT SCALE
C RESET PLOT SCALE
D EDIT DATA POINTS
E PLOT HSTGRM(X)
F RESTORE ORIC DATA
C UNIVAR OF Y
H QUIT & EXIT
10 100 ' *O
5
1 .670S"
-.90000
2
T
3
1
1
.J
6
r
J2
c
ifi
M
&
*
ffi
£R DIAME.TO?
:s'
1
C 4
1
13 2 1
2 1
1 1
•->
2 2 1
- 2 1 2
1
1 1
F, 2 4 1
t A ,1 J 3 A i5
5 1 3
A 1 2 6 3
4
;-,1 1 1
0.87SOOE-02 0 •1I9IOE-OI
KPTS • 4W OF
SPECIFY UP TO 4 INTERVALS FOR DEFINING A GROUP OR FOR
SPECIFICATION OF INCOMPLETE DATA.
INTERVAL X (LEFT) X (RIGHT) Y (BOTH) Y (TOP)
1 0.87SOOE-02 0.04000 -.90000 10.100
0.22135
Figure 25. Frequency diagram showing fiber diameter against fiber length.
-------
Xn
VIEW/EDIT; DATA
CTRL
1 GET NEW DATA
2 -H-VIEW/ED1T DATA
3 MODIFY WEIGHTS
4 CHANGE LAMBDA
5 LINEAR REGRESSION
6 PLOT PDE CONTOURS
7 TRANSFORMATION
8 EXIT TO UNIVAR
9 X & Y RANGE FACTORS
A PLOT FULLSCREEN
B EXPAND PLOT SCALE ,
C RESET PLOT SCALE
D EDIT DATA POINTS
E PLOT HSTGRM(X)
F RESTORE ORIC DATA
G UNIVAR OF Y
H /QUIT & EXIT
-.90000.
FIB
2
2
1
3
1
1
TT
6
C
C
1 M "
&
e
$
s
Eft DIAKETER
,51
: 3
2
:1
2
:e
1
6', 2 A
5 1
A 1 2
4
,1 ,
1
1
2 1
1
1 1
2
'2 1
1 2
1
4 1
A X A & 1 5
3
6 3
1
0.87SOOE-02 0.4I9IOE-OI
WtS
87 DATA POINTS IN GROUP,!, 337 IN CROUP 2
1 GO TO UNIVAR (-XHIT)
2 RESPECIFY GROUPS
Figure 26. Position of frequency diagram during the separation process.
-------
OJ
UNIVARIATE
CTRL
1
2
3
4
5
6
7
8
9
A
I
C
D
E
F
G
GET NEW DATA
1 7S»t«
VIEW/EDIT DATA
HSTCRAM
CHOOSE AB
RESET AB
PROB. DENSITY , 5,,,,
•H-CUM. CURVE
R(X) FCTN
SVIVAL CURVE
LOG XFORM
LKH SVIVAL « ZS«««
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
TTTTTTTTTnTTBWHxJx3ffi6wwDwWOnA**w
.T'T v*
T"" X
j******
/'
' x"
I X
X
T
X
T X
T X
X
T X
T x
X
t
X
T X
X
T
X
T X
X
T
X
t X
X
'x
. i 1 — 1
QUIT t EXIT (.7001 1 «•>«• 46110 68010 90000
PARAMETERS EXCL CRP(X) INCL CRP(T)
SAMPLE 87 337
EST.MU 1.3440 0.90400
EST.SGMA 0.77291 0.54632
EST.MAX 0.46868 0.67530
MEDIAN 1.6000 1.2000
MAX Dir - 0.26971
SIC. PT (0.05) - .16355
Figure 27. Comparison of cumulative fiber length data for narrow and
wide fibers (T - narrow; X - wide).
-------
SECTION 7
COMPARISON OF BEFORE AND AFTER AC PIPE
FOR NARROW FIBER GROUP
Because highly statistically significant results will be described in
this section, care was taken to protect against incorrect inference.
Hence both transformed and smoothed displays were utilized. The same find-
ing, that fibers tended to be shorter before than after AC pipe, was ob-
tained in all instances.
A nonparametric estimation smoothing technique described in (2] Section
5 was used to obtain Figure 29 using the same data utilized to obtain Figure
28. in both figures there is somewhat less of the ambiguous crossover ten-
dency of the estimated population cumulatives shown in Figure 18. In essence
Figures 28 and 18 would have been identical were it not for the removal of
thicker fibers as a preliminary to the construction of Figure 28. However,
much more striking results were obtained by using a logarithmic transforma-
tion as previously described in Section 2.
After using the logarithm of fiber length in place of fiber length
Plotted on an arithmetic scale the degree of distribution separation and the
cause of the ambiguous cumulative crossover tendency was clearly brought into
focus in Figure 30 and particularly 31. Before AC pipe narrow fibers tended
to be significantly shorter than after AC pipe fibers, which substantiates a
result first shown in Figure 5.
The histograms shown in Figure 32 and 33 demonstrate that although
before AC pipe fibers tend in general to be shorter than after AC pipe
fibers, a few, in this case about four before AC pipe fibers, are longer
than their after AC pipe counterparts. This slight aberration is also
illustrated by the upward trend of the right tail of the leftmost curve
shown in Figure 31.
37
-------
FlBtR Ifc'NGTh
UNIVARIATE
CTRL
1
2
3
4
5
6
7
8
£
A
B
C
D
E
F
G
GET NEW DATA
0 TSIOt
VIEW/EDIT DATA
HSTGRAM
CHOOSE AB
RESET AB
PROB. DENSITY
0 5*000
++CUM. CURVE
R(X) FCTH
SVIVAL CURVE
LOG XFORM
LKH SVIVAL • J5»»o
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
0 0
•yyywtnnrinonr^vvw KMYI'¥YyyyKyYYT^¥V*^HHHiuUwt*
,i^\*r
T X
T X
T X
I *
X
t X
X
X
1 X
X
I
X
T X
f *
X
T
X
T
X
f *
X
'x
t *
X
X
• * 1
QUIT i EXIT , j,,1,, j .,,„, « 6000 6 8000
PARAMETERS EXCL GRP(X) INCL GRP(T)
SAMPLE 233 104
EST.MU 1.0800 0.72800
EST.SGHA 0.54112 0.44742
EST.MAX 0.69943 0.83712
MEDIAN 1.2000 1.000
MAX DIF - 0.26500
SIC. PT (0.05) - .16038
9 0000
Figure 28. Comparison of cumulative fiber length data before and after A/C pipe for
the narrowest fiber group (X axis scaled in microns).
-------
vO
UNIVARIATE
CTRL
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F
0
GET NEW DATA
« 75099
VIEW/EDIT DATA
HSTURAM
CHOOSE AB
RESET AB
PROB. DENSITY ^ S|||((|
•H-CI1M. CURVE
R(X) FCTN
SVIVAI. CURVE
LOO XFORM
LKH SVIVAI, 1 75«0»
HEUR. SVIVAL
CUM. HAZAKD
SVt. DENSITY
EXIT SVL TO UNIVAR
x*«*«^nn^ :"" - ' "' WWrHM^"""" " " '""" ' '
.ff^5''1"""""""
X
T X
/ *
T X
I X
T X
1 X
1 X
r "
X
r x
X
I
X
1 X
I X
T X
1 «
1 X
1 X
I X
I «
/
X
X
III
QUIT 4 EXIT , 7n70, 3 ^ofti) 4 r.nno 6 anno q nnnn
PARAMETERS KXCL ORP(X) INCL GRP(T)
SAMPLE 213 104
EST.MU 1 .08OO 0.81600
EST.HOMA 0.65790 0.65316
EST.MAX (1.57700 0.60065
MEDIAN 1.2000 1.0000
MAX. DIP • 0.20610
SIC. PT ((1.05) • .1603R
Figure 29. Comparison of cumulative fiber length data before and after A/C pipe for
the narrowest fiber group using a smoothing technique (X axis scaled in microns).
-------
CTRL
1
2
3
4
*
6
UNIVARIATE
GET NEW DATA
VIEW/EDIT DATA ' /?"°"
HSTCRAM
CHOOSE AB
RESET AB
PROB. DENSITY
e soneo
7 ++C1IM. CURVE
ft
9
A
B
C
D
E
F
r.
R(X) FCTM
SVIVAL CURVE
1.00 XFORM
1 KH SVI VAI
0 25&fl«
HEIIR. SVIVAL
CUM. HAZARD
SVI. DENSITY
EXIT SVL TO UNIVAR
QUIT & EXIT 0 " J
,,**'"**
.»
f/'
If
'''
» l'
* I
«' ,t'''
/ _,/''
"" T1'"'
«vTl'T
»''
I1'
,''•
.1 «
T I
T1 *
I1 «
T ji
1 X
T! K
I1 <
l' »
l' «
,' *
/
/i?I/ ,X
1IJMXW-11''
PARAMETERS F.XCI. GRP(X) INCI. CRP( D
SAMPLE 233 |04
• F.ST.Mll 0.29389 -.20097
EST.SCMA O.A09&8 0.5929ft
EST.MAX 0. 65405 O.(h?072
MEDIAN 0.18232 0.0
MAX DIF - 0.22195
SIC. PTtO.O1)) - .1603B
Figure 30. Comparison of cumulative log fiber length data befor* and after A/C
pipe for the narrowest fiber group (X axis in log amicron scale).
-------
UNIVARIATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 HSTGRAH
4 CHOOSE AB
5 RESET AB
6 4-fpROB. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B IJCH SVIVAI.
C HF.UR. SVIVAL
P CUM. HAZARD
r. SVL DENSITY
F EXIT SVL TO UNIVAR
G QUIT * EXIT
1 49154
• 32792
1 I63SI
0 t
-1 6«
/ "x
,' I X X
r Tx x
1 * i x
/ * ' x
I * ' »
,
>x T
1 * I X
T X „
' X
T
t X ,
X
I ' I
X
tx r
IX T X
, X T «
, X IX
T x 'x
X T v
I X
x r
I *
i i
T X
1 r i
IX 1 "
, x , x
, x r «
' x i x
IX ,
T I "
, ,* I X
' / '', %
. K* 'T x
XX Tl »Tf
^ ' *
M - 6377/ 0 TOJW | 2^76
PARAMETERS EXCI. GRP(X) INCI, fTRP(T)
SAMPLE 213 104
EST.Mll 0.29389 -.20097
EST.SOMA 0.60966 0.59296
EST.MAX 0.65405 0.63072
MEDIAN 0.18232 0.0
,11'"""
VAl.tlE OF T - 0.9409
Figure 31. Comparison of estimated fiber size population density curves of fiber
length A/C pipe for narrowest fiber Kroup (X axis in nicron scale).
-------
UNIVARJATE
CTRL
I GET NEW DATA
2 VIEW/EDIT DATA
3 ++HSTGRM1
4 CHOOSE AB
5 RESET AB
6 PROB. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
F EXIT SVL TO UNIVAR
C QUIT t, EXIT
-I.SM -1 131 - 7S?9 - 3/23 0 8393E-Cff 3891 0 7597
1 150
1531
.9133 -.5676 - I8I9 0 I987 0.5791 0.9601 1.311 1 721
PARAMETERS EXCL GRP(X) INCL GRP(T)
NO. OF INTERVAL - 20
SAMPLE
EST.MU
EST.SGMA
ESTVMAX
H En I AN
233
0.29389
-0.60966
0.65405
0.1R232
104
-.20097
0.59296
0.63072
0.0
Figure 32. Histogram of fiber length data for narrowest fiber group after
A/C pipe (X axis in log micron scale).
-------
UNIVARIATE
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 -H-HSTCRAM
4 CHOOSE AB
5 RESET AB
6 PROB. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
1 LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENS in
F EXIT SVL TO UNIVAR 1 1
C QUIT t EXIT ., 5M .
2t
•T
12T
I7T
III
91
I2T
3T
2T 21 2t ZT
11 IT 11
1 134 - 7529 - 3723 t 8393E-II 3*91 1.7697 1 151 1 531 1 912
-1 334 - 9433 - 56% - 1819 0 19B7 I S794 t 96* 1 I 341 1 771 2 102
NO. Or INTERVAL - 20
PARAMETERS EXCL CRPU) INCL GRP(T)
SAMPLE 233 104
EST.MU 0.29389 -.20097
EST.SCMA 0.60966 0.59296
EST.MAJt 0.65405 0.63072
MEDIAN 0.18232 0.0
Figure 33. Histogram of fiber length data for narrowest fiber group before
A/C pipe (X axis in log micron scale).
-------
SECTION 8
COMPARISON OF BEFORE AND AFTER AC PIPE SAMPLES
FOR THE WIDER FIBER GROUP
There were only eighty-seven fibers which were not members of the narrow-
est fiber group described in the last section. Therefore, the fact that the
fiber length distributions for these wider fibers were not significantly dif-
different before and after AC pipe does not necessarily imply that there is
not a length distribution difference between the two groups of wider fibers.
Figure 34, when compared with either Figure 28 or 29, demonstrates that
there is far less observable difference, before and after AC pipe, within
the wider fiber group than within the narrowest fiber group. This finding
may of course be attributed to the smaller sample sizes available for the
wider fiber group. In fact when one uses a log transform, a shift in size,
similar to that observed in Figure 29, is apparent. (Of course, since a
transformation does not affect the Kolmogorov-Smirnov test, the observed
difference is still not statistically significant).
44
-------
CTRL
1
2
3
4
5
6
1 . OOVO
UNIVARIATE
GET NEW DATA
1 rstlt
VIEW/EDIT DATA
HSTGRAM
CHOOSE AB
RESET AR
PROB. DENSITY
I stitn
7 -t-tCUM. CURVE
8
9
A
R
C
D
E
F
n
R(X) FCTM
SVIVAI. CURVE
ijnc xroRx
UCH SVIVAL I 75IIO
HEUR. RVIVAJ.
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
QUIT 4 EXIT " '
1 ?•"
ri ii r i nil n i T 1 1 ( i n 1 1 it ij |
x^i»&*x«>oixx»xx*»»>1("11"'
xxxxwooooox""^''''
*«x/T
« T'
irf»T"""""1
ff
rX
/
t"
TX
'X
1
T X
T
X
1
1 X
I
X
1
' X
I
1 X
,' x
1 II
T
T *
X
X
jf
nn i 4gni 4 r.ooo e gooo '< oeon
PARAMETERS FJfCI. KRP(X) INCI. CHP(T)
SAMPLE 54 31
F.ST.Mtl 1.4120 1.3440
EST.SCMA 0.70449 1.1423
P.ST.MAX 0.51531 0.31596
MEDIAN 1.8000 1.5000
MAX. DIF - O.15544
Sin. FT (0.05) - .30050
Figure 34. Comparison of cumulative fiber length data before and after A/C pipe for the wider
fiber group (X axis in nicron scale; T = before; X - after).
-------
UNIVARIATE
CTRL
1
2
3
4
5
6
7
8
9
A
B
C
D
E
r
G
GET NEW DATA
t 7S*M .
VIEW/EDIT DATA
RSTCRAH
CHOOSE AB
RESET AB
PROB. DENSITY , 5,,,,
++CUM. CURVE
R(X) FCTN
SVIVAL CURVE
LOG XFORM
IKK SVIVAL • 25**'
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
I t 1
^
/^
X jT
XT'"
/'
,!.*
A
T'«
T X
T X
t X
' X
I
t «
T X
I »
T X
1 X
t x
T X
1 x
t1 x
,1 x*
I1'' *
nil'1'1' „"*
..!?TT J1
/
™™n~,™»»~,n~,.n™™VYV««»XXXXXX
QUIT * EXIT -| &o<)4 . ftS/77 1 39M9 1 2456
PARAMETERS EXCL CRP(X) INCL RRP(T)
SAMPLE 54 33
EST.MU 0.48423 0.29389
EST.SGMA 0.59065 0.79918
EST.MAX 0.66294 0.49131
MEDIAN 0.58779 0.40547
MAX. DIF - 0.16075
S1G. PT (0.05) - .30050
Figure 35. Comparison of cumulative fiber length data before and after A/C pipe for the wider
fiber group (X axis in log micron scale; T = before; X = after).
-------
SECTION 9
RAW WATER AND COMBINED AFTER A/C PIPE FIBER
LENGTH COMPARISONS FOR THICK FIBERS
To obtain further confirmation that before AC piped fibers had a
different length distribution than after AC piped fibers, the data described
in Figure 22 was divided into two subgroups. Specifically the n = 233
narrowest out of the n = 287 total after AC pipe fibers and n = 271
narrowest out of the 318 total Crystal Springs raw water samples were
selected.
The estimated cumulative shown in Figure 36 repeat, with an even higher
level of significance, the general pattern described in Section 5. The
density estimates given in Figure 37 demonstrate that particular before
AC pipe short length fibers tend to be more common than their after AC
pipe counterparts. This finding can be explored in somewhat more detail
by means of the histograms given in Figures 38 and 39.
47
-------
UNIVARIATE
CTRL
1
2
3
4
S
6
7
8
9
A
B
C
D
E
F
CET NEW DATA
VIW/EDIT DATA ' 7S'"
HSTCRAM
CHOOSE AB
RESET AB
PROS. DENSITY
1 SIMO
•H-CUM. CURVE
R(X) FCTN
SVIVAL CURVE
I.OC XPORM
UH SVIVAL t ?s«M
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
I • j
-..•. TTTTT-lllIinllTI]
-1 TTt T "^W|OtitW'if'Oc'A"'.£t*
T *
T'' '"
T X
I X
I X
X
X
' X
I *
X
T X
X
' X
,
I
X
I "
X
1 X
I x
X
1
X
T
T' *
T X
T (
1 X
T X
X
' X
' /
r' x**
I ^x"
-^dx*""
QUIT 4 EXIT T r • *
1 6I»H - W,IH 1 47m 1 57M
PARAMETERS EXCL GRP(X) INCL GRP(T)
SAMPLE 233 271
EST.MU 0.26903 0.18572E-0]
EST.SCMA 0.61530 0.34404
EST.MAX 0.64837 1.1438
MEDIAN 0.18232 0.0
MAX DIP - 0.29213
SIC. PT (0.05) - .12150
Figure 36. Comparisons of pooled fiber length data after A/C pipe and raw water fiber length
data for the narrow fiber ranges (X axis in log micron scale). *
-------
L£N&TjT
UNIVARIATE
CTRL
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F
C
GET NEW DATA
1.85787
VIEW/EDIT DATA
HSTCRAM
CHOOSE AB
RF.SET AB
++PROB. DENSITY 1 S7I9I
CUM. CURVE
R(X) FCTN
SVIVAL CURVE
LOG XFORM
LKH SVIVAL * 3"5"*
HFUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
II
i
T '
I
T
T
I
tl
T T
I
I T
T X>******
/' \
x X
I * X
* I X
« X
t X „
x« x
x" i "
' X ' ,!', »
X 1,1 I *
XI I *
I X . I X
X ' X
XI ' X
I / ' IX
I X* 'l? \
1 IT 'i *X
•WW*"" . . "' 'Trr_)<>'x.jlLv.'*0
QUIT 4 EXIT -1 r*n" - SfiS8< I 17775 1 SPM'"''
PARAMETERS EXCL CRP(X) INCL GRP(T)
SAMPLE 233 271
F.ST.MU 0.2690J 0.18572E-01
EST.SCMA 0.61530 0.34404
EST.MAX 0.64837 1.1438
MEDIAN 0.18232 0.0
VALUE OF T " 5.73B9
Figure 37. Comparison of estimated population density curves for pooled fiber length data after
A/C pipe and raw water fiber length data for the narrow fiber ranges (X axis in log micron scale)
-------
UNIVARIATE
Ul
o
2 VIEW/EDIT DATA
3 4-t-HSTGRAM
4 CHOOSE AB
•> RESET AB
6PPOft nPMCITV
rKUtl • IJtMb 111
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOG XFORM
B IXH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
F EXIT SVL TO UNIVAR
G QUIT & EXIT
.IX
7X
1 SOS -1 088
_
296
18X
9X
15X
38X
26X
I6X
SX
l~n rn
II 1 hn
6702 - 2S2S 0 1647 0 SH71 0 9996 1 417 1 834 2 J52
8789 - 4615 - 44IME l>« 3734 0 7908 1 208 1 626 2 04J 2 461
PARAMETERS EXC1. CRP(X) INCL CRP(T)
SAMPLE 233 271
EST.MU 0.26903 0.18572E-01
EST.SGHA 0.61530 0.34404
EST.MAX 0.64837 1.1438
MEDIAN 0.18232 0.0
Figure 38.
Histogram of narrow fiber length data after A/c pipe (X axis in log micron scale)
-------
I'NIVARIATE
CTRL
1 CET NEW DATA
2 VIEW/EDIT DATA
3 -H-HSTGRAH
4 CHOOSE AB
5 RESET AB
6 PROB. DENSITY
7 CUM. CURVE
8 R(X) FCTN
9 SVIVAL CURVE
A LOO XFORM
B LKH SVIVAL
C HEUR. SVIVAL
D CUM. HAZARD
E SVL DENSITY
IT
F EXIT SVL TO UNIVAR |
4T
C QUIT & EXIT -1.505 -
391
151
301
*$
331
26T
1ST
SI
4" .,-
-2If~l ^ ,-H,
1 «8S - 6702 - 2S28 0 1647 0 5
-------
SECTION 10
RAW WATER AND COMBINED AFTER A/C PIPE FIBER
LENGTH COMPARISONS FOR THICK FIBERS
As shown by the fiber length cumulative given in Figure 40, there was
a non-statistically significant tendency for naturally occurring water to
have shorter fibers than after AC piped water, when only wide fibers were
considered.
Note that the sample sizes n = 54 and n = 47 given after the word SAMPLE
at the bottom of Figure 40, plus the n = 233 and n = 271 sample sizes of Fig-
ure 36 add respectively to n = 287 and n = 318, the sample sizes listed in
Figure 22.
52
-------
I 1000
1 . •D0D
UNIVARIATE
CTRL
1
2
3
4
5
6
7
a
Ln
UJ ,
A
B
C
D
E
F
G
GET NEW DATA
0 75000
VIEW/EDIT DATA
nSTCRAM
CHOOSE AH
RESET AB
PROS. DENSITY
» 50000
•H-CUM. CURVE
R(X) FCTN
SVIVAL CURVE
LOG XFORM
LKH SVIVAL 0 25(00
HEUR. SVIVAL
CUM. HAZARD
SVL DENSITY
EXIT SVL TO UNIVAR
01
TTmmmTTTT7TTTnnmmTmm^^
jT vX*
/ xXXXXXXXXXXX
T y«
r x
x
I X
1 /
' X
i jf
T?
T X-
X
T X
' X
T X
X
T
X
T
X
T
X
T
X
T
X
I
T X
T X
T X
X
X
X
X
_ . 1 1 1 t
QUIT & EXIT • 500 0 2 62SO 4 7500 6 87SO 9 0000
PARAMETERS EXCL GRP(X) INCL GRP(T)
SAMPLE 54 47
EST.MU 1.4350 1.3500
EST.SGMA 0.70349 0.75733
EST.MAX 0.51816 0.52220
MEDIAN 1.8000 1.4000
HAX DIP - 0.20690
SIC. PT (0.05) - .27130
Figure 40. Comparison of wide fiber length data after A/C pipe and wide fiber length data
data from raw water (X axis in micron scale).
-------
SECTION 11
DISCUSSION
The overall conclusions suggested by our study are as follows:
a) There seems to be very little difference between fiber length
distributions before and after water treatment at one site.
b) Before AC pipe samples tend to have shorter fibers than after
AC piped samples.
c) The above difference is more noticeable when the narrow fiber
subgroups are considered apart from their wider fiber counter-
parts.
There are several possible hypotheses concerning these conclusions. The
most reasonable hypothesis Is that fibers coming from the AC pipe interior
are of the commercial variety of asbestos and these fibers tend to be
longer than the fibers found in the natural water around San Francisco.
An alternate, less plausible argument could be made. Consider that larger
bundles of fibers may tend to fragment or chip as they pass through AC pipe.
In the electron microscope fiber measurement process one chunky "fiber" was
observed which might have been composed of a large number of narrower
fibers. Such a fiber could easily leave a trail of fragments along its
path through a long section of pipe.
As shown in Figures 41 and 42 there is a slight, but highly statis-
tically significant positive association between fiber length and width.
(Note the t value of 7.009 at the bottom of Figure 42). Now since a wider
particle would also tend to contain long fibers, the fragments left during
its passage through a long section of pipe would tend to be longer than
the average length of fiber entering the pipe.
54
-------
Wl
VIEW/EDIT DATA
CTRL
1 GET NEW DATA
2 44VIKW/EDIT DATA
3 MODIFY WIGHTS
4 CHANCE LAMBDA
S LINEAR REGRESSION
6 PLOT PDE CONTOURS
7 TRANSFORMATION
B EXIT TO UNIVAR
9 X & V RANGE FACTORS
A PLOT FULLSCREEN
B EXPAND PLOT SCALE
C RESET PLOT SCALE
D EDIT DATA POINTS
E PLOT HSTCRH(X)
F RESTORE ORIG DATA
G UNIVAR OF Y
H QUIT 6 EXIT
COR(X.Y) - 0.233
S.D. (X) - 0.317E-01
S.D. (V) - 0.585
F I L'.tiR D ! AHE FliR
£>7f
853-
1
1
1
1
•i
1
6
C
11 —
o
N
J
H
y
i s
_« i "A .y\
••-•
s
*
*
y
Tj
! 4
1
! 3
1
1
C 2!
1
_§2
F,7,
33
7 1t
J3
B
1
c|
2
(-2
1 1
-. 1SOOOE-OI 0 40230E-Ot
-|
T
2 1
1
1
•— 1
2
6 4 1
3 3
•••>
4 2
2 - 1
1 ^a a i. a? f
B .<
4 1
1
6 1
1
0 3C
1
ffl
P
'i
B
p
R
L
F
N
b
T
H
500
M>IS -1000 OF 1000
Figure 41. Frequency diagram of diameter and length for combined samples before A/C pipe
(maximum sample size that can be analyzed using GRAFSTAT).
-------
U1
LINEAR RGRSN
CTRL
1 GET NEW DATA
2 VIEW/EDIT DATA
3 MODIFY WEIGHTS
4 CHANCE LAMBDA
5 ++LINEAR REGRESSION
6 PLOT PDE CONTOURS
8 4+CONF. BANDS
9 PRED. BANDS
A QUIT & EXIT.
-2.0853.
-j, A-
-.ISOOOE-OI 0.40230E-OI
Y - 4.300 X 4 0.1847
STD ERROR OF ESTIMATE - 0.5693
T STATISTIC • 7.560
0.33700
Figure 42. Plot of linear regression for data shown in Figure 40 (o = .05 confidence band),
-------
REFERENCES
1. Kanarek, M.S. Asbestos in Drinking Water and Cancer Incidence.
Ph.D. Thesis, University of California, Berkeley, California, 1978.
377 pp.
2. Tarter, M.E. and R.A. Kronraal. An Introduction to the Impleraenta-
and Theory of Nonpararaetric Density Estimation. The American Stat-
istician 30(3):105-112, 1976.
3. Siegel, S. Nonpararaetric Statistics for the Behavioral Sciences,
McGraw-Hill, New York, New York 1956.
4. Tarter, M.E., E. 0. Rigsbee and J.T. Wong. Interactive Editing of
Biomedical Data. Computer Programs in Bioraedicine, 6:117-123, 1976.
5. Tarter, M.E. Implementation of Harmonic Data Analysis Procedures.
In: Proceedings of the Computer Science and Statistics: Eleventh
Annual Symposium on the Interface, Durham, North Carolina, 1978.
pp. 234-239.
57
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/1-79-020
3. RECIPIENT'S ACCESSION NO,
4. TITLE AND SUBTITLE
Data Analysis of Drinking Water Asbestos Fiber Size
5. REPORT DATE
May 1979 issuing date
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Michael E. Tarter
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Department of Biomedical and Environmental Health Science
University of California
Berkeley, California 94720
10. PROGRAM ELEMENT NO.
s 614B(d)
11. CONTRACT/GRANT NO.
Order No. CA-7-3036-J-I
12. SPONSORING AGENCY NAME AND ADDRESS
Health Effects Research Laboratory-Cincinnati, Ohio
Office of Research and Development
U.S. Environmental Protection agency
Cincinnati, Ohio 45268
13. TYPE OF REPORT AND PERIOD COVERED
Final 8/16/77-6/30/78
14. SPONSORING AGENCY CODE
EPA/600/10
15. SUPPLEMENTARY NOTES
16. ABSTRACT
A statistical study of asbestos fiber size characteristics was conducted using
data obtained from a variety of San Francisco Bay Area water systems. Particular
emphasis was placed on comparison of fiber length distributions for samples collected
from pre and post asbestos cement (AC) pipe systems. Significant differences were
detected between the fiber size distributions in samples of raw water and water col-
lected after a length of AC pipe. Little difference was detected between the fiber
size distributions of a raw water sample and a treated water sample. It was also
shown that before and after AC pipe, fibers in the water differed most significantly
in the length distributions of narrow fibers.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS C. COSATI Held/Group
Asbestos, Serpentine, Potable water
Health Effects
06 F
8. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (ThisReport)
Unclassified
20. SECURITY CLASS (This page)
Unclassified
10. OF PAGES
68
22. PRICE
EPA Form 2220-1 (Rev. 4-77)
58
U. S. GOVttNMENT PRINTING OfFICE: 1979-657-060/1665 Region No. 5m
------- |