3
-------
EPA 747-R-98-002
July 1998
COMPREHENSIVE ABATEMENT PERFORMANCE PILOT STUDY
VOLUME II: MULTI-ELEMENT DATA ANALYSES
Final Report
Technical Branch
National Program Chemicals Division
Office of Pollution Prevention and Toxics
U.S. Environmental Protection Agency
Washington, DC 20460
-------
DISCLAIMER
The material in this document has been subject to Agency technical and policy review.
Mention of trade names, products, or services does not convey, and should not be interpreted as
conveying, official EPA approval, endorsement, or recommendation.
FURTHER INFORMATION
Additional copies of this report can be obtained by calling the National Lead Information
Center at 1-800-424-LEAD. Information about other technical reports on lead can be found
through internet at the address: "http://www.epa.gov/lead".
This report is copied on recycled paper.
-------
AUTHORS AND CONTRIBUTORS
The study that led to this report was funded and managed by the U.S. Environmental
Protection Agency. The study was conducted collaboratively by two organizations under
contract to the Environmental Protection Agency, Battelle Memorial Institute and Midwest
Research Institute. Each organization's responsibilities are listed below.
Battelle Memorial Institute (Battelle)
Battelle was responsible for the design of the study, for producing the design
documentation and the Quality Assurance Project Plan, for developing training for the field
teams, for recruiting cooperators for the study, for providing team leaders for the field teams, for
auditing the field teams, for data management of combined study data, for auditing the study
data, for identifying the elements that were selected for analysis, for conducting the statistical
analysis of the data, and for writing the final report.
Midwest Research Institute (MRI)
Midwest Research Institute was responsible for participating in the planning for the
study, for writing certain chapters and appendices in the Quality Assurance Project Plan, for
designing and producing a vacuum device for collecting field samples, for developing training
for the field teams, for providing the technicians who collected the field samples, for auditing the
field teams, for conducting the laboratory analysis of the field samples, for identifying the
elements that were selected for analysis, for managing the data associated with the field samples,
for auditing the laboratory results, and for producing the multi-element data on which this report
is based.
U.S. Environmental Protection Agency (EPA)
The Environmental Protection Agency was responsible for managing the study, for
reviewing the design and the Quality Assurance Project Plan, for assessing the performance of
the recruiters and the field teams, for reviewing audit reports, for reviewing draft reports and for
arranging the peer review of the draft final report. The EPA Work Assignment Managers were
Samuel Brown, Benjamin Lim, and John Schwemberger. The EPA Project Leader was
John Schwemberger. The EPA Project Officers were Gary Grindstaff, Joe Breen, Jill Hacker,
Phil Robinson, and Sineta Wooten.
-------
TABLE OF CONTENTS
Page
EXECUTIVE SUMMARY vii
1.0 INTRODUCTION 1
1.1 STUDY DESIGN 1
1.2 DATA 3
2.0 ANALYSIS 6
2.1 COMPARISON OF ELEMENT CONCENTRATIONS FOR HOUSES AND SAMPLE
TYPES 6
2.2 DIFFERENCES IN MULTI-ELEMENT CONCENTRATIONS RELATED TO ABATEMENT
AND RENOVATION HISTORY 16
2.3 RELATIONSHIPS AMONG SAMPLE TYPES 25
2.3.1 Correlations Between Lead and the Other Elements 25
2.3.2 Bivariate Relationships Among the Elements 28
2.3.3 Multivariate Relationships (Principal Components) 38
3.0 PEER REVIEW 42
4.0 REFERENCES 42
APPENDIX A: SUMMARY OF MULTI-ELEMENT DATA A-1
A-1.0 MULTI-ELEMENT DATA LISTING A-3
A-2.0 GEOMETRIC MEAN CONCENTRATIONS BY SAMPLE TYPE AND UNIT A-10
APPENDIX B: DISTRIBUTION AND OUTLIER ANALYSIS FOR THE CAPS PILOT MULTI-ELEMENT
DATA B-1
B-1.0 INTRODUCTION B-3
B-2.0 LOGNORMAL ASSUMPTION B-3
B-3.0 CHARACTERIZATION OF MEASUREMENT RELIABILITY B-4
B-4.0 OUTLIER ANALYSIS B-6
B-4.1 DATA GROUPING B-6
B-4.2 UNIVARIATE OUTLIER TEST B-6
B-4.3 MULTIVARIATE OUTLIER TEST B-7
B-4.4 RESULTS OF OUTLIER ANALYSIS B-8
LIST OF TABLES
Table 1. Abatement and Renovation History by House 2
Table 2. Abbreviations for Sample Types Used in Tables and Figures 4
Table 3. Summary of Planned Samples, Collected Samples, and Analytical Results Used in
Multi-Element Analysis 6
Table 4. Geometric Mean Concentration and Log Standard Deviation Across Houses by Sample
Type 8
Table 5. Results of Analysis of Variance to Test for Significant Differences Among Sample Types,
by Element 9
IV
-------
TABLE OF CONTENTS
(Continued)
Table 6. Model Estimates and Log Standard Errors of Geometric Mean Concentrations in
Unrenovated Unabated Houses 19
Table 7. Ratio of Element Concentrations in Renovated Homes to Concentrations in Unrenovated
Homes, Estimates and Log Standard Errors 20
Table 8. Ratio of Element Concentrations in Abated Homes to Concentrations in Unabated Homes,
Estimates and Log Standard Errors 21
Table 9. Estimated Correlation Between Lead and Remaining Elements, by Sample Type
(Log-transformed concentrations)* 27
Table 10. Principal Components for Model Parameter Estimates (Adjusted House Averages,
Abatement History, and Renovation History) 40
Table A-1a. CAP Pilot Study Multi-Element Data, House 17 A-4
Table A-1b. CAP Pilot Study Multi-Element Data, House 19 A-5
Table A-1c. CAP Pilot Study Multi-Element Data, House 33 A-6
Table A-1d. CAP Pilot Study Multi-Element Data, House 43 A-7
Table A-1e. CAP Pilot Study Multi-Element Data, House 51 A-8
Table A-1f. CAP Pilot Study Multi-Element Data, House 80 A-9
Table A-2. Geometric Mean Concentration by Sample Type and Unit A-11
Table B-1. Test of Normality: Log-transformed and Untransformed Data3 B-4
Table B-2. Log Standard Deviation and Measurement Reliability of Measured Concentrations in
Side-By-Side Dust Samples Collected from Floors and Window Stools B-5
Table B-3. Univariate Outliers Detected by Univariate Methods B-9
Table B-4. Outliers Detected by Multivariate Methods B-10
LIST OF FIGURES
Figure 1a. Lead Concentration vs. Sample Type (Geometric House Mean) 10
Figure 1b. Barium Concentration vs. Sample Type (Geometric House Mean) 10
Figure 1c. Cadmium Concentration vs. Sample Type (Geometric House Mean) 11
Figure 1d. Calcium Concentration vs. Sample Type (Geometric House Mean) 11
Figure 1e. Magnesium Concentration vs. Sample Type (Geometric House Mean) 12
Figure 1f. Nickel Concentration vs. Sample Type (Geometric House Mean) 12
Figure 1 h. Aluminum Concentration vs. Sample Type (Geometric House Mean) 13
Figure 1g. Zinc Concentration vs. Sample Type (Geometric House Mean 13
Figure 1i. Titanium Concentration vs. Sample Type (Geometric House Mean) 14
Figure 1j. Chromium Concentration vs. Sample Type (Geometric House Mean) 14
Figure 1 k. Potassium Concentration vs. Sample Type (Geometric House Mean) 15
Figure 2a. Estimated Average Log concentrations in Unrenovated, Unabated Units, for Each
Element and Each Sample Type. Elements Sorted by Geometric Average Concentration
in Boundary Soil 23
Figure 2b. Log Ratio of Element Concentrations in Renovated Homes to Concentrations in
Unrenovated, Unabated Homes, Sorted by Ratios in Boundary Soil 24
Figure 2c. Log Ratio of Element Concentrations in Abated Homes to Concentrations in Unabated
Homes, Sorted by Ratios in Boundary Soil 25
Figure 3a. Window Channel House Mean Correlation Scatterplot 29
Figure 3b. Window Stool House Mean Correlation Scatterplot 30
Figure 3c. Foundation Soil House Mean Correlation Scatterplot 31
-------
TABLE OF CONTENTS
(Continued)
Figure 3d. Boundary Soil House Mean Correlation Scatterplot 32
Figure 3e. Floor House Mean Correlation Scatterplot 33
Figure 3f. Entryway Dust House Mean Correlation Scatterplot 34
Figure 3g. Entryway Soil House Mean Correlation Scatterplot 35
Figure 3h. Air Duct House Mean Correlation Scatterplot 36
Figure 3i. Bedcover/Rug/Upholstery House Mean Correlation Scatterplot 37
Figure 4. First Two Principal Components for Each Building Component, Plotted versus Each
Other for Unrenovated, Unabated Unit Mean Log-Concentrations, Renovation History,
and Abatement History 41
VI
-------
EXECUTIVE SUMMARY
This report presents the results of the statistical analysis of multi-element data collected
during a pilot study that preceded the Comprehensive Abatement Performance (CAP) Study.
The goal of the CAP Study was to assess the long-term efficacy of lead-based paint abatement.
The pilot study was conducted to test the sampling and analysis protocols for the full study.
For the multi-element analysis, concentrations of lead, as well as of aluminum, barium,
cadmium, calcium, chromium, magnesium, nickel, potassium, titanium, and zinc in dust and soil
samples were measured. Concentrations of barium, cadmium, chromium, titanium, and zinc
were measured because these elements were regarded as possible constituents of paint.
Concentrations of aluminum, calcium, magnesium, nickel, and potassium were measured
because these elements were regarded as likely to be found in soil.
The multi-element analysis was undertaken to determine whether relationships among
these elements could provide a "tracer" for identifying the sources and pathways of lead in
households. Pilot study data were used to 1) characterize the concentrations of lead, aluminum,
barium, cadmium, calcium, chromium, magnesium, nickel, potassium, titanium, and zinc
samples in household dust and soil; 2) measure the differences in these concentrations associated
with renovation and lead-based paint abatement; and 3) investigate the relationship among the
elements by sample type (i.e., samples taken from different locations).
Dust and soil samples from six houses in Denver, Colorado were studied. Two houses
were unabated (identified as relatively free of lead-based paint in Volume 1 of the CAP Pilot
report (US EPA, 1995)). These houses were labeled as "relatively free of lead-based paint"
because the lead loadings in paint usually did not exceed the criterion used to trigger abatement
in the HUD Abatement Demonstration. The remaining four houses were abated using removal
methods and/or encapsulation or enclosure methods. One house was abated using primarily
removal methods on the interior and primarily encapsulation or enclosure methods on the
exterior. Another house was abated using predominantly encapsulation or enclosure methods on
the interior and predominantly removal methods on the exterior. The other two houses were
abated by primarily the same method on the interior as the exterior (one removal, the other
encapsulation or enclosure). Hence most of the lead levels in the paint in the houses studied
were less than 1.0 mg/cm2.
VII
-------
A total of 109 vacuum dust samples was collected. Between 16 and 22 dust samples
were collected at each house from window channels (also called "troughs" or "wells"), window
stools (often referred to as "sills"), air ducts, floors, bedcovers/rugs/upholstery, and entryways.
A total of forty-eight (48) soil samples was collected. Eight samples were collected from each
house: from just outside the front and back entryways, at different locations along the
foundation, and at different locations on the property boundary.
All elements except for potassium and chromium had significant differences in
concentration levels across sample types. Lead, barium, cadmium, calcium, magnesium, nickel,
and zinc typically had higher concentration levels in dust samples than in soil samples.
Aluminum and titanium generally had higher concentration levels in soil samples than in dust
samples. Calcium was the element with the highest concentration in dust samples. Aluminum
had the highest concentration in soil samples.
Tests of hypotheses on the differences due to abatement and renovation resulted in far
more cases of significance for renovation than for abatement. There were thirteen (13) cases of
significant differences for renovation, considerably more than the number of cases that would be
expected by chance alone. For renovation effects, there were several cases of significantly higher
levels in interior dust for lead and for the elements calcium, magnesium, and nickel. Also for
renovation effects, there were cases of significantly lower concentrations in soil sample types for
the elements aluminum, titanium, and potassium. For abatement effects, the number of cases of
significance was equal to the number that would be expected by chance alone. Significantly
higher concentrations of lead and zinc were the case for exterior entryway samples and lead was
significantly higher in interior entryway samples.
After controlling for differences between houses with different abatement and renovation
history, relative concentrations of the elements suggested the following grouping of sample types
in unabated, unrenovated houses: 1) boundary, foundation, and entryway soil samples, and 2)
entryway dust and bedcovers/rugs/upholstery, along with floors and window stools. Window
channels and air ducts did not appear similar to other sample types or each other. For renovated
houses, the three soil samples could be grouped together, and there were similarities between
floor and entryway dust samples, and to a lesser extent, between window channels and window
stools. For abated houses no groupings were clearly apparent.
VIII
-------
Other approaches were used to group sample types. There was no uniformly consistent
grouping of sample types, but some sample types were more likely to be clustered together than
others. In most groupings, either all three soil samples were in a cluster or two of the soil
samples, the foundation and boundary samples, were in a cluster. Typically entryway dust
samples and floor dust samples were in the same cluster, sometimes with other sample types as
well. Air ducts and bedcovers/rugs/upholstery were the sample types most likely to stand apart
from other groups of sample types when grouping approaches were carried out.
There were no definitive identifications of sample types with sources of lead. For
example, window channels were observed to contain high concentrations of lead in dust. Some
of the analyses in the report indicated that there were high levels of barium and zinc, as well as
lead, in the window channel samples. Since barium, zinc, and lead were used in paint, this might
indicate paint was the source of the lead. However, aluminum and titanium were also present at
high levels in window channel samples, and in this study, these elements appeared to be
identified with soil. This would indicate a soil source for the lead. Moreover, titanium was also
used in paint. Overall, the analyses in this report did not result in a definitive answer to the
source of the lead in the window channels.
IX
-------
This page intentionally left blank.
-------
COMPREHENSIVE ABATEMENT PERFORMANCE PILOT STUDY:
MULTI-ELEMENT DATA ANALYSES
1.0 INTRODUCTION
This report presents the results of a multi-element analysis of data obtained during a pilot
study that preceded the Comprehensive Abatement Performance (CAP) Study. This represents
Volume II of the CAP Pilot report. Volume I dealt exclusively with the statistical analysis of
observed levels of lead (US EPA, 1995). The goal of the CAP Study was to assess the long-term
efficacy of lead-based paint abatement. The pilot study was conducted to test the sampling and
analysis protocols that were intended for the full study. These protocols called for determining
the levels of lead in dust and soil samples collected at residential units. The intention of this
report is to summarize the results of an investigation of methods for examining multi-element
data and characterizing the multi-element relationships between different sample types in the
residences sampled.
1.1 STUDY DESIGN
In the CAP Pilot study, six houses of differing abatement histories were sampled. These
houses were located in Denver, Colorado. Two houses were unabated (previously identified as
relatively free of lead-based paint) (US EPA, 1995). The remaining four houses were abated
using removal methods and/or encapsulation or enclosure methods. One house was abated using
primarily removal methods on the interior and primarily encapsulation or enclosure methods on
the exterior. Another house was abated using predominantly encapsulation or enclosure methods
on the interior and predominantly removal methods on the exterior. The other two houses were
abated by primarily the same method on the interior as the exterior (one removal, the other
encapsulation or enclosure). For easy reference, Table 1 displays the abatement and renovation
history of each of the six houses sampled. (Renovation is described later.)
In the six houses, most of the lead levels in paint were less than 1.0 mg/cm2. This might
make it more difficult to develop hypotheses about sources of lead simply based on the levels of
lead observed in different sample types. However, the impetus behind the multi-element analysis
approach was the conception that patterns among different elements might reveal themselves in
different, nearby sample types.
-------
Table 1. Abatement and Renovation History by House
House
17
19
33
43
51
80
Interior
Abatement
History
Abated: Removal3
Unabated
Unabated
Abated: Removal
Abated: E/E
Abated: E/E
Exterior
Abatement
History
Abated: E/Eb
Unabated
Unabated
Abated: Removal
Abated: Removal
Abated: E/E
Renovation
None
Partial
None
None
Full
None
Abated by removal methods.
Abated by encapsulation/enclosure methods.
Along with the determinations of lead obtained in the study, levels often other metals
were measured within dust and soil samples taken at these houses: aluminum, barium, cadmium,
calcium, chromium, magnesium, nickel, potassium, titanium, and zinc. Five of these metals
(barium, cadmium, chromium, titanium, and zinc) have been used in the composition of paint.
The other five elements are likely to be found in soil (Tisdale, Nelson, and Beaton, 1985). For
example, magnesium is found in clay, which is often part of soil samples. The purpose of
measuring the levels of these other metals in the samples was to identify groups of sample types
that appear to have come from similar sources, with the ultimate goal of identifying prominent
sources of lead found in residences.
The major objectives addressed in the analysis of the multi-element data from the pilot
study were the following:
(1) Characterize the concentration levels of lead, aluminum, barium, cadmium,
calcium, chromium, magnesium, nickel, potassium, titanium, and zinc in samples
of household dust and soil;
(2) Determine the effect of renovation and abatement on the concentration of these
elements in household dust and soil; and
-------
(3) Investigate the relationships among these elements by sample type (i.e., samples
taken from different locations).
The intention of this examination was to identify analysis methods for evaluating multi-
element data and to apply these methods to pilot study data to identify any relationships. With
data available for only six housing units, few relationships were strongly detectable.
Subsection 1.2 describes the data and gives a summary of the outlier analysis. Section 2
describes the analyses performed and the results. Section 3 is a summary of the key points of the
peer review of this report. Section 4 lists references. Appendix A contains tabulations of the
data used in the analyses in the report. Appendix B contains technical analyses related to the
distribution of the data, the reliability of the measurements, and the identification of outliers.
1.2 DATA
The study design intended the collection of 25 vacuum dust samples and eight core soil
samples from each of the six houses in the study, for a total of 150 dust samples and 48 soil
samples. The vacuum dust samples were collected from six different locations (window
channels1, window stools2, air ducts, floors, bedcovers/rugs/upholstery, and entryways). Core
soil samples were composite samples of three subsamples. They were taken from just outside the
front and back entryways, at different locations on the foundation, and at different locations on
the property boundary. The number of dust samples actually collected from each house varied
from 16 to 22 for a total of 109 vacuum dust samples. Eight soil samples were collected from
each house for a total of 48 soil samples.
Table 2 contains a description of the acronyms used throughout this report in the tables
and figures to denote the components from which samples were collected (referred to hereafter as
"sample types").
Window channel: The surface below the window sash and inside the screen and/or storm window. Also
called the window trough or the window well.
Window stool: The horizontal board inside the window that extends into the house interior—often called
the window sill.
-------
Table 2. Abbreviations for Sample Types Used in Tables and Figures
Media
Vacuum Dust Samples
Soil Samples
Acronym
ARD
BRU
EWY (-1)
FLR
WCH
WST
BDY
EWY (-0)
FDN
Component/Sample Type
Air duct
Bedcover/rug/upholstery
Entryway (-Inside)
Floor
Window Channel
Window Stool
Boundary
Entryway (-Outside)
Foundation
The dust and soil samples collected during the pilot study were analyzed to determine the
amounts of eleven different elements. Listings of the raw element concentration data are
displayed in Tables A-la through A-If of Appendix A. Each table displays concentrations from
a given house for each of the eleven elements by sample medium, sample type, location, and
sample ED. House number and sample ID uniquely identify each sample. Only element
concentrations (ng/g) were analyzed for this report. Element loadings (ug/ft2) were also
measured for dust samples. However, element loadings are influenced by dust amount, while
concentrations are not. Element loading relationships might be masked by differences in
household cleaning habits. Therefore, loadings were not considered in this analysis.
The samples were prepared using a modified version of EPA SW846 Method 3050. The
modifications were to reagent volumes and final dilution volume. Samples were analyzed by
inductively coupled plasma-atomic emission spectrometry using EPA ITD Method 1620. The
lower reporting limit for all the data was the instrument detection limit. For each batch analysis
an instrument detection limit was calculated. Instrument detection limits were based upon three
times the standard deviation of five determinations of a laboratory fortified blank. The upper
reporting limit was based upon the highest calibration standard used to calibrate the laboratory
instrument.
Twenty-three samples had zinc concentrations above the calibration range of the
measuring instrument. One sample had a cadmium concentration above the calibration range.
For the 23 samples with elevated zinc concentrations, the maximum detectable concentration was
-------
used, after correcting for its dilution factor3. These adjusted values were used in the statistical
analysis and are identified by superscripts in the appendix tables. Because only one sample had a
cadmium concentration above the calibration range, it was excluded from the statistical analysis,
rather than adjusted by its dilution factor.
Table 3 summarizes the numbers of dust (vacuum) and soil samples planned and
collected, the numbers of extra samples collected, the numbers of analytical results reported, and
the numbers of samples included in this multi-element data analysis. Results for seven of the
109 dust samples collected were excluded from the statistical analyses. No soil samples were
excluded. One of the seven dust samples omitted was the sample with the elevated cadmium
concentration described in the previous paragraph (sample 7 in house 17, see Appendix A-l for a
data listing by house and sample number). Another sample (sample 12 in house 19) was dropped
in the laboratory. Four samples (samples 3, 9, and 17 in house 19 and sample 19 in house 43)
were eliminated because only lead concentrations were available due to calcium interference.
Finally, sample 12 in house 51 was excluded due to sampling problems; the cartridge filled with
sawdust prior to completion of the sample collection.
Univariate and multivariate outlier detection tests were applied to the multi-element
concentration data. These two tests were applied to natural logarithms of the concentrations of
the eleven elements. The univariate test is mainly aimed at identifying individual samples with
element concentration outside the range of what is typical. The multivariate test does this also,
but in addition, the multivariate test seeks to identify unusual combinations of different elements.
Lists of potential outliers were sent back to the laboratory for verification. The results for all but
one of the potential outliers were confirmed and included in the analysis as originally reported.
The sample for which an error was reported was updated and the corrected value was used in the
analysis. This sample is documented in the footnotes to Table A-lb. Details regarding the
statistical approach to the outlier analyses and their respective results are provided in
Appendix B.
The maximum detectable concentration was 5 ug/mL. The reported concentration depended on the
actual amount of dilution prior to chemical analysis.
-------
Table 3. Summary of Planned Samples, Collected Samples, and Analytical Results
Used in Multi-Element Analysis
Medium
Vacuum
Dust
Soil
Type of Sample
Regular
Vacuum-Wipe
Comparison
Side-by-side
(QC)
Total Dust
Regular
Side-by-side
(QC)
Side-by-side
(interlab
comparison)6
Total Soil
Planned
Samples to
be Collected
108
36
6
150
36
6
6
48
Planned
Samples
Collected
77
25
5
107
36
6
6
48
Extra
Samples
Collected
1
0
1
2
0
0
0
0
Analytical
Results
Reported
7o(a),(b)
25
6
104
36
6
6
48
Analytical
Results Used
in Data
Analysis
71(c),(d)
25
6
102
36
6
6
48
Sample 19-12 (house 19, sample 12) dropped in lab. No analytical results reported.
IPP analwcie hamnararl Kw /ȣ>l*tiiim in*ai-f Aran/^a -fnr comrtlAO 1 Q_m 1 Q_HQ 1Q-1~7 anrl A*3_1Q- nn multi-
(c)
(d)
(e)
element data reported.
Cadmium concentration was above the upper calibration limit for sample 17-07; excluded from the multi-
element analysis.
Cartridge for sample 51-12 filled with sawdust prior to completion of sample collection; sample excluded
from lead and multi-element data analyses.
These samples were split for analysis by two labs. The result obtained from the primary lab was included
in the multi-element analysis.
2.0 ANALYSIS
The analysis is divided into three parts corresponding to the three major objectives
introduced above. Section 2.1 contains a characterization of the concentration levels of the
different elements in the various sample types. Section 2.2 describes the estimated effects of
abatement and renovation, and Section 2.3 examines the relationships among the elements and
sample types.
2.1 COMPARISON OF ELEMENT CONCENTRATIONS FOR HOUSES AND
SAMPLE TYPES
A lognormal distribution was identified as a reasonable model for characterizing the
concentrations of all of the elements. An analysis leading to this decision is provided in
Appendix B. Thus, commonly used descriptive statistics, such as "mean and standard deviation"
-------
are replaced by the analogous terms "geometric mean and the log standard deviation" throughout
this document. Also provided in appendix B is a quantification of the measurement error
associated with characterizing concentrations of each of the eleven elements included.
Due to the general absence of room-level effects found in the analysis of the CAP pilot
lead data, the basic experimental unit considered in the multi-element data analysis is the house.
House geometric mean concentrations of the eleven elements were the basic quantities used in
the statistical analyses. These are tabulated in Table A-2 of Appendix A by sample type and
house for each of the eleven elements.
Levels of each of the eleven elements observed varied by sample type. Grand geometric
mean concentrations for each element are displayed in Table 4 by sample type. These were
obtained by taking the geometric mean of the house geometric means (displayed in Table A-2)
for each sample type and element. Thus, each house where a sample was taken (for a particular
sample type) is given equal weight in these averages. Each mean is followed by its log standard
deviation. This represents a measure of the between-house variation for that response without
controlling for abatement or renovation history (which are discussed in the next section).
Notice from Table 4 that three of the four sample types with the lowest lead
concentrations were soil samples. The sample types with the highest lead concentration were the
two window components and the air ducts, and these lead concentrations were at least twice as
high as those in the remaining sample types. Aluminum concentrations in soil sample types were
three of the four highest among the nine sample types. Dust samples in window channels had at
least twice the barium concentration as the remaining sample types. For cadmium, calcium,
magnesium, nickel and zinc, concentrations in soil samples were all lower than those in dust
samples. In particular, calcium and magnesium concentrations in dust are more than twice as
high as those in soil samples. One can also observe that chromium concentrations were all lower
in three soil sample types than in dust samples except in window channel dust sample type.
Except for magnesium and potassium, all element concentrations hi boundary soil samples were
lower than those in the foundation or entryway soil samples.
To quantify the degree of variation in the concentrations of each element across sample
types, an analysis of variance was performed on the geometric means for each house. The results
of this ANOVA are summarized in Table 5. For all elements except potassium and chromium,
the differences across sample types were statistically significant at the level of 0.01.
-------
Table 4. Geometric Mean Concentration and Log Standard Deviation Across Houses by Sample Type
Sample
Medium
Dust
Soil
Sample
Type
WCH
WST
ARD
FLR
BRU
EWY-I
EWY-O
FDN
BDY
No. of
Units
Sampled
4
6
5
6
5
6
6
6
6
Dust
Loading
(mgfff)
738
46.8
352
58.3
41.6
71.8
Lead
Geo Mean Log Std.
(ug/g) Oev.
2128 0.97
658 1.20
771 0.31
260 0.81
152 0.72
314 0.91
208 0.90
209 0.87
126 0.79
Aluminum
Geo Mean Log
fog/g) Std. Dev.
12940 0.39
6266 0.36
7136 0.32
6331 0.30
6248 0.47
10761 0.37
16058 0.33
14491 0.40
11373 0.42
Barium
Geo Mean Log
(ug/g) Std. Dev.
1647 1.58
703 1.16
325 0.60
295 0.52
254 0.45
294 0.78
276 0.21
257 0.31
166 0.31
Cadmium
Geo Mean Log
(ug/g) Std. Dev.
19.1 0.61
23.9 1.03
26.3 1.32
9.3 0.68
9.7 0.62
9.5 0.49
5.6 0.85
4.0 0.41
2.8 0.51
Calcium
Geo Mean Log
(ug/g) Std. Dev.
33730 0.23
53230 0.51
40465 0.61
25042 0.44
24598 0.51
32709 1.03
9814 0.40
9812 0.31
8576 0.20
Chromium
Geo Mean Log
(ug/g) Std. Dev
40.1 0.46
54.3 0.54
77.3 0.64
48.7 0.80
55.0 0.52
45.4 0.79
40.8 0.67
28.7 0.28
23.6 0.31
00
Sample
Medium
Dust
Soil
Sample
Type
WCH
WST
ARD
FLR
BRU
EWY-I
EWY-O
FDN
BDY
No. of Units
Sampled
4
6
5
6
5
6
6
6
6
Dust
Loading
(mglff)
738
46.8
352
58.3
41.6
71.8
Magnesium
Geo Mean Log
(ug/g) Std. Dev.
5553 0.32
4807 0.29
3877 0.42
3222 0.25
3094 0.29
4419 0.40
574 0.16
1054 0.66
636 0.39
Nickel
Geo Mean Log
(ug/g) Std. Dev.
24.0 0.35
38.0 0.37
40.7 1.17
27.8 0.60
45.0 1.02
20.7 0.36
13.9 0.74
11.4 0.27
9.7 0.30
Potassium
Geo Mean Log
(ug/s) Std. Dev.
2651 0.444
2818 0.67
4260 0.36
4311 0.70
4046 0.89
4045 0.67
4069 0.26
3476 0.32
3504 0.33
Titanium
Geo Mean Log
(ug/g) Std. Dev.
496 0.27
376 0.13
262 0.38
199 0.29
191 0.57
351 0.33
482 0.23
421 0.24
372 0.26
Zinc
Geo Mean Log
(ug/g) Std. Dev.
3226 1.07
1939 0.66
4458 0.98
770 0.39
656 0.70
722 0.49
296 0.37
372 0.35
178 0.46
-------
Table 5. Results of Analysis of Variance to Test for Significant Differences Among
Sample Types, by Element
Element
Pb
Al
Ba
Cd
Ca
Cr
Mg
Ni
K
Ti
Zn
Root Mean
Squared Error
1.14
0.48
0.95
1.01
0.67
0.78
0.48
0.77
0.74
0.38
0.76
F value
4.47
6.55
3.83
5.54
9.71
1.59
31.27
4.83
0.55
8.44
16.40
P value
0.0006
0.0001
0.0019
0.0001
0.0001
0.1570
0.0001
0.0003
0.8096
0.0001
0.0001
Comment
WCH, WST, and ARD had highest
concentrations; three soil sample type were
among the four lowest
WCH and WST concentrations were more
than twice as high as the remaining sample
types.
Soil all lower than dust
Insignificant differences
Soil all lower than dust, EWY lower than
FDN
Soil all lower than dust
Insignificant differences
ARD, WCH, WST higher than the rest
In interpreting differences in average concentrations across sample types, the reader
should remember that the houses have different abatement and renovation histories. These
effects are discussed later in the report. For example, calcium levels were significantly higher in
the renovated houses than in unrenovated houses for four sample types. Such effects impact the
average concentration across houses, and are not adjusted for in Figures la through Ik.
Figures la through Ik display geometric mean sample concentrations by house and
building component for lead, barium, zinc, aluminum, titanium, cadmium, calcium, chromium,
magnesium, nickel, and potassium. These figures display all the data considered in the analysis.
Mean sample concentrations for each house are plotted with different symbols. The grand
geometric mean concentrations over all houses are plotted with a circle and connected by a solid
line across sample types. The sample types are arranged according to increasing lead
concentration for all elements. The element concentrations summarized in Table 4 can be seen
in these figures. Therefore, the comparisons of grand geometric mean concentrations for all
-------
10000i
§ 1000
I
§
8 10°
s
10
Sample Type
House Legend + + + 17 * * * 19 ***33 00043
51
e-e-e gg
Figure 1a. Lead Concentration vs. Sample Type (Geometric House Mean)
10000
O)
I
3
I
1000
100
*
z
House Lagend
Sample Type
»»*33 00043
99
Figure 1b. Barium Concentration vs. Sample Type (Geometric House Mean)
10
-------
1000
I 100
E
I
Sample Type
House Legend + + + 17 xxx19 »**33 00043 Y Y Y 51 zzzgo ^-e-e 99
Figure 1c. Cadmium Concentration vs. Sample Type (Geometric House Mean)
1000000
•s
o
100000
10000
1000
House Legend
Sample Type
17 xxx 19 ***33 00043 YYYS1 ZZZ80
99
Figure 1d. Calcium Concentration vs. Sample Type (Geometric House Mean)
11
-------
10000-1
O)
o
3
-fc 1000
8
8
I
'53
I m
House Legend
Sample Type
«xx19 **»33 00043
Figure 1e. Magnesium Concentration vs. Sample Type (Geometric House Mean)
1000
o 100
I
o
10
Y Y
House Legend
Sample Type
»»*33 00043
zzzeo
99
Figure 1f. Nickel Concentration vs. Sample Type (Geometric House Mean)
12
-------
100000
$
C 10000
o
I
§
o
N
1000
100
Sample Type
House Legend + + + 17 * * x 19 * * * 33 00043 Y*Y51 zzzeo e-e-« 99
Figure 1g. Zinc Concentration vs. Sample Type (Geometric House Mean)
100000
o
1
•£ 10000
8
|
|
.£ 1000
I
House Legend
17
<* ^ /
Sample Type
'*»33 00043 Y
zzzeo e-e-e 99
Figure 1h. Aluminum Concentration vs. Sample Type (Geometric House Mean)
13
-------
1000
O
1
8
E
100
10
House Legend + + + 17
Sample Type
*»*33 00043 Y Y ^ 51 2 * * 80 e-^-e 99
Figure 1i. Titanium Concentration vs. Sample Type (Geometric House Mean)
1000
1
O 100
I
e
6 10
House Legend
Sample Type
***33 00043
zzzao e-e-e 99
Figure 1j. Chromium Concentration vs. Sample Type (Geometric House Mean)
14
-------
a 10000°
^
O 10000
-------
sample types discussed above can be observed from these figures. Furthermore, these figures as
well as the information in Tables 4 and 5 provide a tool for grouping elements based on the
pattern similarity.
Figures la through Ik are grouped according to similar profiles of element concentrations
across sample types. Three groups of elements were identified. The first group, consisting of
lead, barium, cadmium, calcium, magnesium, nickel, and zinc, generally had higher
concentrations in dust samples than in soil samples. For most of these elements, the highest
concentrations were usually found in window channels, window stools, or air ducts. The second
group, consisting of aluminum and titanium, generally had higher concentrations in soil than in
dust. The third group, consisting of chromium and potassium, had no significant differences in
concentration across sample types.
In summary, all elements except chromium and potassium had significant
differences in concentration levels across sample types. Three groups of elements were
identified: lead, barium, cadmium, calcium, magnesium, nickel, and zinc; aluminum and
titanium; and chromium and potassium. Aluminum was the most prominent element in soil,
and calcium was the element with the greatest concentrations in dust.
2.2 DIFFERENCES IN MULTI-ELEMENT CONCENTRATIONS RELATED TO
ABATEMENT AND RENOVATION HISTORY
The differences in element concentrations associated with abatement and renovation
history was assessed by fitting a statistical model containing terms for both renovation and
abatement to the data in Appendix A. The model fitted to data for each element was
where
Cj represents the observed average log-concentration in house j,
m represents the average log-concentration in unrenovated unabated houses,
a represents the added effect of abatement,
Ij 1 if house j was abated
0 if house j was an unabated house,
r represents the added effect of a full renovation,
16
-------
is the degree of renovation house j was undergoing at the time of sampling
(see below), and
represents house-to-house variation
House 51 was assigned an Rj value of 1 indicating "full renovation" and House 19 a value
of 0.5 indicating "partial renovation". The other four houses were assigned Rj values of zero,
indicating that no renovation was being performed. Although only one home received full
renovation, with one subject to partial renovation, it is necessary to consider its effect.4
In the analysis of the lead data, the method of abatement (E/E or removal) was also
considered as a factor in the statistical model. No significant effect was found; and therefore,
this effect was not included in the above lead model applied to all elements.
Estimates of the model parameters are reported in Tables 6, 7, and 8. Table 6 contains
estimates and log-standard errors of the geometric mean concentration of each element in
unrenovated, unabated houses, by sample type. Table 7 contains estimates and standard errors of
the ratios of element concentrations in homes having undergone renovation compared element
concentrations in unrenovated, unabated homes, by sample type. Table 8 provides analogous
estimates of ratios for homes having undergone abatement. In Tables 7 and 8, a ratio of 1.0
implies no estimated difference. An estimate less than 1 .0 indicates that lower levels were
observed in renovated (abated) houses, while an estimate greater than 1.0 indicates that higher
concentrations were observed in renovated (abated) houses. Those ratios that were significantly
different from 1 .0 at the 5 percent significance level are underlined.
Table 6 shows that air ducts, window stools, and window channels typically had the
highest baseline levels (the geometric mean concentrations for unrenovated/unabated houses) of
lead, calcium and zinc. Soil samples generally had the lowest concentrations for these elements.
The window channel dust samples had especially high baseline concentrations of barium and
lead relative to the concentrations of the other elements. In this manner, window channel dust
samples seemed to differ from the other types of samples.
A
Although having only six houses makes it difficult to control for the effects of renovation, ignoring this
factor might bias estimates of differences between abated and unabated houses and increase the uncertainty in these
estimates. Recall also that this was a pilot study, performed to develop methodology for the subsequent full study
involving many more houses.
17
-------
Close attention should be given to the log standard errors of the estimates in Tables 7
and 8. Most of these are very large in comparison to the logarithm of the multiplicative
estimates.
Note that a total of 198 statistical tests was performed in the analysis supporting the
results in Tables 7 and 8. Each test was performed at the 5 percent level. Therefore, even if
there were no effects of abatement or renovation on any of these element concentrations, it would
still be expected that approximately 5 tests would be significant for each table. Thirteen tests
results were found to be significant for renovation effects. There were two cases (lead on
floors and lead in interior entryways) where lead was significantly higher in dust samples in
renovated houses as compared to unrenovated houses. There were a number of cases where
elements with typically higher concentrations in dust than in soil had significantly higher
concentrations in a renovated house. This was true for calcium on window stools, floors,
beds/rugs/upholstery, and interior entryways, for magnesium in interior entryways, and for nickel
in air ducts. Cadmium in window channels was the exception to this trend. Correspondingly,
there were cases of elements which generally had higher concentrations in soil than in dust
where the soil concentration was significantly lower in a renovated house. This was true
for aluminum and titanium in exterior entryways. Finally, potassium, which showed no
significant differences across sample types in Section 2.1, had significantly lower concentrations
in exterior entryways and foundation samples at renovated houses.
For abated houses, there were five cases of significance. As noted above, there could
occur strictly due to chance, even if there were no differences between abated and unabated
houses. There were two cases where lead was significantly higher in abated houses: lead in
interior entryways and lead in exterior entryways. Zinc was also significantly higher in
exterior entryways at abated houses. Calcium in window stools and chromium in floors were
significantly lower.
18
-------
Table 6. Model Estimates and Log Standard Errors of Geometric Mean Concentrations in Unrenovated
Unabated Houses
Sample
Medium
Dust
Soil
Sample
Type
WCH
WST
ARD
FLR
BRU
EWY-I
EWY-O
FDN
BDY
#
Houses
4
6
5
6
5
6
6
6
6
Lead
Geo Mean Log Std.
(MB'9) Err.
7238 0.64
226 1.17
875 0.41
102 0.33
117 0.45
96 0.19
63 0.43
102 0.89
53 0.81
Aluminum
Geo Mean Log
(ug/g) Std. Err.
13346 0.54
5808 0.39
5341 0.36
7687 0.30
11954 0.39
14146 0.34
22868 0.10
18568 0.33
11492 0.44
Barium
Geo Mean Log
(ug/g) Std. Err.
7058 1.95
478 1.11
216 0.68
313 0.31
163 0.49
255 0.47
261 0.15
252 0.37
128 0.30
Cadmium
Geo Mean Log
(M9/9) Std. Err.
29.7 0.07
21.4 0.89
36.0 1.84
19.1 0.59
25.4 0.34
13.0 0.57
3.9 0.81
3.6 0.48
2.1 0.57
Calcium
Geo Mean Log
(ug'O) Std. Err.
34866 0.37
57057 0.12
53114 0.68
20998 0.25
18230 0.18
25873 0.35
13126 0.36
13395 0.27
9977 0.25
Chromium
Geo Mean Log
(ug/g) Std. Err.
39 0.72
87 0.46
46 0.69
141 0.36
69 0.12
109 0.63
60 0.68
32 0.18
21 0.33
CO
Sample
Medium
Dust
Soil
Sample
Type
WCH
WST
ARD
FLR
BRU
EWY-I
EWY-O
FDN
BDY
#
Houses
4
6
5
6
5
6
6
6
6
Magnesium
Geo Mean Log
(ug/g) Std. Err.
4237 0.45
4501 035
2719 0.50
3337 0.25
3558 0.39
4400 0.23
535 0.11
1175 0.43
703 0.45
Nickel
Geo Mean Log
(ug/g) Std. Err.
17.9 0.50
31.3 0.41
35.2 0.27
41.2 0.69
17.6 0.64
24.6 0.43
13.3 0.81
14.9 0.20
8.47 0.21
Potassium
Geo Mean Log
(ug/g) Std. Err.
2563 0.72
2784 0.77
5553 0.41
4184 0.46
6723 0.92
5575 0.70
4955 0.12
4458 0.19
3500 0.33
Titanium
Geo Mean Log
(ug/g) Std. Err.
656 0.33
370 0.13
188 0.44
222 0.33
387 0.57
444 0.28
601 0.08
443 0.26
338 0.22
Zinc
Geo Mean Log
(ug/g) Std. Err.
13783 0.35
1229 0.37
16504 0.67
555.2 0.40
447.8 0.94
439.1 0.40
183.1 0.19
269.5 0.29
120.8 0.52
-------
Table 7. Ratio of Element Concentrations in Renovated Homes to Concentrations in Unrenovated
Homes, Estimates and Log Standard Errors
Sample
Medium
Dust
Soil
Sample
Type
WCH
WST
ARD
FLR
BRU
EWY-I
EWY-O
FDN
BDY
Lead
Log Std.
Effect Err.
0.45 0.62
1.34 1.57
0.51 1.32
4.67 0.12
17.08 1.64
4.87 0.04
2.12 0.21
2.29 0.91
1.87 0.67
Aluminum
Log
Effect Std. Err.
0.62 0.44
0.96 0.24
2.81 1.05
0.66 0.10
0.14 1.20
0.57 0.13
0.50 0.01
0.49 0.13
0.57 0.22
Barium
Log
Effect Std. Err.
0.31 5.70
0.27 1.40
7.37 3.66
0.35 0.11
5.01 1.92
0.25 0.25
0.72 0.03
0.78 0.16
0.93 0.10
Cadmium
Log
Effect Std. Err.
0.32 0.01
0.21 0.91
0.43 27.17
0.59 0.39
0.17 0.92
0.84 0.37
0.43 0.75
0.78 0.27
0.86 0.37
Calcium
Log
Effect Std. Err.
0.84 0.21
2.83 0.02
1.72 3.66
2.55 0.07
10.45 0.26
9.80 0.14
0.54 0.15
0.77 0.09
0.81 0.06
Chromium
Log
Effect Std. Err.
0.67 0.78
0.43 0.24
9.89 3.84
0.44 0.15
3.19 0.11
0.56 0.45
0.38 0.53
0.57 0.04
0.71 0.12
10
o
Sample
Medium
Oust
Soil
Sample Type
WCH
WST
ARD
FLR
BRU
EWY-I
EWY-O
FDN
BDY
Magnesium
Log
Effect Std. Err.
1.14 0.30
1.27 0.14
3.52 1.98
1.38 0.07
0.88 1.22
2.21 0.06
1.36 0.01
0.27 0.21
1.95 0.14
Nickel
Log
Effect Std. Err.
1.10 0.38
1.59 0.19
79.23 0.58
0.84 0.54
141.04 3.23
1.09 0.21
0.44 0.74
0.63 0.05
0.62 0.05
Potassium
Log
Effect Std. Err.
0.72 0.78
0.52 0.67
0.31 1.34
0.28 0.25
0.04 6.74
0.39 0.55
0.56 0.02
0.52 0.04
0.63 0.13
Titanium
Log
Effect Std. Err.
1.04 0.16
0.85 0.02
3.48 1.55
0.74 0.12
0.18 2.61
0.58 0.09
0.62 0.01
0.72 0.08
0.65 0.05
Zinc
Log
Effect Std. Err.
0.41 0.18
0.47 0.16
0.01 3.61
1.45 0.18
3.35 7.14
1.91 0.18
1.25 0.04
0.90 0.09
1.24 0.30
An underline indicates significant effect at the 0.05 level.
-------
Table 8. Ratio of Element Concentrations in Abated Homes to Concentrations in Unabated
Homes, Estimates and Log Standard Errors
Sample
Medium
Dust
Soil
Sample
Type
WCH
WST
ARD
FLR
BRU
EWY-I
EWY-O
FDN
BDY
Lead
Log Std.
Effect Err.
0.26 0.62
4.45 1.03
0.91 0.22
2.27 0.08
0.96 0.27
3.25 0.03
4.51 0.14
2.13 0.60
2.42 0.44
Aluminum
Log
Effect Std. Err.
1.13 0.44
1.09 0.16
1.36 0.17
0.87 0.07
0.47 0.20
0.82 0.09
0.77 0.01
0.90 0.08
1.14 0.14
Barium
Log
Effect Std. Err.
0.21 5.70
2.93 0.92
1.42 0.61
1.36 0.07
1.61 0.32
2.08 0.16
1.22 0.02
1.13 0.10
1.49 0.07
Cadmium
Log
Effect Std. Err.
0.82 0.01
2.11 0.60
0.68 4.53
0.41 0.26
0.27 0.15
0.67 0.24
2.41 0.49
1.30 0.18
1.58 0.25
Calcium
Log
Effect Std. Err.
1.02 0.21
0.61 0.01
0.58 0.61
0.92 0.05
1.11 0.04
0.60 0.09
0.81 0.10
0.69 0.06
0.92 0.04
Chromium
Log
Effect Std. Err.
1.19 0.78
0.68 0.16
1.60 0.64
0.27 0.10
0.56 0.02
0.33 0.30
0.81 0.34
1.07 0.02
1.23 0.08
Sample
Medium
Dust
Soil
Sample Type
WCH
WST
ARD
FLR
BRU
EWY-I
EWY-O
FDN
BDY
Magnesium
Log
Effect Std. Err.
1.37 0.30
1.01 0.09
1.46 0.33
0.84 0.05
0.81 0.20
0.75 0.04
0.99 0.01
1.38 0.14
0.88 0.09
Nickel
Log
Effect Std. Err.
1.44 0.38
1.12 0.12
0.62 0.10
0.59 0.36
2.10 0.54
0.75 0.14
1.46 0.49
0.80 0.03
1.37 0.03
Potassium
Log
Effect Std. Err.
1.17 0.78
1.30 0.44
0.78 0.22
1.69 0.16
0.72 1.12
0.88 0.36
0.92 0.01
0.88 0.03
1.17 0.08
Titanium
Log
Effect Std. Err.
0.68 0.16
1.09 0.01
1.41 0.26
0.95 0.08
0.41 0.44
0.86 0.06
0.86 0.00
1.05 0.05
1.23 0.03
Zinc
Log
Effect Std. Err.
0.19 0.18
2.63 0.10
0.25 0.60
1.42 0.12
1.55 1.19
1.65 0.12
1.89 0.03
1.68 0.06
1.55 0.20
An underline indicates significant effect at the 0.05 level.
-------
Figure 2a displays the estimates in Table 6 (unrenovated, unabated house geometric
means), portrayed on a log scale. Elements were sorted by geometric mean concentrations in
boundary soil. A distinction between sample types was observed in the shapes depicted in these
figures. Therefore, the sample types were presented in three groups. The first column displays
the results for window channels, window stools, and air ducts. The second column displays the
corresponding results for floors, bedcover/rug/upholstery, and interior entryways. The last
column contains the results for the three soil samples: boundary, entryway, and foundation.
The pattern in the concentrations observed across elements appears similar for the
three soil sample types: a monotonic increase from left to right. A noticeable, but less
consistent pattern, also appears to be present for interior entryway and beds/rugs/
upholstery, and even for floors and window stools. In this pattern, a cluster of four dots
appears in the lower left corner of the plot, a cluster of three dots appears in the middle, and there
is a cluster of four dots in the upper right part of the plot. Air ducts and window channels stand
on their own. Neither air ducts nor window channels appears to be similar to the other sample
types or to each other.
Figure 2b illustrates the renovation effects presented in Table 7. Log scale was used
because of two very high ratios of concentrations in renovated homes to concentrations in
unrenovated homes for nickel in air duct and bedcover/rug/upholstery samples. The log
transformed ratios appear to have a similar pattern for the three soil sample types, except the ratio
for magnesium in foundation soil samples, which looks out of pattern. Another pattern can be
observed in floor and entryway dust samples. For these sample types, magnesium, lead, zinc,
barium, and cadmium appear hi one cluster; chromium, titanium, potassium, nickel, and
aluminum form another cluster; and calcium stands alone. A third pattern shows some
similarities between window channels and window stools. Air duct and bed/rug/upholstery dust
samples are distinct.
22
-------
ro
w
^— •
0
E
0)
LLJ
Al
Ca
K
Mg
Ti
Ba
Zn
Pb
Cr
Ni
Cd
Al
Ca
K
Mg
Ti
Ba
Zn
Pb
Cr
Ni
Cd
Al
Ca
K
Mg
Ti
Ba
Zn
Pb
Cr
Ni
Cd
2 4 6 8 10
i I i i I I I i i I i i i i I
Dust:W. Channel
Dust:W. Stool
Dust:Air Duct
0
0
Dust:Floor
0
Dust:Bed/Rua/~
Dustilnt. Entwv
0
£
Soil:Entrvwav
_
__ _
^
^
Soil'.Foundation
Soil:Boundarv
2468 10 2468 10
Estimated Average Log-Concentration
Figure 2a. Estimated Average Log concentrations in Unrenovated, Unabated Units, for Each Element and
Each Sample Type. Elements Sorted by Geometric Average Concentration in Boundary Soil
-------
_OJ
111
Mg
Pb
Zn
Ba
Cd
Ca
Cr
Ti
K
Ni
Al
Mg
Pb
Zn
Ba
Cd
Ca
Cr
Ti
K
Ni
Al
Mg
Pb
Zn
Ba
Cd
Ca
Cr
Ti
K
Ni
Al
-4-2024
I 1 i 1 i 1 i 1 I I 1 1 1 1 1
Dust:W. Channel
Dust:W. Stool
DustrAir Duct
Dust: Floor
Dust:Bed/Rua/..
Dust:lnt. Entwv
Soil:Entrvwav
SoihFoundation
_
^
Soil:Boundarv
-4-2024 -4-2024
Log Ratio of Element Concentrations in Renovated Homes to Concentrations in Unrenovated Homes
Figure 2b. Log Ratio of Element Concentrations in Renovated Homes to Concentrations in Unrenovated,
Unabated Homes, Sorted by Ratios in Boundary Soil
-------
Figure 2c displays the abatement effects estimated in Table 8. There is no clear similarity
pattern among the sample types.
2.3 RELATIONSHIPS AMONG SAMPLE TYPES
This section explores the relationships among the sample types by analyzing the multi-
element data. There are three subsections. The first subsection discusses pairwise correlations
between lead and the other elements. The second subsection includes correlation scatterplots for
each sample type, with the correlations between elements being displayed for each sample type.
A visual inspection of the correlation scatterplots is used to identify similar sample types. The
last subsection covers a principal components analysis of the multi-element data which reduces
the dimensionality of the analysis, and suggests graphically which sample types are similar. The
analyses in Subsection 2.3.1 and 2.3.2 were done with the (unadjusted) concentration data
described in Section 2.1. The analyses in Subsection 2.3.3 were done with the adjusted
concentrations described in Section 2.2.
2.3.1 Correlations Between Lead and the Other Elements
Table 9 displays the estimated correlation between average logarithmic transformed
concentrations for each house for lead and each of the remaining ten elements by sample type.
Lead was most frequently positively correlated with zinc at a statistically significant level (0.05).
In particular, the correlation between these elements was significantly positive in boundary soil
samples, interior entryway dust samples, exterior entryway soil samples and window stool dust
samples. There was also a significantly positive correlation between lead and calcium in dust
samples taken from bedcover/rug/upholstery. Correlations between lead and calcium in
boundary and foundation soil samples and nickel in foundation soil samples were significantly
negative.
To investigate the overall association of lead with all of the other elements, one can
generally use a multiple correlation procedure. However, there must be at least as many houses
as there are elements of interest. Therefore, this procedure is not applicable for this pilot study
data.
25
-------
NJ
O>
•4—*
-------
Table 9. Estimated Correlation Between Lead and Remaining Elements, by Sample Type (Log-transformed concentrations)1
Sample Type
Air
Ducts
Boundary
Soil
Bedcover/Rug/
Upholstery
Entryway (Inside)
Entryway (Outside)
Foundation
Soil
Floor
Dust
Window
Channel
Window
Stool
#of
Houses
5
6
5
6
6
6
6
6
4
Aluminum
-0.30
-0.38
-0.32
-0.56
-0.72
-0.53
-0.47
-0.47
-0.04
Barium
0.04
0.60
0.07
-0.17
0.34
0.37
-0.64
0.78
0.92
Cadmium
-0.77
0.69
-0.09
-0.40
0.12
0.72
-0.53
-0.28
0.72
Calcium
0.55
-0.86
0.89
0.51
-0.65
-0.89
0.77
-0.15
0.57
Chromium
0.27
-0.30
0.72
-0.67
-0.49
-0.59
-0.68
0.18
0.47
Magnesium
0.14
-0.22
0.01
0.41
0.22
-0.20
0.16
-0.49
-0.85
Nickel
-0.28
-0.26
0.55
-0.29
-0.24
-0.90
-0.59
0.31
-0.20
Potassium
-0.31
-0.34
-0.77
-0.44
-0.53
-0.73
-0.54
-0.43
-0.26
Titanium
0.02
-0.32
0.18
-0.56
-0.60
-0,66
-0.18
0.47
0.83
Zinc
0.43
0.90
0.19
0.86
0.90
0.67
0.62
0.74
0.98
NJ
Underlined correlations were significant at the 0.05 level.
-------
2.3.2 Bivariate Relationships Among the Elements
Displays portraying the bivariate relationships among the eleven elements are provided in
Figures 3a through 3i. For each sample type, average log-concentrations for each house are
plotted for each pair of elements. Ellipses are drawn on each plot that represent 95% of the
estimated bivariate distribution. Those plots for which the ellipse is narrow represent pairs of
elements for which there was a strong observed correlation. Pairs of elements which are
negatively correlated have an ellipse with the major axis running from upper left to lower right.
The magnitude of the correlation can be inferred from the shape of the ellipse by comparing it to
the key at the bottom of each graph.
On the plots in Figures 3a through 3i, each house is identified with a different symbol.
This permits determining whether certain houses have similar characteristics with respect to the
various elements and/or sample types.
Although it is difficult to interpret the plots in Figures 3a through 3i, some patterns can
be seen in these correlation scatterplots. Window channel and window stool samples are
characterized by positive correlations among lead, barium, titanium, and zinc. Among the other
samples types, it is interesting to note that the foundation sample type nearly had this same
pattern in its scatterplot. Foundation and boundary samples were characterized by positive
correlations among aluminum, chromium, nickel, potassium, and titanium. Floor samples and
interior entryway samples displayed positive pairwise correlations between lead and zinc,
aluminum and titanium, barium and potassium, calcium and magnesium, cadmium and
chromium, and lead and calcium. These samples also had a negative pairwise correlation
between lead and chromium. Exterior entryway samples had some of the characteristics of both
the other soil samples and the floor and interior entryway samples. However, the exterior
entryway samples generally did not have positive correlations as strong as those of the other soil
samples among the five elements aluminum, chromium, nickel, potassium, and titanium. Also,
the exterior entryway samples had a negative correlation between calcium and magnesium and a
negative correlation between lead and calcium, in contrast to the positive correlations for the
floor and interior entryway samples.
28
-------
Pb
Al
Ba
Cd
Ca
Or
Mg
Ni
K
Ti
Zn
Corr.
Ellipse
90%
0
60%
o
30%
O
0%
O
House Legend: * = 17 a=19 +=33
X = 43 Z = 51 Y = 80
Figure 3a. Window Channel House Mean Correlation Scatterplot
29
-------
Pb
Al
Ba
Cd
Ca
Or
Mg
Ni
K
TI
Zn
Corr.
Ellipse
90%
0
60%
o
30%
O
0%
O
House Legend: * = 17 '
X = 43
D = 19
+ = 33
Figure 3b. Window Stool House Mean Correlation Scatterplot
30
-------
Pb
Al
Ba
Cd
Ca
Cr
Mg
V
Ni
K
TI
Zn
COIT.
Ellipse
90%
0
60%
(9
30%
o
0%
O
House Legend: * = 17 D = 19 + = 33
X = 43 Z = 51 Y = 80
Figure 3c. Foundation Soil House Mean Correlation Scatterplot
31
-------
Pb
Al
Ba
Cd
Ca
Or
Mg
Ni
K
TI
Zn
Corr.
Ellipse
90%
^
60%
o
30%
O
0%
O
House Legend:
* = 17
X = 43
D = 19
Z = 51
+ = 33
Y = 80
Figure 3d. Boundary Soil House Mean Correlation Scatterplot
32
-------
Pb
Al
tf
Vs* rj
Ba
Cd
Ca
Or
Mg
Ni
K
Tl
Zn
Corr.
Ellipse
90%
^
60%
O
30%
O
0%
O
House Legend: * = 17 D = 19 + = 33
X = 43 Y = 80
Figure 3e. Floor House Mean Correlation Scatterplot
33
-------
Pb
Al
Ba
Cd
Ca
Or
Mg
Ni
K
Ti
Zn
Corr.
Ellipse
90%
^
60%
o
30%
O
0%
O
House Legend: * = 17
X = 43
D = 19
Z = 51
+ = 33
Y = 80
Figure 3f. Entryway Dust House Mean Correlation Scatterplot
34
-------
Pb
Al
Ba
Cd
Ca
Or
Mg
Ni
K
Ti
Zn
COIT.
Ellipse
90%
0
60%
(9
30%
O
0%
O
House Legend:
* = 17 D = 19
X = 43 Z = 51
+ = 33
Y = 80
Figure 3g. Entryway Soil House Mean Correlation Scatterplot
35
-------
Pb
Al
TX+J
Y*
Ba
Cd
Ca
Or
Mg
Ni
K
Tl
Zn
Corr.
Ellipse
90%
^
60%
^?
30%
o
0%
O
House Legend: * = 17 n=19
X = 43 Y = 80
+ = 33
Figure 3h. Air Duct House Mean Correlation Scatterplot
36
-------
Pb
Al
Ba
+ y
?*
Cd
Ca
Or
Mg
Ni
K
Ti
Zn
Corn
Ellipse
90%
60%
(9
30%
o
0%
O
House Legend:
* = 17
X = 43
D = 19
Z = 51
+ = 33
Y = 80
Figure 3i. Bedcover/Rug/Upholstery House Mean Correlation Scatterplot
37
-------
Air duct samples and bedcover/rug/upholstery samples had patterns that were unlike each
other and the rest of the samples. Air duct samples were characterized by strong positive
pairwise correlations between titanium, aluminum, chromium, and magnesium, and by a strong
negative correlation between barium and potassium. Bedcover/rug/upholstery samples did not
have the strong correlations seen for air ducts, but did have a pattern of positive correlations
among lead, calcium, chromium, and nickel.
Because lead, barium, titanium, and zinc were used in paints in the past, the positive
correlations among these elements for the window channel and window stool samples might
be reflective of dust generated from paint.
2.3.3 Multivariate Relationships (Principal Components)
For the estimated model parameters displayed in Tables 6, 7, and 8 (average log-
concentrations in unrenovated unabated houses, increments in log-concentration associated with
renovation, and increments in log-concentration associated with abatement), a principal
components analysis was performed across the nine sample types. The purpose of this analysis
was to identify consistent patterns across sample types and to determine whether there were
patterns in the differences between homes that were abated or renovated.
Principal component analyses can be performed based on either correlations or
covariances. Analyses based on correlations standardize the range of each of the elements'
concentrations. This prevents the most widely fluctuating elements from dominating the analysis
and gives equal attention to all variables regardless of their range. Covariance-based analyses
leave all element concentrations in their original scale. Since the scales observed varied
substantially by element, and a priori there was no reason to weight more heavily the elements
with greater absolute variation, correlations were used.
The numerical results of the principal components analyses and plots of the first two
principal components are displayed in Table 10 and Figure 4. Table 10 displays estimates of the
coefficients for the first two principal components followed by the cumulative proportion of total
variation explained by these components. Figure 4 displays the relationship among the nine
different sample types relative to the first two principal components. Although there are eleven
38
-------
elements, the two principal component axes represent the two perpendicular directions (in the
eleven-dimensional space) in which the greatest variability was observed.
The first two principal components accounted for at least 68% of the total variability in
the model parameter estimates in each of the three analyses. This means that although eleven
elements were measured (lead, aluminum, barium, cadmium, calcium, chromium, magnesium,
nickel, potassium, titanium, and zinc), most of the variation among the nine sample types
occurred within a two-dimensional space (i.e., two linear combinations of the eleven element
concentrations). In each case, this was highly significant.
In three of the principal components, aluminum and titanium both appear with negative
coefficients. This is an interesting pattern because, as pointed out in Section 2.1, these two
elements generally had higher concentrations in soil than dust, whereas all the other elements
either were typically higher in dust than soil or had no significant differences across sample
types.
Figure 4 shows that for averages in unrenovated, unabated houses it can be argued that
the three soil sample types can be grouped into one cluster; floor, entryway, window stool,
bedcover/rug/upholstery, and air duct dust sample types form another cluster; and window
channels stand alone. This is similar to the groupings of Section 2.2 with one exception. In
Section 2.2, air ducts stood alone, whereas here air ducts are grouped with a number of other
sample types.
For the differences associated with renovated houses, all samples can be grouped into one
cluster except for air ducts and bedcover/rug/upholstery, which are distinct from the rest of the
samples and each other. This grouping has similarities to the grouping of Subsection 2.2. In that
section, the three soil samples were grouped together, entryway and floor samples made up a
second group, and it could be argued that the window channels and window stools should be
grouped together. It is worth noting that the two sample types that did not fit into any group in
either Figure 2b or Figure 4 (air ducts and bedcovers/rugs/upholstery) were not sampled in the
only fully renovated house.
39
-------
Table 10. Principal Components for Model Parameter Estimates (Adjusted House Averages,
Abatement History, and Renovation History)
Response
Unrenovated
Unabated
Unit Means
Abatement
History
Renovation
History
Principal
Component
1
2
1
2
1
2
Principal Component Coefficients1
Pb
0.20
0.48
0.34
0.35
0.02
0.47
Al
-0.37
0.20
-0.37
0.29
0.40
-0.22
Ba
0.17
0.48
0.30
0.31
0.43
0.13
Cd
0.43
0.04
0.11
0.31
-0.13
-0.15
Ca
0.41
-0.00
0.07
-0.46
0.03
0.40
Cr
0.15
-0.32
-0.43
0.07
0.46
-0.01
Mg
0.37
-0.00
-0.43
0.06
0.34
-0.15
Ni
0.36
-0.28
0.16
-0.34
0.30
0.33
K
-0.09
-0.27
0.09
0.10
-0.22
-0.37
Ti
-0.17
0.43
-0.23
0.44
0.40
-0.23
Zn
0.32
0.25
0.42
0.22
-0.10
0.45
Cumulative
Explained
Variability
0.40
0.71
0.36
0.68
0.43
0.83
Significance
Level2
<0.001
0.005
<0.001
1. Coefficients are applied to the estimated parameters for each sample type to obtain maximum spread among sample types in two dimensions.
2. Significance level of the proportion of variability explained by the first two principal components under the null hypothesis of uncorrelated element
concentrations. Under the null hypothesis, the distribution for the proportion of variability explained by the first two components was estimated based on a
simulation study. The significance level is the probability that the proportion of variability explained by the first two components is larger than or equal to the
observed proportion.
-------
Unrenovated, Unabated
House Means
x*
Renovation Effects
Abatement Effects
OA
Y
Component
+ + + Window Channels
z z z Entryway Dust
xxx Window Stools
« * « Entryway Soil
* * * Air Ducts
ooo Foundation Soil
D D D Floors
A A A Boundary Soil
Y Y Y Bed/Rug/Upholstery
Figure 4. First Two Principal Components for Each Building Component, Plotted
versus Each Other for Unrenovated, Unabated Unit Mean
Log-Concentrations, Renovation History, and Abatement History
41
-------
Figure 4 also displays differences associated with abated houses in the lower right hand
corner of the figure. The foundation and boundary soil samples are close to each other, and
nearby are floor, entryway dust, and entryway soil samples. Hence, for abatement, Figure 4
conveys far more clustering of sample types than does Figure 2c.
3.0 PEER REVIEW
This report was peer reviewed by four peer reviewers with expertise and background in
the subject area of the report. Comments from the reviewers which had an important effect on
the report and which are important for interpreting the report are described below.
One reviewer stated that the report was lacking in testable hypotheses. In response,
testable hypotheses were added to the report. However, the report was intended to be an
exploratory analysis in some respects and that exploratory aspect was retained. Another reviewer
suggested an alternative graphical approach, which was incorporated into Figure 2 in the final
report. The reviewer also suggested an ordering for graphs which produced monotonic plots that
could be readily used as reference points for comparison. Significant changes were made to the
description of the data in response to one of the reviewers. Reviewers also commented on the
assumption of the lognormal distribution and the reliability of elemental measurements as could
be measured by side-by-side samples. In response, sections on tests for lognormality and
measurement reliability were added to Appendix A. Finally, in response to reviewer comments,
the findings and conclusions of the report were reviewed and revised or replaced as necessary.
4.0 REFERENCES
Morrison, D., 1976, Multivariate Statistical Methods. Second Edition, McGraw-Hill.
Tisdale, S. L., Nelson, W. L., and Beaton, J. D., Soil Fertility and Fertilizers, 4th edition,
Macmillan Publishing Co., NY, 1985.
US EPA, 1995, "Comprehensive Abatement Performance Pilot Study, Volume I: Results of Lead
Data Analyses", EPA 747-R-93-007.
42
-------
US EPA, 1996, "Comprehensive Abatement Performance Study, Volume I: Summary Report,"
EPA230-R-94-013a.
US EPA, 1996, "Comprehensive Abatement Performance Study, Volume II: Detailed Statistical
Results," EPA 230-R-94-013b.
43
-------
This page intentionally left blank.
44
-------
APPENDIX A:
SUMMARY OF MULTI-ELEMENT DATA
A-1
-------
This page intentionally left blank.
A-2
-------
APPENDIX A:
SUMMARY OF MULTI-ELEMENT DATA
A-1.0 MULTI-ELEMENT DATA LISTING
Tables A-la through A-If contained the raw element concentration data for each of the
six houses. Each table displays concentrations for a given house for each of the eleven elements
by sample medium (dust or soil), sample type, location and sample ED. Each sample was
uniquely identified by its house number and sample ID.
A-3
-------
Table A-1a. CAP Pilot Study Multi-Element Data, House 17
Sample Identification
Medium
Dust
Soil
Tvoe
ARD
BRU
EWY-I
FLR
WCH
WST
BDY
EWY-0
FDN
Location
KIT
BD1
BD1
EWY
EWY
KIT
KIT
BD1
BD1
BD1
LVG
LVG
KIT
KIT
BD1
BD1
LVG
LVG
LVG
LFT
BAG
LFT
FRO
BAC
LFT
BAC
BAC
Sample ID
09
19
18
20
21
01
03
11
12
13
31
32
07
06
14
16
36
39
40
26
27
28
22
23
24
25
29
Concentrations
Pb
363
717
66.9
282
259
50.0
254
373
328
225
153
63.7
1140
221
727
338
506
270
337
52.2
70.5
56.4
70.4
364
70.2
69.4
65.7
Al
8970
8660
5140
10200
10200
1690
6950
7290
9280
6090
5170
6460
268
6600
16300
12500
4480
12500
9770
26700
20200
25100
20400
19600
20800
18000
18200
Ba
187
173
434
367
1100
742
1840
742
875
698
442
165
915
440
627
725
377
1820
2170
221
183
206
196
440
199
262
171
Ca
16400
16900
19000
12300
16500
14200
23100
15400
8770
33700
13900
7080
22700
48000
39100
41700
29700
21200
27200
13100
8260
13300
12800
14200
12200
11300
11700
Cd
65.7
615
9.97
19.6
11.5
3.10
13.4
26.1
14.6
8.74
10.6
3.71
b
114
198
191
39.5
307
146
2.68
2.33
2.61
2.75
241
2.81
2.62
2.51
Cr
40.3
64.9
43.9
36.5
34.7
16.2
29.4
43.7
42.7
29.3
26.0
24.6
45.0
23.6
35.8
38.6
42.9
43.3
50.8
44.6
38.5
43.8
37.7
269
40.9
39.2
38.0
//a/a)
K
7740
5790
8100
8120
8420
14400
17200
10000
11500
14900
9870
4600
481
31900
3820
4990
8800
4920
6290
6400
5940
5870
5360
4570
5410
4420
4460
Ma
3780
3730
3210
2290
3090
2720
3950
2790
2240
4180
2490
1600
4870
8460
8040
7360
10900
3980
6380
984
500
1030
540
614
668
2570
2960
Ni
22.7
19.6
76.8
27.3
27.9
13.0
16.3
120
45.5
33.9
222
18.6
20.5
159
23.1
22.2
188
16.8
27.1
17.1
15.5
16.4
15.1
238
15.7
13.9
14.3
Ti
245
296
84.9
285
332
55.1
104
207
188
243
159
209
957
323
552
368
243
627
505
692
454
643
486
582
422
391
385
Z
517'
39900
572
426°
620"
502
1340
516'
284"
1750
486"
229"
14900
1730
10000
4220
2520
1310
1910°
116
177
108
181
499'
279
345
299
Analysis result was greater than upper calibration
' Analysis result was greater than upper calibration
limit; reported value is the maximum detectable concentration.
limit for cadmium; sample excluded from data analysis.
-------
Table A-1b. CAP Pilot Study Multi-Element Data. House 19
Sample Identification
Medium
Dust
Soil
Tvoe
ARD
BRU
EWY-I
FLR
WCH
WST
BDY
EWY-0
FDN
Location
LVG
BD1
LVG
BD1
EWY
EWY
LVG
LVG
BD1
BD1
BD1
KIT
KIT
BD1
LVG
BD1
KIT
FRO
LFT
LFT
FRO
BAG
FRO
FRO
LFT
Sample
ID
09b
19
08
18
20
21
01
03b
11
12C
13
31
32
17"
04
16
36
26
27
29
22
23
28
24
25
Concentrations I
Pb
69.5
624
482
485
201
184
190
69.5
301
.
402
99.5
67.9
368
70.8
215
177
98.2
43.3
44.2
49.7
40.4
197
49.2
238
Al
8950
6810
2900
8660
6740
4560
5500
5690
4250
4330
,
4130
7760
4190
10900
8340
8030
12800
9280
31300
10200
10500
Ba
585
695
190
275
56.8
179
598
831
103
53.1
74.1
281
209
121
116
110
131
128
409
116
228
Ca
69600
93800
37000
1 40000
94800
1 77000
20000
58500
9280
8140
149000
74200
92700
8320
11200
11700
12200
13400
15100
12600
12500
Cd
23.7
12.7
8.51
6.16
10.1
6.15
19.5
.
13.6
5.71
3.24
4.14
37.4
17.0
2.30
2.30
1.63
2.27
2.04
3.23
2.02
4.85
Cr
146
187
81.4
40.8
40.1
36.1
114
157
44.9
41.9
50.3
77.8
30.3
24.6
16.0
15.3
23.7
17.9
34.3
19.7
27.8
i/a/a)
K
3100
1900
1020
5400
2050
2890
.
2470
2140
2290
2270
1200
2450
1690
3430
3490
2950
3430
2840
6980
3010
3190
Mq
5100
4600
2430
6890
5990
7940
3370
3990
2970
2900
12400
4050
2620
430
1510
1510
491
370
985
403
378
Ni
313
389
112
30.6
47.6
31.5
152
306
43.2
40.7
19.1
116
47.3
8.91
6.58
6.49
10.3
11.7
13.8
8.01
21.0
Ti
351
265
104
290
241
130
157
166
136
143
416
385
166
379"
257
223
383
285
753
295
374
Zn
1470
1970
341
551
583
706
683
1520
316"
267°
231
2050
944
161
107
130
161
278
281
143
461
Analysis result was greater than upper calibration limit; reported value is the maximum detectable concentration.
ICP analysis hampered by calcium interference; no multi-element data reported.
Sample dropped in laboratory; therefore, no data reported.
The titanium concentration was originally reported as 0.38 p*g/g. This concentration was flagged in the outlier analysis, investigated, and revised to 379
The outlier analysis is described in Appendix B.
-------
Table A-1c. CAP Pilot Study Multi-Element Data, House 33
Sample Identification
Medium
Dust
Soil
Type
ARD
BRU
EWY-I
FLR
WST
WCH
BDY
EWY-0
FDN
Location
BD2
LVG
LVG
EWY
EWY
BD2
BD2
LVG
LVG
LVG
KIT
KIT
BD2
LVG
LVG
LDY
LVG
LFT
FRO
FRO
BAG
FRO
LFT
FRO
LFT
Sample ID
09
19
18
20
21
01
03
11
12
13
31
32
04
14
16
36
17
26
27
22
23
28
24
25
29
Concentrations (fjglg)
Pb
477
1610
117
128
88.4
135
183
189
128
107
116
88.2
575
175
562
581
7240
44.1
168
63.2
136
57
167
108
176
Al
8030
3550
12000
21700
17900
4910
4880
13100
12400
13400
13600
13200
7040
9740
8050
3960
13300
10900
13200
22800
26200
21500
22000
22700
25500
Ba
206
225
163
226
298
357
139
300
453
167
288
301
488
594
1830
510
7060
121
161
252
401
280
356
309
369
Ca
76700
36800
18200
21000
15900
42300
41800
20800
21500
23900
19000
20200
37300
26900
55800
1 55000
34900
12000
5270
8130
12500
8090
12400
12900
12300
Cd
19.8
65.6
25.4
12.9
30.0
13.1
40.9
88.9
66.1
20.8
35.7
33.0
19.6
24.7
11.0
10.1
29.7
2.18
2.01
2.52
14.4
1.75
3.51
3.27
4.17
Cr
53.0
40.7
69.1
94.2
523
96.7
85.2
180
190
146
516
676
135
101
87.0
85.8
39.1
27.0
19.9
29.4
952
26.9
31.3
28.4
36.8
K
3670
8410
6720
5800
5830
1830
1210
5100
5710
5850
5990
5600
5960
3730
3350
1510
2560
2980
3060
4190
6240
3530
4960
3620
5540
Mg
3380
2190
3560
5180
3870
3250
2940
3170
2950
4060
3490
3670
4150
3220
4440
6780
4240
474
497
495
849
494
3060
616
3350
Ni
27.7
44.6
17.6
21.5
12.7
33.5
15.1
18.6
19.6
22.9
20.9
16.8
52.2
21.5
24.6
17.1
17.9
9.59
7.58
10.8
13.1
6.78
15.9
11.9
13.3
Ti
297
120
387
572
558
165
195
389
314
325
386
355
625
373
480
283
656
321
443
730
667
575
423
601
498
Zn
2620
1 04000
448
458
482
426
646
939
866
608
609
577
1180
1500
1610
1180
13800
165
112
140
243
122
258
263
285
>
CD
-------
Table A-1d. CAP Pilot Study Multi-Element Data, House 43
Sample Identification
Medium
Dust
Soil
Tvoe
ARD
BRU
EWY-I
FLR
WST
WCH
BDY
EWY-0
FDN
Location
LVG
DIN
LVG
DIN
EWY
EWY
LVG
LVG
DIN
DIN
DIN
KIT
KIT
LVG
DIN
KIT
LVG
KIT
FRO
BAG
FRO
BAG
BAG
FRO
BAG
FRO
Sample
ID
09
19"
08
18
20
21
01
03
11
12
13
31
32
04
16
36
05
38
26
27
22
23
28
24
25
29
Concentrations I
Pb
1140
611
102
195
263
589
147
205
234
256
149
308
309
964
378
397
963
1430
290
60.8
623
205
304
337
181
245
Al
9150
6500
11500
13400
14300
6600
7830
6920
8630
7490
10400
13400
5170
10500
9170
13700
35400
12600
5340
13800
19400
15700
18500
21600
19400
Ba
243
209
304
331
2110
220
288
420
393
210
873
593
521
512
443
384
367
203
83.2
374
374
284
460
339
337
Ca
63500
28000
22100
18200
23300
15100
43300
30100
21900
15000
17800
25000
47400
20200
33800
56400
13100
12500
5790
12100
13100
13000
10000
15800
8240
Cd
11.0
6.00
6.15
5.26
6.91
4.71
7.26
7.73
8.12
4.59
8.23
8.79
18.2
20.6
221
8.93
72.3
4.53
0.94
6.58
2.83
2.45
5.39
3.80
4.29
Cr
165
40.0
37.6
35.3
26.2
33.8
30.2
51.5
42.0
44.1
47.0
45.6
82.7
28.6
44.3
23.8
32.5
28.3
13.8
28.7
32.1
25.9
41.8
36.4
34.6
[/a/a)
K
4100
9200
7770
9790
6060
7020
31700
8610
6270
6800
7390
6910
4590
6630
3550
5340
4640
4780
1740
3810
4550
3880
3800
4740
3800
Ma
6720
3860
4100
4530
4460
2940
8090
3720
3450
2920
3150
4430
4450
4020
4210
14000
4540
491
301
494
506
493
3070
610
2410
Ni
28.7
25.6
25.8
18.9
16.5
23.7
26.1
44.9
21.5
15.0
20.4
61.5
25.3
17.5
22.6
17.3
18.8
12.1
9.57
11.8
11.7
10.8
12.6
14.1
12.0
Ti
408
198
344
486
467
198
257
231
237
262
291
422
440
312
353
509
244
473
314
326
741
497
601
723
577
Zn
7810
2990
1250
763
2070
1640
989
2870'
1160
1320
949
981
1340
6950
1160
2540
1720
221
88.7
492"
300
272
812
561
488°
Analysis result was greater than upper calibration limit; reported value is the maximum detectable concentration.
ICP analysis hampered by calcium interference; no multi-element data reported.
-------
>
do
Table A-1e. CAP Pilot Study Multi-Element Data, House 51
Sample Identification
Medium
Dust
Soil
Tvoe
EWY-I
FLR
WST
WCH
BDY
EWY-0
FDN
Location
EWY
EWY
BAT
BD3
BD3
BD3
BD1
BD1
BD3
BAT
BD3
BD3
BD1
BAT
BD3
BD3
FRO
BAG
BAG
FRO
BAG
FRO
BAG
BAG
Sample
ID
20
21
01
11
12"
13
31
32
44
06
14
16
40
07
15
17
26
27
29
22
23
24
25
28
Concentrations
Pb
640
4030
2450
966
467
712
1780
1760
646
6370
774
670
3580
2730
421
493
346
329
300
899
505
938
539
426
Al
8490
7110
4410
6340
116
5060
5690
6090
3290
4020
7950
9160
6950
4830
13300
12500
7760
8190
7390
8710
9130
9170
9210
9320
Ba
234
75.5
93.0
43.2
86.2
135
1430
325
27.0
679
278
314
746
1190
288
300
207
177
178
232
269
258
262
257
Ca
1 30000
127000
1 34000
26400
14800
1 1 3000
91300
39300
17700
1 54000
92300
77300
77500
1 23000
13500
15600
5930
6560
7070
4100
5800
5450
7960
7520
Cd
6.98
11.6
8.78
7.50
1.72
5.30
7.19
6.44
4.37
19.9
4.73
6.08
7.00
13.2
6.71
5.21
3.86
2.55
2.40
4.51
3.74
4.13
3.81
3.16
Cr
22.5
37.8
20.7
25.8
5.59
16.8
26.7
22.9
14.1
31.1
22.9
30.3
26.1
26.1
33.6
33.8
24.6
19.2
16.9
22.4
23.1
15.9
22.5
20.2
\ualal
K
2320
1630
1860
2080
815
1920
1690
2050
1760
905
2170
3110
2780
901
3280
3410
2220
2600
2430
2290
2650
1610
2430
2310
Ma
7220
7430
8590
3010
1020
5590
3690
3140
2010
9290
4730
4820
5120
14500
4390
4560
304
1490
1690
1900
302
384
1520
295
Ni
13.7
19.8
36.0
15.3
3.40
13.5
12.8
11.9
8.02
165
90.4
18.9
24.4
52.5
22.0
19.5
11.2
6.14
5.83
6.90
7.70
7.51
7.10
6.90
Ti
294
211
149
188
44.1
175
226
260
117
259
345
407
486
362
485
570
306
305
271
342
324
378
343
346
Zn
743
2760
3390
966'
304
782
1440
1470'
657'
4110
835
866
2170
3200
753
549
314
235
217
433
376
533
377
340
Analysis result was greater than upper calibration limit; reported value is the maximum detectable concentration.
During initial sampling attempt, cartridge filled with sawdust prior to completion of sample collection. Sample was excluded from lead analysis
and multi-element analysis.
-------
Table A-1f. CAP Pilot Study Multi-Element Data. House 80
Samole Identification
Medium
Dust
Soil
Tvoe
ARD
BRU
EWY-I
FLR
WST
WCH
BDY
EWY-0
FDN
Location
BAT
BD3
KIT
BAT
BD3
EWY
EWY
BAT
BAT
BD3
BD3
BD3
KIT
KIT
BAT
BD3
PAN
KIT
KIT
BD3
KIT
KIT
FRO
BAG
FRO
BAG
BAG
LFT
BAG
BAG
Sample
ID
09
19
45
08
18
20
21
01
03
11
12
13
31
32
06
14
36
39
40
15
41
42
26
27
22
23
28
24
25
29
Concentrations I
Pb
1700
965
389
344
66.3
342
222
1210
649
180
175
243
182
223
61600
680
535
7880
4660
938
4550
5790
308
343
380
350
412
942
459
317
Al
5810
5270
3610
7780
2100
11800
7440
6870
8730
3720
4810
6430
4950
5510
610
6120
5200
3830
6260
11600
8140
11400
13000
13400
16400
15200
17600
17300
8810
8890
Ba
1640
366
470
263
101
303
257
1010
572
186
176
240
323
350
30300
1380
658
29400
6560
846
22500
10900
246
279
282
288
340
414
202
198
Ca
49700
32200
13400
41100
7620
25000
9620
51000
32800
13900
18000
9710
18200
15100
21300
38200
105000
29300
45900
51000
65400
29500
8320
7260
6960
10500
8230
6940
5160
7430
Cd
6.65
7.79
5.52
5.69
4.79
8.61
4.00
5.37
4.37
9.25
5.09
5.33
4.23
7.98
30.8
17.2
7.85
23.3
20.1
17.6
23.1
30.4
9.30
6.19
9.88
7.69
8.29
14.0
6.06
7.56
Cr
84.0
78.8
16.9
36.1
33.3
33.3
28.9
31.2
32.4
46.1
59.4
44.3
25.4
56.7
151
66.0
60.1
104
206
49.7
94.8
97.7
24.0
24.8
31.3
31.9
31.8
32.9
23.0
23.8
c/a/a)
K
2210
3480
3420
2510
1140
4990
670
3850
4380
3520
5050
3840
2540
3840
1540
348
2470
745
3150
3340
959
1810
4220
4660
4970
4710
5220
4440
2470
2570
Ma
3760
2280
1820
2950
1180
3710
2350
2990
2860
1510
1940
1720
1890
2290
5080
3160
2740
2430
2710
5030
4060
3750
489
493
489
502
487
772
1510
1500
Ni
37.6
12.4
10.1
15.2
42.5
27.8
8.82
14.6
18.5
51.2
19.5
14.8
10.6
21.5
42.4
99.3
15.9
35.4
140
15.6
21.5
147
9.51
11.0
12.3
11.5
13.8
13.8
7.43
8.05
Ti
225
209
103
272
117
389
301
226
198
155
177
224
239
243
181
426
630
494
461
439
715
568
437
326
486
501
528
564
322
288
Zn
5960
1170'
1240
664
136
703
468
1640
1180
436
508
326"
436
514
35100
1630
2590
7560
3470
1850
4830°
4510°
394
396
385
417
492°
973
345
377
Analysis result was greater than upper calibration limit; reported value is the maximum detectable concentration.
-------
A-2.0 GEOMETRIC MEAN CONCENTRATIONS BY SAMPLE TYPE AND UNIT
House geometric mean concentrations of the eleven elements were the basic quantities
used in the statistical analyses. They are listed in Table A-2. Also included in Table A-2 are
indicators of interior and exterior abatement for each house. A "U" indicates that no abatement
was performed in the house, an "R" indicates that the house was abated primarily by removal
methods, and an "E" indicates that the house was abated primarily by encapsulation/enclosure
methods. Table A-2 also contains the number of samples for which concentrations were
determined for all eleven elements. Any sample in Table A-la through A-If for which at least
one element had a missing value was not included in the calculations for the Table A-2 summary.
A-10
-------
Table A-2. Geometric Mean Concentration by Sample Type and Unit
Sample
Type
WCH
WST
ARD
FLR
BRU
EWY-I
House
33
43
51
80
17
19
33
43
51
80
17
19
33
43
80
17
19
33
43
51
80
17
19
33
43
80
17
19
33
43
51
80
Interior
Abatement
History
U
R
E
E
R
U
U
R
E
E
R
U
U
R
E
R
U
U
R
E
E
R
U
U
R
E
R
U
U
R
E
E
Exterior
Abatement
History
U
R
R
E
E
U
U
R
R
E
E
U
U
R
E
E
U
U
R
R
E
E
U
U
R
E
E
U
U
R
R
E
Renovation
None
None
Full
None
None
Partial
None
None
Full
None
None
Partial
None
None
None
None
Partial
None
None
Full
None
None
Partial
None
None
None
None
Partial
None
None
Full
None
Samples
Taken in
Unit
1
2
3
3
6
3
4
3
4
5
2
1
2
1
3
7
5
7
7
6
7
1
2
1
2
2
2
2
2
2
2
2
Geometric Mean Concentrations (//
Pb
7238.3
1174.9
827.6
2913.6
368.3
139.2
425.4
525.1
1854.4
3828.3
510.6
624.4
874.6
1137.7
861.2
165.5
173.1
130.7
220.9
1227.1
304.8
66.9
483.3
116.9
141.3
151.1
269.9
192.6
106.4
394.0
1605.4
275.4
Al
13345.9
22025.3
9305.9
10248.9
9505.7
5120.0
6836.3
7928.2
6718.3
3416.5
8813.8
8948.0
5340.9
9152.4
4800.0
5548.9
4830.6
9921.4
8504.5
5024.0
5668.4
5139.3
4444.6
11954.3
8630.5
4040.3
10232.5
7640.8
19721.0
13844.0
7773.1
9357.4
Ba
7057.8
375.5
469.3
5915.2
817.6
163.3
721.8
490.7
458.6
5556.1
179.9
585.1
215.5
243.1
655.5
642.6
217.4
267.2
380.0
137.8
338.8
433.8
363.7
162.5
252.1
163.2
636.0
125.0
259.4
835.8
133.0
279.1
Ca
34866.1
27181.3
29601.8
46139.1
33201.3
100782.1
54305.9
31861.3
96019.2
40917.9
16615.1
69610.2
53114.2
63535.5
27795.2
14686.3
27470.5
25622.3
22451.7
54204.6
19605.1
19032.9
58943.0
18229.9
24888.9
17692.3
14240.0
114992.3
18226.2
20562.2
128563.4
15521.7
Cd
29.66
25.40
7.73
23.09
140.20
13.80
15.22
43.54
7.96
18.10
201.09
23.72
36.02
11.03
6.59
9.17
7.87
35.89
6.86
6.43
5.71
9.97
10.41
25.44
6.07
5.22
14.99
7.88
19.71
6.03
9.00
5.87
Cr
39.05
27.82
30.92
77.22
38.12
49.10
100.43
47.14
27.40
105.18
51.14
145.95
46.41
164.75
48.20
28.83
65.56
203.43
41.41
20.63
40.40
43.85
123.27
69.06
38.78
34.67
35.56
40.40
221.98
30.42
29.12
30.99
K
2563.4
4977.4
2158.1
1794.5
7410.0
1708.4
3260.0
4764.2
2030.3
1254.4
6695.5
3097.1
5553.4
4100.3
2971.9
10974.0
2398.4
3854.7
8828.0
1890.0
3789.6
8097.0
1395.4
6723.0
8456.8
1688.3
8268.5
3326.8
5812.2
7701.6
1944.0
1828.9
/g)
Mg
4237.1
7979.5
6625.4
4244.4
7191.9
5090.4
4478.2
4222.3
5742.0
3105.9
3759.1
5103.5
2719.4
6724.6
2497.4
2726.5
3911.9
3341.5
3844.8
3870.5
2109.9
3210.0
3342.6
3558.3
3977.0
1867.1
2659.3
6423.5
4477.7
4497.6
7325.5
2953.4
Ni
17.86
18.02
28.21
36.69
43.74
47.15
26.21
21.50
51.21
50.58
21.10
312.90
35.15
28.70
16.77
39.95
76.26
20.42
27.29
14.44
19.03
76.84
208.42
17.55
25.69
25.43
27.59
38.19
16.55
17.63
16.46
15.66
Ti
655.7
352.5
464.2
563.0
414.4
298.3
422.0
364.5
364.8
406.3
269.7
351.4
188.4
408.4
169.6
151.1
145.9
290.5
264.2
179.6
206.6
84.9
166.3
387.4
260.9
178.7
307.6
264.3
565.2
476.5
249.1
341.8
Zn
13782.9
2089.1
1097.5
3426.6
2781.5
765.2
1354.3
2212.6
1594.7
5223.5
4537.5
1465.9
16503.7
7806.0
2053.8
568.7
573.1
647.6
1310.1
1235.2
609.7
572.3
819.4
447.8
1931.2
300.7
513.8
566.6
469.9
1255.7
1432.1
573.6
U = unabated, R = removal, and E = encapsulation/enclosure.
-------
Table A-2. (Continued)
Sample
Type
EWY-O
FDN
BDY
House
17
19
33
43
51
80
17
19
33
43
51
80
17
19
33
43
51
80
Interior
Abatement
History
R
U
U
R
E
E
R
U
U
R
E
E
R
U
U
R
E
E
Exterior
Abatement
History
E
U
U
R
R
E
E
U
U
R
R
E
E
U
U
R
R
E
Renovation
None
Partial
None
None
Full
None
None
Partial
None
None
Full
None
None
Partial
None
None
Full
None
Samples
Taken in
Unit
2
3
3
3
2
3
3
2
3
3
3
3
3
3
2
2
3
2
Geometric Mean Concentrations (//g/g)
Pb
160.1
73.3
78.8
338.4
673.7
379.6
68.4
108.3
146.9
246.0
599.4
515.4
59.2
57.3
86.0
132.6
324.7
324.8
Al
19994.7
15510.8
23437.0
16152.0
8916.5
16376.5
18939.6
10368.0
23354.5
19783.7
9231.2
11057.3
23827.4
9015.6
11982.1
8192.7
7773.8
13198.4
Ba
293.5
190.1
304.8
341.1
249.5
302.3
207.4
162.2
343.9
374.5
259.0
254.5
202.5
115.6
139.7
130.0
186.7
261.9
Ca
13488.9
13527.8
9360.6
12687.1
4876.6
8453.0
11734.0
12527.0
12542.9
10929.7
6884.4
6432.2
11296.0
10270.8
7961.7
8519.8
6504.0
7770.7
Cd
25.75
2.46
3.99
3.57
4.11
8.57
2.64
3.13
3.63
4.45
3.68
8.62
2.54
2.05
2.09
2.06
2.87
7.59
Cr
100.69
24.38
90.97
28.76
22.75
31.68
39.32
23.43
31.99
37.50
19.34
26.19
42.19
18.17
23.19
19.74
19.98
24.40
K
4950.2
4083.4
4517.1
4069.0
2461.6
4962.2
4740.2
3096.0
4632.8
4092.1
2081.4
3045.3
6063.7
3280.7
3019.1
2880.7
2413.1
4435.5
Mg
575.9
563.6
591.9
497.6
757.7
492.6
1718.7
390.2
1848.8
1652.2
556.2
1204.9
797.2
993.6
485.7
384.3
914.8
491.3
Ni
59.90
11.83
9.85
11.39
7.29
12.51
14.63
12.96
13.59
12.88
7.17
9.38
16.32
7.25
8.53
10.77
7.37
10.25
Ti
532.3
434.5
654.3
493.4
333.0
504.5
398.8
331.9
502.1
630.4
355.5
374.2
587.0
278.8
377.2
385.3
293.8
377.5
Zn
300.2
232.7
160.7
342.3
403.5
429.4
306.6
257.1
268.6
605.7
408.9
502.1
130.4
131.0
135.6
140.1
252.1
395.2
U = unabated, R = removal, and E = encapsulation/enclosure.
-------
APPENDIX B:
DISTRIBUTION AND
OUTLIER ANALYSIS FOR THE CAPS
PILOT MULTI-ELEMENT DATA
B-1
-------
This page intentionally left blank.
B-2
-------
APPENDIX B:
DISTRIBUTION AND OUTLIER ANALYSIS
B-1.0 INTRODUCTION
This appendix documents an analysis leading to the selection of the lognormal
distribution for characterizing element concentrations, provides a quantification of the reliability
of the element concentrations measured for this report, and presents the statistical outlier analysis
performed on the CAPS Pilot multi-element data. The statistical approach employed, the outliers
identified, and the results of the laboratory review of the outlier data are discussed.
B-2.0 LOGNORMAL ASSUMPTION
To investigate the appropriateness of the lognormal assumption, a goodness-of-fit test
was applied. The results of this test are presented in Table B-l. For each element and each
component sampled, a Shapiro-Wilk goodness of fit test was applied to both the untransformed
and the log-transformed concentrations. A 'Yes' appears in Table B-l for each case where the
hypothesis of normality is rejected at the 0.05 level, and a 'No' is for the case where hypothesis
of normality is not rejected.
Of the 99 element-component combinations examined, there were 42 cases where neither
untransformed data nor log-transformed data were rejected as non-normal. There were 32 cases
where the untransformed data were rejected as non-normal while the log-transformed data were
not rejected as non-normal. On the other hand, there were only 3 cases in which the log-
transformed data were rejected as non-normal and the untransformed data were not rejected.
There were 22 cases rejected for non-normality of both the log-transformed and untransformed
data. Overall, for approximately 75% of the tests, the lognormal distribution was not rejected,
whereas for approximately 45% of the tests, the normal distribution was not rejected. Hence the
lognormal distribution was chosen over the normal distribution for the analysis of the data.
When interpreting these results, however, it is important to remember that these tests do
not control for systematic differences between observations. For example, differences between
abated and unabated homes and substrate effects are not adjusted for. The tests are also based on
very little data (sample size for each element-component combination is no more than 6). Thus,
this should only be regarded as a cursory analysis, leading to a decision regarding whether or not
B-3
-------
to transform the data before modeling. It is not a full-fledged declaration that the variability in
these elements is well characterized by the lognormal distribution.
Table B-1. Test of Normality: Log-transformed and Untransformed Data"
Elements
Aluminum
Barium
Calcium
Magnesium
Nickel
Potassium
Cadmium
Chromium
Lead
Titanium
Zinc
Dust
ARD
Yes/No
No/Yes
No/No
No/No
No/Yes
No/No
No/Yes
No/No
No/No
No/No
No/Yes
BRU
No/No
No/No
No/Yes
No/No
No/Yes
No/No
No/Yes
No/Yes
No/No
No/No
No/No
EWI
No/No
No/Yes
Yes/Yes
No/No
No/No
No/No
No/Yes
Yes/Yes
Yes/Yes
No/No
Yes/Yes
FLR
No/Yes
No/Yes
Yes/Yes
Yes/Yes
Yes/Yes
No/Yes
Yes/Yes
Yes/Yes
No/Yes
No/No
No/Yes
WSL
Yes/No
Yes/Yes
No/Yes
No/Yes
Yes/Yes
No/Yes
Yes/Yes
No/Yes
Yes/Yes
No/No
No/Yes
WST
No/Yes
No/Yes
No/No
Yes/Yes
Yes/Yes
No/No
No/Yes
No/Yes
No/No
No/No
No/Yes
Soil
BDY
No/Yes
No/No
No/No
Yes/Yes
No/No
No/No
No/Yes
No/Yes
Yes/Yes
No/No
No/Yes
EWY
No/No
No/No
Yes/No
Yes/Yes
Yes/Yes
No/No
Yes/Yes
Yes/Yes
No/No
No/No
No/No
FDN
Yes/Yes
No/No
No/No
No/Yes
No/No
No/No
No/Yes
No/No
No/Yes
No/No
No/Yes
a Test results are presented first for log-transformed data and then for untransformed data. 'Yes' indicates
the hypothesis of normality was rejected. 'No' indicates the hypothesis of normality was not rejected.
B-3.0 CHARACTERIZATION OF MEASUREMENT RELIABILITY
Side-by-side dust samples were collected on floors and window stools. Eleven pairs were
collected on floors, and two pairs were collected on window stools. Side-by-side soil samples
were also collected near entryways, foundations, and property boundaries. These allow
characterization of the degree of variability introduced by local spatial variability combined with
the variability introduced by the chemical analysis process. Table B-2 provides estimates of the
log standard deviation of measured concentration in side-by-side samples for each of the eleven
elements considered. Also provided is an estimate of the proportion of total variability that is
attributed to real variation in element concentrations. This is measured as the total variance
B-4
-------
minus the variance of side-by-side measures, divided by the total variance. This quantity is
labeled as the measurement reliability. The closer this quantity is to 1, the more reliable the
measurement.
Table B-2. Log Standard Deviation and Measurement Reliability of
Measured Concentrations in Side-By-Side Dust Samples
Collected from Floors and Window Stools
Elements
Aluminum
Barium
Cadmium
Calcium
Chromium
Magnesium
Nickel
Lead
Potassium
Titanium
Zinc
Floor
Log
Standard
Deviation"
0.18
0.44
0.37
0.30
0.24
0.18
0.70
0.25
0.22
0.16
0.30
Measurement
Reliability"
0.86
0.84
0.86
0.77
0.94
0.73
0.57
0.93
0.88
0.84
0.81
Window Stool
Log
Standard
Deviation
0.28
0.76
0.38
0.26
0.35
0.24
0.73
0.29
0.73
0.11
0.43
Measurement
Reliability
0.78
0.74
0.93
0.61
0.81
0.76
0.61
0.97
0.63
0.58
0.76
Soil
Log
Standard
Deviation"
0.27
0.34
0.16
0.10
0.13
0.51
0.17
0.42
0.22
0.24
0.18
Measurement
Reliability"
0.68
0.43
0.89
0.88
0.78
0.45
0.78
0.84
0.72
0.60
0.90
a Log standard deviation is the square root of the variance of the log-transformed concentrations due to the
side-by-side measurement error.
b The measurement reliability is the proportion of the total variance not attributed to side-by-side
measurement error. It characterizes the precision of the measurements.
Measurement reliability was above 70% for 17 of the 22 element measures for dust. It
was above 70% for 7 of the 11 elements in soil. The lowest reliabilities, 0.43 and 0.45, were
observed for soil measurements of barium and magnesium, respectively.
B-5
-------
B-4.0 OUTLIER ANALYSIS
Two outlier tests were applied to the multi-element data. The first was a univariate
outlier test, which evaluates one element at a time. This is the same test that was previously
applied to the lead data. The test was applied to the natural logarithms of the concentrations for
lead, aluminum, barium, cadmium, calcium, chromium, magnesium, nickel, potassium, titanium,
and zinc. The second test was a multivariate outlier test, which evaluates measurements for all
eleven elements simultaneously. The multivariate test detects measurements which for a single
element may not be an outlier, but when viewed in combination with the other elements is
inconsistent with the majority of the data. Groupings of the data were defined before performing
the outlier tests.
B-4.1 DATA GROUPING
The following homogeneous groups of data were identified for each indicated sample
type:
• Vacuum Cassette Samples (7 groups): air duct, upholstery (including bed
coverings and throw rugs), interior entryway, floor (excluding entryway), window
stool, window channel, and floor (including entryway);
• Soil Samples (4 groups): boundary, foundation, exterior entryway, and all exterior
samples combined.
Initially, data for all six units in the Pilot Study were combined before performing the univariate
and multivariate outlier tests on these groups. When there were sufficient data, subsequent
univariate outlier tests were also performed by segregating the data in each group by abatement
method and by housing unit. Segregating by abatement method and unit was not done for the
multivariate test due to the need for larger sample sizes with the increase in dimensionality.
B-4.2 UNIVARIATE OUTLIER TEST
Formal statistical outlier tests were performed on the natural logarithms of the
concentrations for lead, aluminum, barium, cadmium, calcium, chromium, magnesium, nickel,
potassium, titanium and zinc. Data were placed into groups of comparable values, and a
maximum absolute studentized residual procedure was used to identify potential outliers. The
B-6
-------
SAS procedure GLM (SAS PC, ver. 6.08) was used to compute the studentized residual for each
data value in a group by fitting a "constant" model (i.e., mean value plus error term) to the log-
transformed data in each group. The absolute values of the studentized residuals were then
compared to the upper .05/n quantile of a student-t distribution with n-2 degrees of freedom,
where n is the number of data values in the group. If the maximum absolute studentized residual
was greater than or equal to the .05/n quantile, the corresponding data value was flagged as a
potential outlier. When a potential outlier was identified, that value was excluded from the
group, and the outlier test was performed again. This procedure was repeated until no more
outliers were detected.
B-4.3 MULTIVARIATE OUTLIER TEST
The multivariate outlier test is based on the Hotelling T-squared statistic, with one major
difference. The Hotelling T-squared statistic is discussed in most multivariate statistics texts.
The difference in the statistic used here is that, in computing the statistic for the i* observation,
that observation is excluded from the computation of the mean vector and the variance-
covariance matrix. This yields estimates of location and covariance that are unaffected by the
observation in question and lead to a more robust outlier test. This is a multivariate extension of
the univariate studentized residual used for the univariate outlier test. Under assumptions of
normality, the resulting statistic has an F distribution, with numerator degrees of freedom equal
to p (the number of elements) and denominator degrees of freedom equal to a function of p and
the sample size, N. In this case, p was equal to eleven.
The observation corresponding to the maximum value of the statistic in a data group was
declared a potential outlier if the statistic exceeded the (1-.10/N) quantile of the F distribution
with appropriate degrees of freedom. When a potential outlier was identified, that sample was
excluded from the group, and the outlier test was performed again. This procedure was repeated
until no more outliers were detected.
B-7
-------
B-4.4 RESULTS OF OUTLIER ANALYSIS
The potential outliers identified by these two tests were screened by a statistician to
eliminate those that were merely numerical anomalies due to very small sample sizes. The
remaining outliers identified by the univariate test are listed in Table B-3, and those identified by
the multivariate test are listed in Table B-4. These lists of the remaining outliers were sent back
to the laboratory for verification. One outlier was confirmed by the laboratory as an error and is
documented in the footnote to Table A-lb. All remaining outliers were verified and declared by
the laboratory to be correct as reported.
B-8
-------
Table B-3. Univariate Outliers Detected by Univariate Methods
Sample
Processing
Batch
CIS
CRS
SSS
CSS
CSS
SSS
SSS
SSS
CSS
CSS
CKC
CKC
CIS
SSS
CIS
CIS
CLS
SSS
SSS
CLS
SKI
SSS
CLS
CRS
SSS
SSS
SSS
CLS
CLS
CSS
CSS
CRS
CLS
SSS
SSS
CRS
CRS
House ID/
Sample ID
33/20
33/21
33/23
33/31
33/32
43/22
43/26
43/27
43/11
43/32
43/36
17/01
17/03
17/23
19/04
19/08
19/13
19/25
19/26
17/19
43/24
19/28
19/36
80/06
80/24
80/26
80/27
80/09
80/45
80/39
80/41
51/12
51/20
51/24
51/26
33/19
43/16
Concentration (//g/g)*
Al
609.89
Ba
30315.04
29402.19
22466.22
Cd
14.43
6.58
4.53
0.94
220.60
241.07
4.85
615.27
5.39
13.98
9.30
6.19
1.72
3.86
Cr
94.18
523.19
951.74
515.97
676.48
13.75
16.00
268.94
186.60
16.92
5.59
22.45
99.07
Ni
238.11
Ti
422.12
55.00
104.36
0.38
753.13
165.58
181.30
564.27
44.14
Zn
2866.97
502.00
231.35
1520.83
35121.27
972.71
5963.48
533.06
306.14
* No outliers were detected for calcium, magnesium, lead and potassium.
B-9
-------
Table B-4. Outliers Detected by Multivariate Methods
Sample
Processing
Batch
CIS
CKC
CRS
SSS
SSS
SSS
House
17
17
80
51
17
33
Sample
ID
03
01
06
26
23
23
Concentration (//g/g)
Pb
253.91
50
61573.85
345.81
363.88
135.78
Al
6949.83
1694
609.89
7761.56
19585.58
26178.44
Ba
1841.07
742
30315.04
206.56
439.75
401 .46
Cd
13.39
3.1
30.83
3.86
241 .07
14.43
Ca
23113.58
14246
21251.35
5934.11
14160.18
12471.77
Cr
29.35
16.19
151.36
24.57
268.94
951.74
Mg
3950.1 1a
2724
5080.89
303.99
614.15
848.89
Ni
16.27
13
42.43
11.18
238.11
13.06
K
17158.68
14419
1536.03
2224.18
4570.6
6241.22
Ti
104.36a
55.07
181.3
306.4
582.48
667.37
Zn
1338.25
502
35121.27
313.77
499.30
243.15
-------
REPORT DOCUMENTATION PAGE
Form Approved
OMB No 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions,
searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments
regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington
Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to
the Office of Management and Budget, Paperwork Reduction Project (0704-0188), Washington, DC 20503.
1. AGENCY USE ONLY (Leave blank)
2. REPORT DATE
July 1998
3. REPORT TYPE AND DATES COVERED
Final Report
4. TITLE AND SUBTITLE
Comprehensive Abatement Performance Pilot Study
Volume II: Multi-element Data Analyses
6. AUTHOR(s)
John Kinateder and Z. James Ma
5. FUNDING NUMBERS
C: 68-D5-0008
7. PERFORMING ORGANIZATION NAME(s) AND ADDRESS(ES)
Battelle Memorial Institute
505 King Avenue
Columbus, Ohio 43201
8. PERFORMING ORGANIZATION
REPORT NUMBER
Not Applicable
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES)
U.S. Environmental Protection Agency
Office of Pollution Prevention and Toxics
401 M Street SW (7401)
Washington, D.C. 20460
10. SPONSORING/MONITORING
AGENCY
REPORT NUMBER
EPA 747-R-98-002
11. SUPPLEMENTARY NOTES
Other Battelle staff involved in the production of this report included Bruce Buxton, Steve Rust, Tamara Collins, Fred
Todt, Nancy McMillan, Matt Palmgren, Nick Sasso, Robin Hertz, and Casey Boudreau. Key Midwest Research Institute
(MRI) staff included Gary Dewalt, Paul Constant, Jim McHugh, and Jack Balsinger.
12.a DISTRIBUTION/AVAILABILITY STATEMENT
12b. DISTRIBUTION CODE
13. ABSTRACT (Maximum 200 words)
This report presents the results of the statistical analysis of multi-element data collected during a pilot study that
preceded the Comprehensive Abatement Performance (CAP) Study. The goal of the CAP Study was to assess the long-
term efficacy of lead-based paint abatement. In this report, multi-element analysis was undertaken to determine
whether relationships among lead, aluminum, barium, cadmium, calcium, chromium, magnesium, nickel, potassium,
titanium, and zinc in dust and soil samples could provide a way to identify the source and pathways of lead in
households. This report summarizes the results of the investigation of the multi-element analysis.
14. SUBJECT TERMS
Lead, Multi-elements, Dust, Soil, Elements in Dust and Soil, Metals, Lead-Based Paint
Abatement, Renovation, Principal Components Analysis, Correlation Scatterplot, Pairwise
Correlation, Dot Plots.
15. NUMBER OF PAGES
72
16. PRICE CODE
17. SECURITY CLASSIFICATION
OF REPORT
Unlimited
18. SECURITY CLASSIFICATION
OF THIS PAGE
Unlimited
19. SECURITY
CLASSIFICATION
OF ABSTRACT
Unlimited
20. LIMITATION OF
ABSTRACT
NSN 7540-01-280-5500
Prescribed by ANSI Std. Z39-18
Standard Form 298 (Rev 2-89)
298-102
------- |