EPA 747-R-96-008
May 1997
LEAD EXPOSURE ASSOCIATED WITH
RENOVATION AND REMODELING ACTIVITIES:
N
ENVIRONMENTAL FIELD SAMPLING STUDY
VOLUME II: APPENDICES
Prepared By
Battelle
505 King Avenue
Columbus, OH 43201
for
Technical Programs Branch
Chemical Management Division
Office of Pollution Prevention and Toxics
Office of Prevention, Pesticides, and Toxic Substances
U.S. Environmental Protection Agency
Washington, DC 20460
0 S Environmental Protection Agency
Region 5, Library (PL-12J) f]M
77 West Jackson Boulevard, IZtn floor
Chicago, IL 60604-3590
-------
DISCLAIMER
Mention of trade names, products, or services does not
convey, and should not be interpreted as conveying, official EPA
approval, endorsement, or recommendations.
This report is copied on recycled paper.
Appendices ii
-------
TABLE OF CONTENTS
Page
A.O SUPPORT TABLES AND FIGURES OF EFSS DATA A-2
A.1 Descriptive Statistics on Sample Weights, Lead Loadings, and Lead
Concentrations by Sample Type, in the Carpet Removal Phase A-2
A.2 Descriptive Statistics on Sample Weights, Lead Loadings, and Lead
Concentrations by Sample Type, in the Window Replacement Phase A-20
A.3 Descriptive Statistics of Data from the CED Phase A-38
A.4 Descriptive Statistics of Data from the Cleanup Investigation A-64
B.O DATA PROCESSING AND OUTLIER DETECTION B-2
B.1 Data Processing B-2
B.1.1 Data Storage, Transfer, and Tracking B-2
B.1.2 Data Verification B-2
B.1.3 Data Manipulation B-3
B.2 Outlier Detection B-3
B.2.1 Logic Checks B-3
B.2.2 Formal Statistical Tests B-6
B.2.3 Graphical Review B-11
B.2.4 Outlier Summary B-15
C.O STATISTICAL METHODS AND MODELS IN THE EFSS C-2
C.1 Models in the Carpet Removal and Window Replacement Phases to
Evaluate Factors or Measurements in Relation to Lead Exposure C-2
C.2 Estimating Components to Total Variability in Log-transformed Lead
Loading Data in the Carpet Removal and Window Replacement Phases C-8
C.3 Modeling the Relationship Between Lead Exposure and Distance from
Activity Area in the Window Replacement Phase C-9
C.4 Models in the CED Phase to Evaluate Factors or Measurements in Relation
to Lead Exposure C-10
C.4.1 Variance Components Associated with Worker Personal
Exposure Levels C-11
C.4.2 Differences in Worker Personal Exposure Lead Levels that
are Attributable to Substrate (Plaster Versus Wood) C-13
C.4.3 Association Between Potential Lead Hazard to Occupants
and CED Activity/Substrate Combinations C-15
C.4.4 Differences in the Relationship Between Lead in Settled Dust
and Distance that can be Attributed to Substrate (Plaster
Versus Wood) C-17
C.4.5 Investigating the Substrate Effect on Worker Personal
Exposure Monitoring Results after Adjusting for the Effects
of Pre-activity Paint Lead Loading C-20
C.4.6 Investigating the Substrate Effect on the Relationship
Between Lead Loading in Settled Dust and Distance After
Adjusting for the Effects of Pre-Activity Paint Lead Loading C-21
C.5 Quantifying Lead Disturbance in a 6' X 1' Dustfall Gradient C-22
C.6 Methodology Underlying Normal Distribution Theory Confidence Intervals,
Confidence Interval for a Percentile, and Confidence Interval for the
Proportion less (Greater) than a Specified Value C-23
Appendices iii
-------
TABLE OF CONTENTS
(Continued)
Page
C.6.1 Normal Theory Two-Sided 100(1-a) Percent Confidence
Interval on the p-th Distribution Percentile C-25
C.6.2 Normal Theory Two-Sided 100(1-a) Percent Confidence
Interval on P(Y>a) C-30
C.7 Models Used to Estimate Mean Exposure Levels Across Studies in
Combining Other Sources of Surface Preparation Data C-32
D.O QUALITY CONTROL D-2
D.1 Field Blanks D-2
D.2 Field Side-by-side Samples D-4
D.3 Summary of Laboratory QA/QC Findings D-8
LIST OF TABLES
Table CR-1. Numbers of Field Samples Collected and Results Reported in the EFSS
Carpet Removal Substudy, by Study Unit A-3
Table CR-2a. Descriptive Statistics (Across All Units) of Lead Concentrations (//g/m3)
Associated with Personal Air and Ambient Air Samples During Carpet
Removal A-3
Table CR-2b. Descriptive Statistics (for Each Unit) of Lead Concentrations (//g/m3)
Associated with Personal Air and Ambient Air Samples During Carpet
Removal A-4
Table CR-3a. Descriptive Statistics (Across All Units) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from Floors
Collected Before Carpet Removal ("pre") and at 1-hour Following
Completion of Carpet Removal ("post") A-5
Table CR-3b. Descriptive Statistics (for Each Unit) of Sample Weights (g) Associated
with Vacuum Samples from Floors Collected Before Carpet Removal ("pre")
and at 1-hour Following Completion of Carpet Removal ("post") A-6
Table CR-3c. Descriptive Statistics (for Each Unit) of Lead Loadings (//g/ft2) Associated
with Vacuum Samples from Floors Collected Before Carpet Removal ("pre")
and at 1-hour Following Completion of Carpet Removal ("post") A-7
Table CR-3d. Descriptive Statistics (for Each Unit) of Lead Concentrations (//g/g) Associated
with Vacuum Samples from Floors Collected Before Carpet Removal ("pre")
and at 1-hour Following Completion of Carpet Removal ("post") A-8
Table CR-4a. Descriptive Statistics (Across All Units) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from Stainless
Steel Dustfall Collectors at 1-hour and 2-hours Following Completion of
Carpet Removal A-9
Table CR-4b. Descriptive Statistics (for Each Unit) of Sample Weights (g) Associated with
Vacuum Samples from Stainless Steel Dustfall Collectors at 1-hour and
2-hours Following Completion of Carpet Removal A-10
Table CR-4c. Descriptive Statistics (for Each Unit) of Lead Loadings fc/g/ft2) Associated
with Vacuum Samples from Stainless Steel Dustfall Collectors at 1-hour
and 2-hours Following Completion of Carpet Removal A-11
Appendices iv
-------
TABLE OF CONTENTS
(Continued)
Page
Table CR-4d. Descriptive Statistics (for Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Stainless Steel Dustfall Collectors
at 1-hour and 2-hours Following Completion of Carpet Removal A-12
Table CR-5a. Descriptive Statistics (Across All Units) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from Window
Sills Collected Before Carpet Removal ("pre") and at 1-hour Following
Completion of Carpet Removal ("post") A-13
Table CR-5b. Descriptive Statistics (for Each Unit) of Sample Weights (g) Associated
with Vacuum Samples from Window Sills Collected Before Carpet Removal
("pre") and at 1-hour Following Completion of Carpet Removal ("post") A-14
Table CR-5c. Descriptive Statistics (for Each Unit) of Lead Loadings (//g/ft2)
Associated with Vacuum Samples from Window Sills Collected Before
Carpet Removal ("pre") and at 1-hour Following Completion of Carpet
Removal ("post") A-15
Table CR-5d. Descriptive Statistics (for Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Window Sills Collected Before
Carpet Removal ("pre") and at 1-hour Following Completion of Carpet
Removal ("post") A-16
Table CR-6a. Descriptive Statistics (Across All Units) of Lead Loadings (//g/ft2) Associated
with Vacuum and Wipe Samples from Stainless Steel Dustfall Collectors
Collected at 1-hour Following Completion of Carpet Removal A-17
Table CR-6b. Descriptive Statistics (for Each Unit) of Lead Loadings (//g/ft2) Associated
with Vacuum and Wipe Samples from Stainless Steel Dustfall Collectors
Collected at 1-hour Following Completion of Carpet Removal A-17
Table WR-1. Numbers of Field Samples Collected and Results Reported by Unit in the
R&R/EFSS Window Replacement Phase A-21
Table WR-2a. Descriptive Statistics (Across All Units) of Lead Concentrations (//g/m3)
Associated With Personal Air, Pre-Activity Area Air, and During-Activity Area
Air Samples Collected During Window Replacement A-22
Table WR-2b. Descriptive Statistics (Within Each Unit) of Lead Concentrations (fjg/m3)
Associated With Personal Air, Pre-Activity Area Air, and During-Activity Area
Air Samples Collected During Window Replacement A-22
Table WR-3a. Descriptive Statistics (Across All Units) of Sample Weight (g) Associated
with Vacuum Samples from Floors Collected Before ("pre") and at One
Hour Following Completion of Window Replacement ("post") A-23
Table WR-3b. Descriptive Statistics (Across All Units) of Lead Loading (//g/ft2) Associated
with Vacuum Samples from Floors Collected Before ("pre") and at One Hour
Following Completion of Window Replacement ("post") A-23
Table WR-3c. Descriptive Statistics (Across All Units) of Lead Concentration (//g/g)
Associated With Vacuum Samples from Floors Collected Before ("pre") and
at One Hour Following Completion of Window Replacement ("post") A-24
Table WR-3d. Descriptive Statistics (Within Each Unit) of Sample Weight (g) Associated
with Vacuum Samples from Floors Collected Before ("pre") and One Hour
Following Completion of Window Replacement ("post") -- at a Distance of
0 Feet from the Windows A-24
Appendices v
-------
TABLE OF CONTENTS
(Continued)
Page
Table WR-3e. Descriptive Statistics (Within Each Unit) of Sample Weight (g) Associated
with Vacuum Samples from Floors Collected Before ("pre") and One Hour
Following Completion of Window Replacement ("post") - at a Distance of
3 Feet from the Windows A-25
Table WR-3f. Descriptive Statistics (Within Each Unit) of Sample Weight (g) Associated
with Vacuum Samples from Floors Collected Before ("pre") and One Hour
Following Completion of Window Replacement ("post") -- at a Distance of
6 Feet from the Windows A-25
Table WR-3g. Descriptive Statistics (Within Each Unit) of Lead Loadings (//g/ft2)
Associated with Vacuum Samples from Floors Collected Before ("pre") and
One Hour Following Completion of Window Replacement ("post") - at a
Distance of 0 Feet from the Windows A-26
Table WR-3h. Descriptive Statistics (Within Each Unit) of Lead Loadings (//g/ft2)
Associated with Vacuum Samples from Floors Collected Before ("pre") and
One Hour Following Completion of Window Replacement ("post") -- at a
Distance of 3 Feet from the Windows A-26
Table WR-3L Descriptive Statistics (Within Each Unit) of Lead Loadings (//g/ft2)
Associated with Vacuum Samples from Floors Collected Before ("pre") and
One Hour Following Completion of Window Replacement ("post") -- at a
Distance of 6 Feet from the Windows A-27
Table WR-3J. Descriptive Statistics (Within Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Floors Collected Before ("pre")
and One Hour Following Completion of Window Replacement ("post") — at
a Distance of 0 Feet from the Windows A-27
Table WR-3k. Descriptive Statistics (Within Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Floors Collected Before ("pre") and
One Hour Following Completion of Window Replacement ("post") -- at a
Distance of 3 Feet from the Windows A-28
Table WR-31. Descriptive Statistics (Within Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Floors Collected Before ("pre") and
One Hour Following Completion of Window Replacement ("post") -- at a
Distance of 6 Feet from the Windows A-28
Table WR-4a. Descriptive Statistics (Across All Units) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from Stainless
Steel Dustfall Collectors at One and Two Hours Following Completion of
Window Replacement A-29
Table WR-4b. Descriptive Statistics (for Each Unit) of Sample Weight (g) Associated with
Vacuum Samples from Stainless Steel Dustfall Collectors at One and Two
Hours Following Completion of Carpet Removal A-30
Table WR-4c. Descriptive Statistics (for Each Unit) of Lead Loading (//g/ft2) Associated
with Vacuum Samples from Stainless Steel Dustfall Collectors at One and
Two Hours Following Completion of Carpet Removal A-31
Table WR-4d. Descriptive Statistics (for Each Unit) of Lead Concentration (//g/g) Associated
with Vacuum Samples from Stainless Steel Dustfall Collectors at One and
Two Hours Following Completion of Carpet Removal A-32
Table WR-5a. Descriptive Statistics (Across All Units) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from Window
Wells Taken Prior to Window Replacement A-33
Appendices vi
-------
TABLE OF CONTENTS
(Continued)
Page
Table WR-5b. Descriptive Statistics (Within Each Unit) of Sample Weights, Lead
Loadings, and Lead Concentrations Associated with Vacuum Samples
from the Window Wells Taken Prior to Window Replacement A-33
Table WR-6a. Descriptive Statistics (Across All Units) of Sample Weight, Lead Loadings,
and Lead Concentrations Associated with Interior and Exterior Paint Chip
Samples Collected Following Completion of Window Replacement A-34
Table WR-6b. Descriptive Statistics (Within Each Unit) of Sample Weight, Lead Loadings,
and Lead Concentrations Associated with Interior and Exterior Paint Chip
Samples Collected Following Completion of Window Replacement A-35
Table CED-1. Summary Statistics for Task-Length Average Personal Worker Lead Levels
(/;g/m3) During Each Combination of CED Activity and Substrate A-40
Table CED-2. Parameter Estimates P0 and P, from Model (A-1) Predicting Dust Lead
Loadings as a Function of Distance for Each Target Activity Execution A-41
Table CI-1. Descriptive Statistics for Measured Lead Loadings (mg/cm2) in Paint Chip
Samples Taken from Surfaces Disturbed by R&R Activity in the Cleanup
Investigation A-64
Table CI-2. Descriptive Statistics for Measured Lead Loadings (//g/ft2) in Post-Activity
Settled Dust Samples Taken Prior to Cleanup in the Cleanup Investigation . . . A-65
Table CI-3. Descriptive Statistics for Measured Lead Loadings (/yg/ft2) in Post-Cleanup
Settled Dust Samples Taken in the Cleanup Investigation A-66
Table CI-4. Descriptive Statistics for Measured Lead Loadings (//g/ft2) in Next-Day
SSDC Vacuum-Dust Samples Taken in the Cleanup Investigation A-67
Table B-1. Sample Pairs in the Carpet Removal and Window Replacement Phases
Whose Lead Loadings Were Flagged by Logic Checks B-5
Table B-2. Data Values in the Carpet Removal Phase Identified as Outliers by Formal
Statistical Tests B-8
Table B-3. Data Values in the Window Replacement Phase Identified as Outliers by
Formal Statistical Tests B-9
Table B-4. Data Values in the CED Phase Identified as Outliers by Formal
Statistical Tests B-10
Table B-5. Sample Pairs in the Carpet Removal Phase Identified as Outliers by
Graphical Review and Regression Analysis B-13
Table B-6. Sample Results in the Window Replacement Phase Identified as Outliers
by Graphical Review and Regression Analysis B-14
Table C-1. Data Categories and Covariates Considered in Fitting Model Form (C-1)
to Lead Loading Data in the Carpet Removal Phase C-4
Table C-2. Data Categories and Covariates Considered in Fitting Model Form (C-1 a)
to Lead Loading Data in the Window Replacement Phase C-5
Table C-3. Parameter Estimates (and Standard Errors) from Fitting Statistical Models
to Evaluate Factors Relating to Lead Disturbance in Carpet Removal Data .... C-6
Table C-4a. Parameter Estimates (and Standard Errors) from Fitting Model (WR-1)
to Evaluate Factors Relating to Lead Disturbance for Personal Exposure
Samples in the Window Replacement Phase C-6
Table C-4b. Parameter Estimates (and Standard Errors) from Fitting Model (WR-2)
to Evaluate Factors Relating to Lead Disturbance for During-Activity
Area Air Samples in the Window Replacement Phase C-7
Appendices vii
-------
TABLE OF CONTENTS
(Continued)
Page
Table C-4c. Parameter Estimates (and Standard Errors) from Fitting Models (WR-3)
and (WR-4) to Evaluate Factors Relating to Lead Disturbance for SSDCs
at Zero and Six Feet from Windows Being Removed C-7
Table C-5. Statistical Models Considered in the CED Phase to Characterize Lead
Loadings C-11
Table C-6. Estimated Variance Components and Geometric Mean Resulting from
Fitting Model (CED-1) to Personal Exposure Results for a Given CED
Activity/Substrate Combination C-12
Table C-7a. Parameter Estimates from Fitting Model (CED-2) to Personal Exposure
Results for Drilling, Sawing, and Cleanup Activities C-14
Table C-7b. Results of the Likelihood Ratio Testing Procedure for Comparing Model
(CED-2) Relative to Model (CED-1) C-14
Table C-8. Estimates of the Intercept and Slope (and Associated Standard Errors)
from the Two Stage and the Population Models of Settled Dust Lead
Loading as a Function of Distance C-17
Table C-9a. Parameter Estimates from Fitting Model (CED-6) to Settled Dust Lead
Loadings for Drilling and Sawing Activities C-19
Table C-9b. Results of the Likelihood Ratio Testing Procedure for Comparing
Model (CED-6) to Model (CED-5) C-19
Table C-10. Parameter Estimates from Fitting Model (CED-7) to Personal Exposure
Results for Drilling, Sawing, and Cleanup Activities C-20
Table C-11. Parameter Estimates from Fitting Model (CED-8) to SSDC Lead Loading
Data for Drilling and Sawing Activities C-22
Table D-1a. Descriptive Summaries of Field Blank Sample Results Within the
Carpet Removal Phase D-3
Table D-1b. Descriptive Summaries of Field Blank Sample Results Within the
Window Replacement Phase D-3
Table D-1c. Descriptive Summaries of Field Blank Sample Results Within the
CED Phase D-4
Table D-2a. Sample Loadings, Concentrations, and Weights for Each Sample Pair
Defined by a Regular Dust Sample and an Adjoining Side-by-Side QC
Sample, in the R&R Carpet Removal Phase D-5
Table D-2b. Sample Loadings, Concentrations, and Weights for Each Sample Pair
Defined by a Regular Dust Sample and an Adjoining Side-by-Side QC
Sample, in the R&R Window Replacement Phase D-7
Table D-2c. Loadings for Each Sample Pair Defined by a Regular Dust Sample and an
Adjoining Side-by-Side QC Sample, Both Collected by Vacuum
Techniques, in the R&R CED Phase D-8
Table D-3a. Summary of Conclusions Made in Analysis of Laboratory QA/QC
Samples in the Carpet Removal Phase D-9
Table D-3b. Summary of Conclusions Made in Analysis of Laboratory QA/QC
Samples in the Window Replacement Phase D-10
Table D-3c. Summary of Conclusions Made in Analysis of Laboratory QA/QC
Samples in the CED Phase D-11
Appendices viii
-------
TABLE OF CONTENTS
(Continued)
Page
LIST OF FIGURES
Figure CR-1a. Boxplots of Lead Loadings (//g/ft2) for Three Types of Dust Samples
Taken Prior to Carpet Removal Activities A-18
Figure CR-1b. Boxplots of Lead Concentrations U/g/g) for Three Types of Dust
Samples Taken Prior to Carpet Removal Activities A-18
Figure CR-2a. Boxplots of Lead Loadings (/vg/ft2) for Five Types of Post-Activity
Settled Dust Samples A-19
Figure CR-2b. Boxplots of Lead Concentrations (//g/g) for Four Types of Post-Activity
Settled Dust Samples A-19
Figure WR-1a. Boxplots of Lead Loadings (jug/ft2) for Vacuum Dust Samples From the
Floors and Window Wells Taken Prior to Window Replacement Activities . . . A-36
Figure WR-1b. Boxplots of Lead Concentrations (/vg/g) for Vacuum Dust Samples From
Floors and Window Wells Taken Prior to Window Replacement Activities . . . A-36
Figure WR-2a. Boxplots of Lead Loadings (jug/ft2) for Vacuum Dust Samples From Floors
and Stainless Steel Dustfall Collectors Taken One-Hour After Completion
of the Window Replacement Activities A-37
Figure WR-2b. Boxplots of Lead Concentrations (jwg/g) for Vacuum Dust Samples From
Floors and Stainless Steel Dustfall Collectors Taken One-Hour After
Completion of the Window Replacement Activities A-37
Figure WR-3. Boxplots of Lead Loadings (mg/cm2) for Paint Chip Samples Taken from
the Interior and Exterior of the Window Frames After Completion of the
Window Replacement Activities A-38
Figure CED-1 A-43
Figure CED-2 A-46
Figure CED-3 A-48
Figure CED-4 A-51
Figure CED-5 A-53
Figure CED-6 A-56
Figure CED-7 A-58
Figure CED-8 A-60
Figure B-1. Scatterplot of Paint Chip Log Lead Loadings versus Paint Chip Log Lead
Concentrations for the Demolition CED Activities B-12
Appendices ix
-------
APPENDIX A
SUPPORTING TABLES AND FIGURES OF EFSS DA TA
A-1
-------
A.O SUPPORT TABLES AND FIGURES OF EFSS DATA
A.1 DESCRIPTIVE STATISTICS ON SAMPLE WEIGHTS. LEAD LOADINGS.
AND LEAD CONCENTRATIONS BY SAMPLE TYPE. IN THE CARPET
REMOVAL PHASE
This section presents descriptive summaries on lead exposure data for the various sample
types considered in the EFSS carpet removal substudy. These summaries, cited in Chapters 6
and 7A, are presented within six sets of tables, grouped according to the sample type considered:
Total number of samples collected and analyzed: Table CR-1.
Personal and ambient air results: Tables CR-2a, CR-2b.
Pre- and post-activity floor vacuum dust results: Tables CR-3a through CR-3d.
Stainless steel dustfall collector fSSDC) dust sample results using vacuum techniques.
collected at 1-hour and 2-hours following the activity: Tables CR-4a through CR-4d.
Pre- and post-activity window sill dust results: Tables CR-5a through CR-5d.
SSDC dust sample results collected at 1-hour following the activity, using vacuum and
wipe techniques: Tables CR-6a, CR-6b.
Table CR-1 reports the number of samples planned, the number of samples collected, and the
number of analytical results received. The remaining tables present summary statistics, such as
the number of samples with nonmissing data values, the arithmetic and geometric means, the
standard deviation of the log-transformed data, and the minimum and maximum data values.
Within each set of tables, the first table represents statistics over all study data, while subsequent
tables include statistics calculated for each study unit. Lead exposure data summarized in these
tables include the physical sample weights (g) and lead concentrations (ng/g) for vacuum
samples, lead loadings ((^g/ft2) for vacuum and wipe samples, and lead concentrations (ug/m3)
for personal air and ambient air samples. The personal air lead concentrations are expressed in
terms of task-length average (TLA) exposures for a given worker, defined as the average
exposure over the duration of activity.
Sample types are placed within a given set of tables above so that their results can be
compared. Therefore, these tables also include differences in sample results between pairs of
adjoining samples at a given location. Tables CR-3a and CR-3b consider differences in results
between adjoining pre- and post-activity floor dust samples. Tables CR-4a through CR-4d
consider differences between 1-hour and 2-hour results from SSDC vacuum dust samples.
Tables CR-5a through CR-5d include differences in results between adjoining pre- and post-
activity window sill samples, and Tables CR-6a through CR-6b include differences between 1-
hour vacuum and wipe SSDC sample results. The same statistics are presented for the paired
differences as for the individual sample results. However, the statistic identified as the
"geometric mean" represents the geometric mean of the ratio between the two adjoining sample
results at a given location (e.g., post- vs. pre-activity, 2-hour vs. 1-hour, wipe vs. vacuum).
A-2
-------
In addition to the above tables, this section also includes boxplots of lead loading and
concentration data for the different types of settled dust samples. Figures CR-la and CR-lb
present lead loading and concentration data, respectively, for the three types of settled dust
samples collected prior to the start of carpet removal activity (pre-activity carpet data, window
sill data, and floor data). Figures CR-2a and CR-2b contain boxplots for lead loading and
concentration data, respectively, for the five types of settled dust samples collected following
completion of carpet removal activities (1-hour vacuum from floors and SSDCs, 1-hour wipe
from SSDCs, and 2-hour vacuum from SSDCs). Chapter 5 includes a discussion of how to
interpret boxplots, as well as the boxplots for personal air and ambient air sample lead loadings.
Table CR-1. Numbers of Field Samples Collected and Results Reported in the EFSS Carpet
Removal Substudy, by Study Unit
Unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Total
Number of
Proposed
Samples111
Reg.
26
26
26
26
26
26
26
26
208
QC
9
9
9
9
9
9
9
9
72
Number of
Proposed Samples
Collected121
Reg,
26
26
26
26
25
26
25
26
206
QC
9
9
9
9
9
9
9
9
72
Number of Extra
Samples
Collected13'
2
0
0
0
0
0
1
0
3
Total Number of
Samples
Collected
37
35
35
35
34
35
35
35
281
Number of
Analytical
Results
Received
37
35
35
35
34
35
35
35
281
ID
12}
A breakdown of the number of proposed samples by sample type is presented in Table 7A-1.
Two proposed samples were not collected: both were personal air samples at units where a single R&R worker
performed carpet removal.
Extra samples consisted of additional cassette samples necessary for personal air monitoring over the duration of carpet
removal activity.
Table CR-2a.
Descriptive Statistics (Across All Units) of Lead Concentrations (//g/m3)
Associated with Personal Air and Ambient Air Samples During Carpet
Removal
Sample Type
During-Activity
Personal Exposure
Samples121
Pre-Activity Ambient
Air Samples
During-Activity
Ambient Air Samples
N
14
8
16
Arithmetic
Mean
35.87
0.10
1.48
Geometric
Mean
8.44
0.09
0.33
Log Std.
Dev,01
1.77
0.43
1.58
Minimum
Value
0.86
0.05
0.06
Maximum Value
221.3
0.17
13.38
ID
121
Standard deviation of the log-transformed data (expressed in log measurement units).
Results summarize worker exposure over the entire Job. Three of the 14 workers in this study had multiple cassette
samples taken within non-overlapping time intervals during the activity. For these workers, cumulative results over the
entire job were calculated from the multiple samples.
A-3
-------
Table CR-2b. Descriptive Statistics (for Each Unit) of Lead Concentrations (//g/m3)
Associated with Personal Air and Ambient Air Samples During Carpet
Removal
Unit ID
N
Arithmetic
Mean
Geometric
Mean
Log Std.
Oev,"1
Minimum
Value
Maximum
Value
Baseline (Pro-
Activity)
Value
Personal Air Sample Results121
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
2
2
2
2
1
2
1
2
55.58
3.81
7.00
174.7
1.48
4.56
7.39
0.97
55.41
3.67
6.85
168.4
1.48
4.08
7.39
0.97
0.11
0.40
0.30
0.39
0.68
0.16
51.25
2.77
5.56
128.1
59.91
4.86
8.44
221.3
1.48
2.52
6.60
7.39
0.86
1.08
Ambient Air Sample Results
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
2
2
2
2
2
2
2
2
2.87
0.13
0.10
6.83
0.12
0.22
1.45
0.14
1.42
0.11
0.10
1.92
0.12
0.19
1.44
0.14
1.88
0.89
0.20
2.75
0.16
0.88
0.16
0.29
0.38
0.06
0.09
0.28
0.11
0.10
1.29
0.11
5.36
0.20
0.12
13.38
0.14
0.35
1.61
0.17
0.10
0.09
0.17
0.07
0.09
0.05
0.14
0.06
111 Standard deviation of the log-transformed data (expressed in log measurement units).
121 For units 1-01 and 2-03, results summarize worker exposure over the entire iob. Three of the 14 workers
in this study had multiple cassette samples taken within non-overlapping time intervals during the activity.
For these workers, cumulative results over the entire job were calculated from the multiple samples.
A-4
-------
Table CR-3a. Descriptive Statistics (Across All Units) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from Floors
Collected Before Carpet Removal ("pre") and at 1-hour Following Completion
of Carpet Removal ("post")111
Data Representation
N
Arithmetic
Mean
Geometric
Mean
tog Std.
Dev.(ZI
Minimum
Value
Maximum
Value
Sample Weight (g)
Pre-activity
Post-activity
Paired Differences
(post minus pre)
24
24
24
0.0607
0.1677
0.1070
0.0304
0.1205
3.9666'31
1.1323
0.8087
0.0057
0.0357
-.2910
0.4536
0.6285
0.5179
Loadings (//g/ft2)
Pre-activity
Post-activity
Paired Differences
(post minus pre)
24
24
24
84.14
591.3
507.1
14.44
130.4
9.03131
1.98
1.67
1.38
6.38
-195
564.5
6135
6132
Concentrations (//g/g)
Pre-activity
Post-activity
Paired Differences
(post minus j>re)
24
24
24
1336.8
2875.8
1539.0
475.1
1081.4
2.3'31
1.4
1.4
36.7
71.9
-3554
9179.2
20662
20571
111 Only results for regular samples (i.e., no side-by-side QC samples) are represented in this table.
(2) Standard deviation of the log-transformed data (expressed in log measurement units).
131 Geometric mean of the ratio of the post-activity result to the result of the adjoining pre-activity sample.
No measurement units are associated with this value.
A-5
-------
Table CR-3b. Descriptive Statistics (for Each Unit) of Sample Weights (g) Associated with
Vacuum Samples from Floors Collected Before Carpet Removal ("pre") and at
1-hour Following Completion of Carpet Removal ("post")111
Unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Time
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
N
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
Arithmetic
Mean
0.0112
0.1800
0.0395
0.1526
0.0337
0.0851
0.0299
0.3225
0.0543
0.0751
0.0096
0.0802
0.1691
0.1435
0.1386
0.3026
Geometric
Mean
0.0110
0.1545
0.0344
0.0993
0.0216
0.0848
0.0263
0.2801
0.0499
0.0746
0.0089
0.0654
0.0583
0.1149
0.1313
0.2186
Log Std.
Oev.'a
0.2343
0.7018
0.6261
1.1448
1.1183
0.1111
0.6463
0.6707
0.5330
0.1409
0.4566
0.7558
1.9238
0.9053
0.4234
1.0012
Minimum
Value
0.0090
0.0741
0.0207
0.0357
0.0113
0.0747
0.0132
0.1394
0.0272
0.0646
0.0057
0.0390
0.0100
0.041 1
0.0806
0.0858
Maximum
Value
0.0142
0.2999
0.0693
0.3419
0.0784
0.0919
0.0476
0.5311
0.0741
0.0856
0.0142
0.1556
0.4536
0.2267
0.1723
0.6285
Paired Differences (Post minus pre)131
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
3
3
3
3
3
3
3
3
0.1688
0.1131
0.0515
0.2926
0.0208
0.0706
-.0256
0.1640
14.099
2.8810
3.9342
10.670
1.4963
7.3181
1.9713
1.6649
0.0599
-.0336
0.0135
0.0918
0.0010
0.0302
-.2910
0.0052
0.2909
0.3212
0.0775
0.5179
0.0374
0.1499
0.2167
0.4656
m
(2)
(3)
Only results for regular samples (i.e., no side-by-side QC samples) are represented in this table.
Standard deviation of the log-transformed data (expressed in log measurement units).
The geometric mean column contains the geometric mean of the ratio of the post-activity result to the
result of the adjoining pre-activity sample. No measurement units are associated with this value.
A-6
-------
Table CR-3c. Descriptive Statistics (for Each Unit) of Lead Loadings (//g/ft2) Associated
with Vacuum Samples from Floors Collected Before Carpet Removal ("pre")
and at 1-hour Following Completion of Carpet Removal ("post")11'
Unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Time
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
N
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
Arithmetic
Mean
1.70
146.2
12.45
35.02
13.97
108.5
3.19
2183
280.9
494.7
2.86
185.8
32.49
167.7
325.5
1410
Geometric
Mean
1.67
107.5
11.43
28.27
6.30
36.87
3.09
637.2
188.1
192.1
2.86
103.9
29.89
119.9
315.2
488.2
Log Std.
Dev.'21
0.23
0.99
0.52
0.77
1.52
1.93
0.29
1.96
1.20
1.70
0.01
1.64
0.52
1.17
0.30
1.83
Minimum
Value
1.38
41.28
6.70
16.69
2.61
6.38
2.61
183.2
52.21
54.87
2.85
15.77
16.66
30.92
261.4
119.6
Maximum
Value
2.16
295.5
18.78
68.68
36.68
292.3
4.34
6135
564.5
1332
2.89
316.6
45.77
240.2
447.5
3857
Paired Differences (Post minus pre)13'
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
3
3
3
3
3
3
3
3
144.5
22.58
94.54
2179
213.8
182.9
135.2
1084
64.50
2.47
5.85
205.9
1.02
36.28
4.01
1.55
39.90
-2.09
3.77
178.8
-171
12.92
-4.12
-195
293.9
61.98
255.6
6132
767.8
313.7
223.5
3595
(1)
(2)
(3)
Only results for regular samples (i.e., no side-by-side QC samples) are represented in this table.
Standard deviation of the log-transformed data (expressed in log measurement units).
The geometric mean column contains the geometric mean of the ratio of the post-activity result to the
result of the adjoining pre-activity sample. No measurement units are associated with this value.
A-7
-------
Table CR-3d. Descriptive Statistics (for Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Floors Collected Before Carpet
Removal ("pre") and at 1-hour Following Completion of Carpet Removal
("post")111
unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Time
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
N
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
Arithmetic
Mean
152.9
868.9
337.0
304.8
310.2
1204.0
126.7
7469.8
4716.0
5932.2
342.6
2910.9
1805.6
1084.0
2903.7
3231.8
Geometric
Mean
152.0
695.5
331.7
284.8
292.6
434.9
117.9
2274.7
3773.3
2575.2
320.8
1590.3
513.1
1043.5
2401.1
2233.8
Lag Std.
Dev.'4'
0.1
0.9
0.2
0.4
0.4
1.9
0.4
2.0
0.8
1.6
0.4
1.4
2.4
0.3
0.7
1.2
Minimum
Value
133.9
248.7
271.1
200.9
231.4
71.9
90.8
433.5
1919.3
730.6
203.8
404.2
36.7
752.2
1553.5
617.5
Maximum
Value
172.9
1372.8
416.4
467.6
467.9
3180.4
198.1
20662
9179.2
15564
500.1
6883.0
4576.5
1477.2
5552.6
6136.6
Paired Differences (Post minus pre)131
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
3
3
3
3
3
3
3
3
716.1
-32.2
893.8
7343.1
1216.3
2568.4
-721.6
328.2
4.6
0.9
1.5
19.3
0.7
5.0
2.0
0.9
114.8
-170.5
-159.5
235.5
-2319
80.3
-3554
-2611
1221.1
196.6
2712.5
20571
6385.1
6679.2
1440.5
4531.7
01 Only results for regular samples (i.e., no side-by-side QC samples) are represented in this table.
121 Standard deviation of the log-transformed data (expressed in log measurement units).
(31 The geometric mean column contains the geometric mean of the ratio of the post-activity result to the
result of the adjoining pre-activity sample. No measurement units are associated with this value.
A-8
-------
Table CR-4a. Descriptive Statistics (Across All Units) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from Stainless
Steel Dustfall Collectors at 1-hour and 2-hours Following Completion of
Carpet Removal'1'
Data Representation
N
Arithmetic
Mean
Geometric
Mean
Log Std.
Oev.1*'
Minimum
Value
Maximum
Value
Sample Weight (g)
One hour
post-activity
Two hours
post-activity
Paired Differences
(2-hr minus 1-hr)
24
24
24
0.0288
0.0423
0.0135
0.0223
0.0305
1 .3682(3)
0.6999
0.8151
0.0077
0.0043
-.0452
0.1127
0.2245
0.2087
Loadings (//g/ft2)
One hour
post-activity
Two hours
post-activity
Paired Differences
(2-hr minus 1-hr)
24
24
24
72.68
109.8
37.14
24.33
38.63
1.59(3)
1.50
1.50
2.61
2.61
-193
621.0
937.8
316.8
Concentrations fc/g/g)
One hour
post-activity
Two hours
post-activity
Paired Differences
(2-hr minus 1-hr)
24
24
24
2427.8
2935.1
507.4
1089.9
1264.8
1 .2'31
1.2
1.2
214.3
280.8
-6244
19839
29867
10028
111 Only results for regular samples (i.e., no side-by-side QC samples) are represented in this table.
121 Standard deviation of the log-transformed data (expressed in log measurement units).
131 Geometric mean of the ratio of the two-hour result to the result of the adjoining one-hour sample. No
measurement units are associated with this value.
A-9
-------
Table CR-4b. Descriptive Statistics (for Each Unit) of Sample Weights (g) Associated with
Vacuum Samples from Stainless Steel Dustfall Collectors at 1-hour and 2-
hours Following Completion of Carpet Removal111
Unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Time Post-
Activity
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
N
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
Arithmetic
Mean
0.0499
0.0548
0.0435
0.0361
0.0206
0.0276
0.0269
0.0465
0.0190
0.0232
0.0172
0.0235
0.0264
0.0981
0.0266
0.0286
Geometric
Mean
0.0327
0.0495
0.0341
0.0327
0.0195
0.0229
0.0260
0.0422
0.0165
0.0226
0.0139
0.0148
0.0203
0.0516
0.0235
0.0280
Log Std.
Dev.'zl
1.1205
0.5957
0.8809
0.5296
0.4229
0.8154
0.3296
0.5207
0.6880
0.2671
0.7768
1.2455
0.8773
1.5458
0.6520
0.2593
Minimum
Value
0.0127
0.0249
0.0141
0.0234
0.0122
0.0091
0.0178
0.0311
0.0079
0.0192
0.0077
0.0043
0.0098
0.0103
0.0114
0.0212
Maximum
Value
0.1127
0.0721
0.0821
0.0602
0.0277
0.0425
0.0317
0.0770
0.0309
0.0308
0.0335
0.0519
0.0537
0.2245
0.0405
0.0354
Paired Differences (2-hr, minus 1-hr)131
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
3
3
3
3
3
3
3
3
0.0049
-.0073
0.0070
0.0196
0.0042
0.0063
0.0717
0.0020
1.5139
0.9592
1.1752
1 .6208
1.3741
1 .0620
2.5482
1.1937
-.0452
-.0219
-.0031
0.0001
-.0001
-.0061
0.0005
-.0068
0.0477
0.0093
0.0205
0.0453
0.0117
0.0184
0.2087
0.0179
111 Only results for regular samples (i.e., no side-by-side QC samples) are represented in this table.
121 Standard deviation of the log-transformed data (expressed in log measurement units).
131 The geometric mean column contains the geometric mean of the ratio of the two-hour result to the result
of the adjoining one-hour sample. No measurement units are associated with this value.
A-10
-------
Table CR-4c. Descriptive Statistics (for Each Unit) of Lead Loadings (//g/ft2) Associated
with Vacuum Samples from Stainless Steel Dustfall Collectors at 1 -hour and
2-hours Following Completion of Carpet Removal111
Unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Time Post-
Activity
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
N
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
Arithmetic
Mean
103.3
80.47
18.66
14.34
9.44
13.04
282.1
437.1
82.21
38.96
39.98
143.7
34.93
117.9
10.79
33.04
Geometric
Mean
60.55
56.31
13.88
13.07
7.55
8.95
197.7
316.5
36.87
34.67
10.04
40.83
27.72
77.76
9.53
21.57
Log Std.
Dav.(Z1
1.48
1.10
0.98
0.53
0.92
1.16
1.00
0.96
1.53
0.57
2.10
2.37
0.82
1.21
0.59
1.23
Minimum
Value
11.65
18.25
5.10
7.74
2.61
2.61
96.91
154.3
13.67
23.00
2.85
3.33
13.96
22.10
6.78
5.95
Maximum
Value
205.7
163.2
36.51
22.39
13.35
25.89
621.0
937.8
216.0
66.69
114.0
372.9
68.60
244.7
18.82
68.58
Paired Differences (2-hr, minus 1-hr)131
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
3
3
3
3
3
3
3
3
-22.8
-4.32
3.60
155.0
-43.3
103.7
82.98
22.24
0.93
0.94
1.19
1.60
0.94
4.07
2.81
2.26
-42.5
-14.1
-2.74
57.40
-193
0.22
8.14
-0.83
6.60
2.65
13.53
316.8
53.01
259.0
222.5
49.76
(1)
(2)
(3)
Only results for regular samples (i.e., no side-by-side QC samples) are represented in this table.
Standard deviation of the log-transformed data (expressed in log measurement units).
The geometric mean column contains the geometric mean of the ratio of the two-hour result to the result
of the adjoining one-hour sample. No measurement units are associated with this value.
A-11
-------
Table CR-4d. Descriptive Statistics (for Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Stainless Steel Dustfall Collectors at
1-hour and 2-hours Following Completion of Carpet Removal'1'
Unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Time Post-
Activity
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
1-hr.
2-hrs.
N
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
Arithmetic
Mean
2179.4
1327.5
408.8
407.5
419.3
412.2
10038
12972
3216.1
1854.9
1357.3
3922.3
1369.9
1565.5
433.7
1018.9
Geometric
Mean
1852.5
1137.9
407.2
399.9
387.1
390.4
7592.7
7499.2
2238.5
1532.1
722.3
2766.9
1368.3
1506.2
405.9
769.8
Log Std.
Dev.'»
0.7
0.7
0.1
0.2
0.5
0.4
0.9
1.4
1.0
0.8
1.3
1.1
0.1
0.3
0.5
1.0
Minimum
Value
917.7
732.9
361.6
330.9
214.3
287.3
3057.2
2004.0
927.2
746.7
299.2
774.4
1277.5
1090.0
242.2
280.8
Maximum
Value
3795.5
2418.2
444.6
519.7
561.6
609.1
19839
29867
6990.3
3402.4
3402.4
7185.7
1424.7
2145.3
594.3
1937.3
Paired Differences (2-hr, minus 1-hr)131
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
3
3
3
3
3
3
3
3
-851.9
-1.2
-7.1
2934.6
-1361
2565.0
195.6
585.2
0.6
1.0
1.0
1.0
0.7
3.8
1.1
1.9
-2964
-72.7
-141.8
-1053
-6244
475.2
-317.5
38.5
593.1
99.6
73.0
10028
1671.6
3783.4
720.6
1472.7
111 Only results for regular samples (i.e., no side-by-side QC samples) are represented in this table.
121 Standard deviation of the log-transformed data (expressed in log measurement units).
(3> The geometric mean column contains the geometric mean of the ratio of the two-hour result to the result
of the adjoining one-hour sample. No measurement units are associated with this value.
A-12
-------
Table CR-5a. Descriptive Statistics (Across All Units) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from Window
Sills Collected Before Carpet Removal ("pre") and at 1-hour Following
Completion of Carpet Removal ("post")
Data Representation
Pre-activity
Post-activity
Paired Differences
(post minus pre)
N
Arithmetic
Mean
Geometric
Mean
Sample Weight (g)
16
16
16
0.2535
0.3087
0.0552
0.0970
0.1380
1 .4235(2)
tog Std.
Oev.">
Minimum
Value
1 .6046
1.3282
0.0059
0.0195
-.7865
Maximum
Value
1.1394
1.6537
1.4214
Loadings d/g/ft2)
Pre-activity
Post-activity
Paired Differences
(post minus pre)
16
16
16
4208
4404
196.2
417.5
661.3
1.58(2)
2.32
2.22
21.58
11.81
-14879
41459
26581
16392
Concentrations (//g/g)
Pre-activity
Post-activity
Paired Differences
(post minus pre)
16
16
16
7396.5
7878.4
481.9
2161.8
2303.8
1.1'2'
1.7
1.5
102.0
250.0
-6248
52985
66776
13791
111 Standard deviation of the log-transformed data (expressed in log measurement units).
121 Geometric mean of the ratio of the post-activity result to the result of the adjoining pre-activity sample.
No measurement units are associated with this value.
A-13
-------
Table CR-5b. Descriptive Statistics (for Each Unit) of Sample Weights (g) Associated with
Vacuum Samples from Window Sills Collected Before Carpet Removal
("pre") and at 1-hour Following Completion of Carpet Removal ("post")
Unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Time
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
N
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
Arithmetic
Mean
0.1686
0.1162
0.0618
0.2488
0.1776
0.0831
0.6859
1 .0033
0.2120
0.5123
0.0124
0.0238
0.6808
0.4092
0.0292
0.0729
Geometric
Mean
0.1685
0.1162
0.0617
0.1199
0.0677
0.0661
0.5145
0.7639
0.1205
0.2869
0.0106
0.0234
0.6797
0.3910
0.0249
0.0713
Log Std.
Dev.'1'
0.0138
0.0037
0.0103
1.9222
2.2901
0.9920
1.1245
1.0922
1 .6473
1.6731
0.8232
0.2583
0.0815
0.4299
0.8161
0.3033
Minimum
Value
0.1669
0.1159
0.0613
0.0308
0.0134
0.0328
0.2323
0.3529
0.0376
0.0879
0.0059
0.0195
0.6416
0.2885
0.0140
0.0575
Maximum
Value
0.1702
0.1165
0.0622
0.4668
0.3417
0.1334
1.1394
1.6537
0.3863
0.9366
0.0189
0.0281
0.7200
0.5299
0.0444
0.0883
Paired Differences (Post minus pre)121
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
2
2
2
2
2
2
2
2
-.0523
0.1871
-.0945
0.3175
0.3003
0.0114
-.2716
0.0437
0.6894
1.9418
0.9776
1 .4849
2.3808
2.2167
0.5753
2.8580
-.0543
-.0305
-.2083
-.7865
0.0503
0.0006
-.3531
0.0131
-.0504
0.4046
0.0194
1.4214
0.5503
0.0222
-.1901
0.0743
111 Standard deviation of the log-transformed data (expressed in log measurement units).
121 The geometric mean column contains the geometric mean of the ratio of the post-activity result to the
result of the adjoining pre-activity sample. No measurement units are associated with this value.
A-14
-------
Table CR-5c. Descriptive Statistics (for Each Unit) of Lead Loadings (//g/ft2) Associated
with Vacuum Samples from Window Sills Collected Before Carpet Removal
("pre") and at 1-hour Following Completion of Carpet Removal ("post")
Unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Time
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
N
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
Arithmetic
Mean
535.2
157.9
1850
990.0
82.46
393.4
111.5
1236
5576
16052
64.31
155.6
25395
16098
51.62
152.9
Geometric
Mean
339.2
156.8
1612
812.3
71.67
392.1
111.2
1012
4795
13449
48.06
59.47
19668
12217
46.70
74.08
Log Std.
Dev.11'
1.46
0.17
0.76
0.92
0.77
0.12
0.09
0.92
0.80
0.87
1.13
2.29
1.05
1.10
0.64
1.91
Minimum
Value
121.19
139.05
942.19
424.12
41.69
361.44
104.06
526.05
2730.3
7289.5
21.58
11.81
9330.3
5615.3
29.63
19.14
Maximum
Value
949.2
176.7
2757
1556
123.2
425.4
118.9
1946
8422
24814
107.0
299.5
41459
26581
73.60
286.7
Paired Differences (Post minus pre)121
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
2
2
2
2
2
2
2
2
-377
-860
311.0
1125
10475
91.33
-9297
101.3
0.46
0.50
5.47
9.10
2.80
1.24
0.62
1.59
-772.5
-1201
238.22
407.19
4559.2
-9.77
-14879
-54.46
17.86
-518
383.7
1842
16392
192.4
-3715
257.1
111 Standard deviation of the log-transformed data (expressed in log measurement units).
<2i The geometric mean column contains the geometric mean of the ratio of the post-activity result to the
result of the adjoining pre-activity sample. No measurement units are associated with this value.
A-15
-------
Table CR-5d.
Descriptive Statistics (for Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Window Sills Collected Before
Carpet Removal ("pre") and at 1-hour Following Completion of Carpet
Removal ("post")
Unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Time
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
Pre
Post
N
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
Arithmetic
Mean
3182.3
1042.3
4538.6
1294.0
735.5
3404.2
412.4
3304.9
27328
34308
3462.6
2234.3
17029
16439
2482.6
999.8
Geometric
Mean
1639.6
1030.4
3960.2
1027.9
506.2
2833.2
271.6
1663.9
9411.9
11088
1791.3
1026.9
13665
14755
2319.7
946.0
Log Std.
Dev.™
1.8
0.2
0.8
1.0
1.3
0.9
1.4
1.9
2.4
2.5
1.8
2.0
1.0
0.7
0.5
0.5
Minimum
Value
454.9
885.0
2321.6
508.0
201.9
1517.0
102.0
449.4
1671.9
1841.0
499.4
250.0
6867.1
9191.2
1598.2
676.2
Maximum
Value
5909.6
1199.7
6755.5
2080.0
1269.0
5291.5
722.9
6160.4
52985
66776
6425.9
4218.5
27192
23687
3366.9
1323.3
Paired Differences (Post minus pre)12'
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
2
2
2
2
2
2
2
2
-2140
-3245
2668.8
2892.4
6980.0
-1228
-590.1
-1483
0.6
0.3
5.6
6.1
1.2
0.6
1.1
0.4
-5025
-6248
1315.1
-273.5
169.2
-2207
-3504
-2691
744.8
-241.6
4022.4
6058.3
13791
-249.4
2324.0
-274.9
111 Standard deviation of the log-transformed data (expressed in log measurement units).
121 The geometric mean column contains the geometric mean of the ratio of the post-activity result to the
result of the adjoining pre-activity sample. No measurement units are associated with this value.
A-16
-------
Table CR-6a. Descriptive Statistics (Across All Units) of Lead Loadings (//g/ft2) Associated
with Vacuum and Wipe Samples from Stainless Steel Dustfall Collectors
Collected at 1-hour Following Completion of Carpet Removal'1'
Data Representation
Vacuum
Wipe
Paired Difference (Wipe
minus vacuum)
N
16
16
16
Arithmetic
Mean
85.59
58.84
-26.8
Geometric
Mean
23.04
27.06
1.17131
Log Std.
Dev.la
1.65
,_ 1.31
: Minimum
Value
2.61
3.19
-287
Maximum
Value
621.0
333.6
77.96
'" Only results for regular samples (i.e., no side-by-side QC samples) are represented in this table. Only regular vacuum
samples located side-by-side with a wipe sample are included in the calculation of vacuum sample results.
121 Standard deviation of the log-transformed data (expressed in log measurement units).
131 Geometric mean of the ratio of the wipe result to the result of the adjoining sample collected by vacuum. No
measurement units are associated with this value.
Table CR-6b.
Descriptive Statistics (for Each Unit) of Lead Loadings (//g/ft2) Associated
with Vacuum and Wipe Samples from Stainless Steel Dustfall Collectors
Collected at 1-hour Following Completion of Carpet Removal'11
Unit ID
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
Sample
Type
Vacuum
Wipe
Vacuum
Wipe
Vacuum
Wipe
Vacuum
Wipe
Vacuum
Wipe
Vacuum
Wipe
Vacuum
Wipe
Vacuum
Wipe
N
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
Arithmetic Mean
108.7
48.82
9.74
21.56
7.48
35.46
358.9
254.2
116.5
22.14
58.55
48.29
18.10
32.69
6.78
7.51
Geometric
Mean
48.95
35.66
8.56
19.29
5.68
22.06
245.3
241.5
60.54
22.09
18.83
17.82
17.62
32.46
6.78
6.14
Log Std.
Dev*'
2.03
1.18
0.73
0.68
1.10
1.49
1.31
0.46
1.80
0.09
2.55
2.34
0.33
0.17
0.00
0.93
Minimum
Vatue
11.65
15.48
5.10
11.94
2.61
7.70
96.91
174.9
16.97
20.66
3.11
3.41
13.96
28.81
6.78
3.19
Maximum
Value
205.7
82.16
14.37
31.18
12.35
63.22
621.0
333.6
216.0
23.62
114.0
93.17
22.24
36.57
6.78
11.82
Paired Differences (wipe minus vacuum)131
1-01
1-02
1-03
1-04
2-01
2-02
2-03
2-05
2
2
2
2
2
2
2
2
-59.9
11.83
27.98
-105
-94.3
-10.3
14.59
0.72
0.73
2.25
3.89
0.98
0.36
0.95
1.84
0.91
-124
6.84
5.09
-287
-192
-20.8
6.57
-3.59
3.83
16.81
50.87
77.96
3.69
0.30
22.61
5.04
Only results for regular samples (i.e., no side-by-side QC samples) are represented in this table. Only regular vacuum
samples located side-by-side with a wipe sample are included in the calculation of vacuum sample results.
Standard deviation of the log-transformed data (expressed in log measurement units).
The geometric mean column contains the geometric mean of the ratio of the wipe result to the result of the adjoining
sample collected by vacuum. No measurement units are associated with this value.
A-17
-------
e-
sr
1000CXH
10000
1000
100
10
1-
Pre-Activity
carpet
(n=24)
Pre-Activity
window sill
(n-16)
Sample Type
Pre-Activity
floor vacuum
(n=24)
Figure CR-1a. Boxplots of Lead Loadings (//g/ft2) for Three Types of Dust Samples Taken
Prior to Carpet Removal Activities
100000
10000
1000
100
10
1
Pre-Activity
carpet
(n-24)
Pre-Activity
window sill
(n-16)
Sample Type
Pre-Activity
floor vacuum
(n-24)
Figure CR-1b. Boxplots of Lead Concentrations (//g/g) for Three Types of Dust Samples
Taken Prior to Carpet Removal Activities
A-18
-------
e-
sr
100000
10000
1000
100
10
1-1
1-Hour 1-Hour 1-Hour 1-Hour 2-Hour
Post-Activity Post-Activity Post-Activity Post-Activity Post-Actfvfly
window Sid floor vacuum wipe vacuum (SSDC) vacuum (SSDC)
(n-16) (n-24) (n-16) (n-24) (n-24)
Sample Type
SSDC - Stainless Steel Oustfall Collectors
Figure CR-2a. Boxplots of Lead Loadings (//g/ft2) for Five Types of Post-Activity Settled
Dust Samples
J
5
100000
10000
1000
100
10
1
1-Hour
Post-Activity
window sil
(n-16)
SSDC
1-Hour
Post-Activity
floor vacuum
(n-24)
1-Hour
Post-Activity
vacuum (SSOC)
(n-24)
2-Hour
Post-Activity
vacuum (SSOC)
(n-24)
Sample Type
Stainless Steel Dustfall Collectors
Figure CR-2b. Boxplots of Lead Concentrations (//g/g) for Four Types of Post-Activity
Settled Dust Samples
A-19
-------
A.2 DESCRIPTIVE STATISTICS ON SAMPLE WEIGHTS. LEAD LOADINGS.
AND LEAD CONCENTRATIONS BY SAMPLE TYPE. IN THE WINDOW
REPLACEMENT PHASE
This section presents descriptive summaries on lead exposure data for the various sample
types considered in the EFSS window replacement phase. These summaries, cited in Section 7B
and Chapter 6, are presented within six sets of tables, grouped according to the sample type
considered:
Total number of samples collected and analyzed: Table WR-1.
Personal and ambient air results: Tables WR-2a, WR-2b.
Pre- and post-activity floor vacuum dust results: Tables
WR-3a through WR-31.
Stainless steel dustfall collector (SSDC) dust sample
results using vacuum techniques, collected at 1 -hour and
2-hours following the activity: Tables WR-4a through WR-4d.
Pre-activity window well dust results: Tables WR-5a, WR-5b.
Interior and Exterior Paint Chip results: Tables WR-6a, WR-6b.
Table WR-1 reports the number of samples planned, the number of samples collected, the
number of analytical results received, and the number of analytical results used in the statistical
analysis. The remaining tables present summary statistics, including the number of samples with
non-missing data values, the arithmetic and geometric means, the standard deviation of the log-
transformed data, and the minimum and maximum data values. Within each set of tables, the
first table represents statistics over all study data, while subsequent tables include statistics
calculated for each study unit. Lead exposure data included in these tables include the physical
sample weights (g) and lead concentrations (|ag/g) for vacuum samples, lead loadings (|ig/ft2) for
vacuum samples, lead loadings for paint chip samples (mg/cm2) and lead concentrations (|ug/m3)
for personal air and ambient air samples. The personal air lead concentrations are expressed in
terms of task-length average (TLA) exposures for a given worker, defined as the average
exposure over the duration of activity.
Similar sample types are placed within a given set of tables above so that their results can
be compared. These comparisons are made using the differences in sample results between pairs
of adjoining samples at a given location. Tables WR-2a through WR-21 present differences in
results between adjoining pre- and post-activity floor dust samples. Tables WR-3a through WR-
3d present differences between 1-hour and 2-hour results from SSDC vacuum dust samples. The
same statistics are presented for the paired differences as for the individual sample results.
However, the statistics identified as the "geometric mean" represents the geometric mean of the
ratio between the two adjoining sample results at a given location (e.g., post- vs. pre-activity, 2-
hourvs. 1-hour).
A-20
-------
In addition to the above tables, this section also includes boxplots of lead loading and
concentration data for the different types of settled dust samples. Figures WR-la and WR-lb
present lead loadings and lead concentrations, respectively, for four pre-activity settled dust
samples (1 window well, 3 floor). Figures WR-2a and WR-2b present lead loadings and lead
concentrations, respectively, for five post-activity settled dust samples (3 floor, 2 SSDC). Figure
WR-3 presents boxplots of interior and exterior paint chip lead loadings from removed windows.
Chapter 5 includes a discussion of how to interpret boxplots.
Table WR-1. Numbers of Field Samples Collected and Results Reported by Unit in the
R&R/EFSS Window Replacement Phase
Unit
ID
1-01
2-01
3-01
4-01
Number of
Proposed Samples
Beg.
42
42
42
42
ac
6
6
6
6
Number of Proposed
Samples Collected"'
Beg.
42
40
39
41
QC
6
6
6
6
Number of Extra
Samples
Collected1*
Reg.
1
8
3
3
QC
1
0
0
1
Total
Number of
Samples
Collected
50
54
48
51
Number of
Analytical Results
Receded131
47
50
47
49
Number of
Analytical Results
Used In Analysis'41
46
50
46
49
The number of proposed samples collected differs from number of proposed samples for units 2-01, 3-01 and 4-01. For
unit 2-01, one pre- and one post-activity settled dust (floor) sample at 3 feet were not collected. For unit 3-01, two
post-activity tarpaulin samples and one post-activity settled dust (floor) sample at 3 feet were not collected. For unit 4-
01, one post-activity tarpaulin sample was not collected.
Two additional regular samples were collected at unit 2-01: one post-activity tarpaulin and one exterior paint chip. One
additional 2-hour post-activity settled dust (stainless steel dust collector) sample was collected at unit 3-01. Two
additional QC samples, both paint chip field blanks, were collected at units 1-01 and 4-01. Additional personal air
monitor samples collected when filled cassettes were replaced account for the other regular samples in this column.
Tarpaulin samples were collected but not analyzed. They will be archived.
One window well sample from unit 1-01 was deleted from the analysis because it was identified as an outlier. One 2-
hour post-activity settled dust (stainless steel dust collector) sample was deleted from analysis because of a protocol
deviation. This collector was placed after activity had been completed.
A-21
-------
Table WR-2a. Descriptive Statistics (Across All Units) of Lead Concentrations (//g/m3)
Associated With Personal Air, Pre-Activity Area Air, and During-Activity
Area Air Samples Collected During Window Replacement
Sample Type
During-Activity
Personal
Exposure Samples
Pre-Activity
Ambient Air
Samples
Post-Activity
Ambient Air
Samples
N
8
4
8
Arithmetic
Mean
13.95
0.83
1.54
Geometric
Mean
7.48
0.30
1.16
Log
Std.
Dev.(1)
1.19
1.58
0.82
Minimum
Value
2.41
0.10
0.29
Maximum
Value
44.29
2.86
4.16
111 Standard deviation of the log transformed data (expressed in log measurement units).
Table WR-2b. Descriptive Statistics (Within Each Unit) of Lead Concentrations (//g/m3)
Associated With Personal Air, Pre-Activity Area Air, and During-Activity
Area Air Samples Collected During Window Replacement
Unit
ID
N
Arithmetic
Mean
Geometric
Mean
log Std.
Dev.m
Minimum
Vatue
Maximum
Value
Personal Air Sample Results
1-01
2-01
3-01
4-01
2
2
2
2
2.88
3.45
37.98
11.50
2.87
3.29
37.46
8.85
0.17
0.44
0.24
1.07
2.55
2.41
31.67
4.15
3.22
4.49
44.29
18.84
Baseline
(Pre-Activity
Value)
Ambient Air Sample Results
1-01
2-01
3-01
4-01
2
2
2
2
0.56
1.00
3.41
1.18
0.50
0.55
3.33
1.17
0.74
0.48
0.32
0.16
0.29
0.68
2.66
1.05
0.84
1.33
4.16
1.31
0.10
0.10
0.27
2.86
Standard deviation of the log transformed data (expressed in log measurement units).
A-22
-------
Table WR-3a.
Descriptive Statistics (Across All Units) of Sample Weight (g) Associated
With Vacuum Samples from Floors Collected Before ("pre") and at One
Hour Following Completion of Window Replacement ("post")
Data
Representation
Pre-activity
Pre-activity121
Pre-activity131
Post-activity
Post-activity141
Post-activity131
Paired Difference
(Post minus Pre)
Paired Difference
(Post minus Pre)13-41
Paired Difference
(Post minus Pre)13-61
Distance
(feet)
0
3
6
0
3
6
0
3
6
N
12
11
16
12
10
16
12
10
15
Arithmetic
Mean
4.1668
0.7965
0.6935
3.1297
0.9499
0.9403
-1.0371
0.3552
0.2584
Geometric
Mean
0.7219
0.3436
0.1588
0.7992
0.2878
0.2498
1.1 070161
1 .0335161
1 .4730151
Log Std.
Dav.'11
2.5855
1.8286
2.4357
2.5822
2.3679
2.0747
Minimum
Value
0.0076
0.0055
0.0008
0.0058
0.0037
0.0088
-14.7512
-0.3251
-0.8229
Wlaximum
Value
20.4911
2.8148
2.9056
7.1463
3.3809
3.2372
4.4551
2.8118
2.2498
111
121
131
14)
15)
Standard deviation of the log-transformed data (expressed in log measurement units).
One pre-activity floor sample at 3 feet was not collected at unit 2-01.
Results include both regular and QC samples.
Two post-activity floor samples at 3 feet were not collected: one at unit 2-01 and one at unit 3-01.
Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
One pre-activity/post-activity pair is excluded because the pre-activity sample location differed from the post-activity
sample location.
Table WR-3b. Descriptive Statistics (Across All Units) of Lead Loading (//g/ft2) Associated
With Vacuum Samples from Floors Collected Before ("pre") and at One
Hour Following Completion of Window Replacement ("post")
Data
Representation
Pre-activity
Pre-activity12'
Pre-activity131
Post-activity
Post-activity141
Post-activity131
Paired Difference
(Post minus Pre)
Paired Difference
(Post minus Pre)13'41
Paired Difference
(Post minus Pre)13'61
Distance
(feet)
0
3
6
0
3
6
0
3
6
N
12
11
16
12
10
16
12
10
15
Arithmetic
Mean
58016.8
5424.6
3562.0
35499.4
5504.4
9357.9
-22517.3
448.8
6179.1
Geometric
Mean
1913.3
490.6
334.3
3912.1
1293.5
878.4
2.0""
3.5'5'
2.6151
Log Std.
D«V,"'
3.63
3.22
2.99
3.02
2.43
3.01
Minimum
Value
9.0
2.3
1.7
19.0
56.0
10.4
-354160.0
-19060.0
-540.3
Maximum
Value
439845.0
29402.0
18443.2
109740.0
12702.0
54515.0
92013.0
10856.4
36071.8
111
121
131
141
151
Standard deviation of the log-transformed data (expressed in log measurement units).
One pre-activity floor sample at 3 feet was not collected at unit 2-01.
Results include both regular and QC samples.
Two post-activity floor samples at 3 feet were not collected: one at unit 2-01 and one at unit 3-01.
Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
One pre-activity/post-activity pair is excluded because the pre-activity sample location differed from the post-activity
sample location.
A-23
-------
Table WR-3c. Descriptive Statistics (Across All Units) of Lead Concentration (//g/g)
Associated With Vacuum Samples from Floors Collected Before ("pre")
and at One Hour Following Completion of Window Replacement ("post")
Data
Representation
Pre-activity
Pre-activity121
Pre-activity13'
Post-activity
Post-activity1*'
Post-activity13'
Paired Difference
(Post minus Pre)
Paired Difference
(Post minus Pre)13'4'
Paired Difference
(Post minus Pre)13-6'
Distance
(feet)
0
3
6
0
3
6
0
3
6
N
12
11
16
12
10
16
12
10
15
Arithmetic
Mean
7047.0
5735.5
4243.5
11299.6
11112.0
6898.3
4251.7
5126.8
2838.0
Geometric
Mean
2650.2
1503.0
2105.1
4886.6
4494.7
3516.2
1.816'
3.215'
1 .7'61
tog Std.
Oev.'11
1.58
1.74
1.37
1.80
2.19
1.25
•
•
Minimum
Value
228.5
155.6
136.0
234.8
31.8
632.4
-15767.1
-27343.0
-5856.75
Maximum
Value
33003.6
37114.5
16470.1
26874.7
32346.0
31728.1
23378.0
31925.8
28448.5
111
12)
13)
14)
15)
Standard deviation of the log-transformed data (expressed in log measurement units).
One pre-activity floor sample at 3 feet was not collected at unit 2-01.
Results include both regular and QC samples.
Two post-activity floor samples at 3 feet were not collected: one at unit 2-01 and one at unit 3-01.
Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
One pre-activity/post-activity pair is excluded because the pre-activity sample location differed from the post-activity
sample location.
Table WR-3d. Descriptive Statistics (Within Each Unit) of Sample Weight (g) Associated
With Vacuum Samples from Floors Collected Before ("pre") and One
Hour Following Completion of Window Replacement ("post") -- at a
Distance of 0 Feet from the Windows
Unit
ID
1-01
2-01
3-01
4-01
Data
Representation
Kre-Activity
Post-Activity
Paired Difference
(Post minus pre)
Pre-Activity
Post-Activity
Paired Difference
(Post minus pre)
Pre-Activity
Post-Activity
Paired Difference
(Post minus pre)
Pre-Activity
Post-Activity
Paired Difference
(Post minus pre)
N
d
3
3
3
3
3
3
3
3
3
3
3
Arithmetic
Mean
U.U014
0.0414
-0.0100
1.1503
3.6090
2.4588
3.7352
4.9993
1.2641
11.7303
3.8689
-7.8613
Geometric
Mean
U.U4bO
0.0212
0.4522121
0.2333
1.3251
5.678812'
3.4471
4.7908
1 .389812'
7.2101
3.0337
0.420812'
Log Std.
Dev."1
U.0121
1.4559
3.0698
2.4222
0.4749
0.3486
•
1.4526
0.9652
Minimum
Value
U.U34/
0.0058
-0.0292
0.0076
0.0809
0.0733
2.5883
3.7681
1.1145
1.3725
0.9993
-14.7512
Maximum
Value
U.UB^b
0.1024
0.0178
2.8584
5.7063
4.4551
5.9636
7.1463
1.4951
20.4911
5.7399
-0.3732
"' Standard deviation of the log-transformed data (expressed in log measurement units).
121 Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
A-24
-------
Table WR-3e. Descriptive Statistics (Within Each Unit) of Sample Weight (g) Associated
with Vacuum Samples from Floors Collected Before ("pre") and One Hour
Following Completion of Window Replacement ("post") - at a Distance of 3
Feet from the Windows
Unit
»D
1-01
2-0 1131
3-0 1141
4-01
Data
Representation
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
N
3
3
3
2
2
2
3
2
2
3
3
3
Arithmetic
Mean
0.0727
0.0550
-0.0177
0.3519
1.8266
1 .4747
1.6617
1.1396
0.0544
0.9516
1.1339
0.1823
Geometric
Mean
0.0337
0.0146
0.4328121
0.2769
0.9595
3.4655121
1 .4741
0.9819
0.9204'2'
0.9422
1.1213
1.1901121
tog Stct.
Dev.m
1.7291
2.0594
1.0189
1.7812
.
0.5900
0.7911
0.1726
0.1808
Minimum
Value
0.0055
0.0037
-0.0352
0.1347
0.2723
0.1376
0.8863
0.5612
-0.3251
0.7922
0.9700
-0.1487
Maximum
Value
0.1721
0.1559
-0.0018
0.5691
3.3809
2.8118
2.8148
1.7179
0.4339
1.1187
1.3732
0.4294
111 Standard deviation of the log-transformed data (expressed in log measurement units).
121 Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
131 One set of the pre- and post-activity samples at this unit were not collected due to space and time constraints.
141 The post-activity sample was not collected. Sampler ran out of sampling bottles.
Table WR-3f. Descriptive Statistics (Within Each Unit) of Sample Weight (g) Associated
with Vacuum Samples from Floors Collected Before ("pre") and One Hour
Following Completion of Window Replacement ("post") -- at a Distance of 6
Feet from the Windows11'
Unit
ID
1-01
2-01
3-01
4-01
Data
Representation
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)141
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
N
4
4
3
4
4
4
4
4
4
4
4
4
Arithmetic
Mean
0.0466
0.0500
-0.0196
0.0744
0.1002
0.0258
1.1280
1.5855
0.4575
1.5252
2.0256
0.5004
Geometric
Mean
0.0184
0.0413
1.8163131
0.0252
0.0409
1 .6228131
1.0180
1.3877
1 .3632131
1.3453
1 .6609
1 .2346'31
Log Std.
Dev.«'
2.1805
0.7165
1.6986
1.6376
0.5600
0.6043
0.5583
0.7565
Minimum
Value
0.0008
0.0224
-0.0438
0.0062
0.0088
0.0009
0.4699
0.7324
-0.8229
0.7727
0.7517
-0.5511
Maximum
Value
0.0967
0.0952
0.0223
0.2505
0.2981
0.0478
1.5553
2.7197
2.2498
2.9056
3.2372
1.9870
'" Includes QC samples and regular samples.
121 Standard deviation of the log-transformed data (expressed in log measurement units).
131 Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
141 One pre-activity/post-activity pair is excluded because the pre-activity sample location differed from the post-activity
sample location.
A-25
-------
Table WR-3g. Descriptive Statistics (Within Each Unit) of Lead Loadings (//g/ft2)
Associated with Vacuum Samples from Floors Collected Before ("pre") and
One Hour Following Completion of Window Replacement ("post") -- at a
Distance of 0 Feet from the Windows
Unit
ID
1-01
2-01
3-01
4-01
Data
Representation
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
N
3
3
3
3
3
3
3
3
3
3
3
3
Arithmetic
Mean
79.4
246.4
167.0
419.3
1191.2
771.9
35699.2
96981.7
61282.5
195869.0
43578.4
-152291.0
Geometric
Mean
36.4
226.9
6.2'21
220.3
389.8
1 .812'
31474.6
96448.9
3.1121
53110.0
27459.9
0.5'2>
LOS Std.
Dev."»
1.58
0.50
4
1.78
2.62
0.61
0.13
2.76
1.35
Minimum
Value
9.0
140.0
113.7
28.4
19.0
-9.4
17727.0
84851.0
24743.0
2342.1
6224.0
-354160.0
Maximum
Value
202.9
379.1
211.1
653.0
1980.5
1403.8
60108.0
109740.0
92013.0
439845.0
85685.1
3881.9
"' Standard deviation of the log-transformed data (expressed in log measurement units).
121 Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
Table WR-3h. Descriptive Statistics (Within Each Unit) of Lead Loadings (//g/ft2)
Associated with Vacuum Samples from Floors Collected Before ("pre") and
One Hour Following Completion of Window Replacement ("post") -- at a
Distance of 3 Feet from the Windows
Unit
ID
1-01
2-0 1'3'
3-0 1141
4-01
Data
Representation
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
N
3
3
3
2
2
2
3
2
2
3
3
3
Arithmetic
Mean
15.1
884.0
868.9
93.5
84.7
-8.8
8255.4
10427.0
2601.6
11557.4
10456.0
-1101.3
Geometric
Mean
10.0
255.1
25. 512'
67.7
81.5
1 .2121
7746.0
10427.0
1 .512'
5706.5
10302.4
1 .8121
Log Std.
Dev.'11
1.29
2.00
1.20
0.39
0.46
0.00
1.45
0.21
Minimum
Value
2.3
56.0
39.8
29.0
61.7
-50.3
4623.8
10425.0
-598.0
1845.6
8324.1
-19060.0
Maximum
Value
26.8
2476.4
2449.6
158.0
107.6
32.7
11027.0
10429.0
5801.2
29402.0
12702.0
10856.4
111 Standard deviation of the log-transformed data (expressed in log measurement units).
121 Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
131 Both the pre- and post-activity samples at this unit were not collected due to space and time constraints.
141 The post-activity sample was not collected. Sampler ran out of sampling bottles.
A-26
-------
Table WR-31. Descriptive Statistics (Within Each Unit) of Lead Loadings (//g/ft2)
Associated with Vacuum Samples from Floors Collected Before ("pre") and
One Hour Following Completion of Window Replacement ("post") -- at a
Distance of 6 Feet from the Windows'11
Unit
ID
1-01
2-01
3-01
4-01
Data
Representation
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)'41
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
N
4
4
3
4
4
4
4
4
4
4
4
4
Arithmetic
Mean
20.3
104.7
96.5
41.6
70.5
28.9
6117.5
10641.9
4524.4
8068.4
26614.4
18545.9
Geometric
Mean
11.8
72.0
7.2'31
40.1
41.9
1 .013'
4048.8
9004.5
2.2<3'
6496.9
21937.5
3.4'3>
log Sttt.
Dev.<»
1.40
1.03
0.32
1.21
1.05
0.74
0.70
0.72
Minimum
Value
1.7
20.2
18.5
25.9
10.4
-20.2
1227.2
3162.9
-540.3
4272.7
9668.8
5369.0
Maximum
Value
48.8
252.8
239.7
51.5
190.4
151.8
15796.0
17136.4
11542.8
1 8443.2
54515.0
36071.8
111 Includes QC samples and regular samples.
121 Standard deviation of the log-transformed data (expressed in log measurement units).
131 Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
141 One pre-activity/post-activity pair is excluded because the pro-activity sample location differed from the post-activity
sample location.
Table WR-3J.
Descriptive Statistics (Within Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Floors Collected Before ("pre")
and One Hour Following Completion of Window Replacement ("post") --
at a Distance of 0 Feet from the Windows
unit
ID
1-01
2-01
3-01
4-01
Data
Representation
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
N
3
3
3
3
3
3
3
3
3
3
3
3
Arithmetic
Mean
1136.8
13862.3
12725.5
1650.1
301.2
-1348.9
11469.1
20958.7
9489.6
13935.6
10076.3
-3859.3
Geometric
Mean
776.9
10711.0
13.8121
944.1
294.2
0.3'21
9130.8
20132.1
2.2121
7366.1
8988.3
1 .2'2'
Log Std.
Dev.(1t
1.11
0.96
1.40
0.26
0.80
0.36
1.48
0.57
Minimum
Value
260.2
3702.4
1303.6
228.5
234.8
-3501.1
4906.9
13483.1
-133.2
1706.4
6228.4
-15767.1
Maximum
Vatue
2398.8
24129.3
23378.0
3735.9
393.0
47.4
22651.5
26874.7
20025.8
33003.6
17236.4
4521.9
Standard deviation of the log-transformed data (expressed in log measurement units).
Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
A-27
-------
Table WR-3k. Descriptive Statistics (Within Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Floors Collected Before ("pre") and
One Hour Following Completion of Window Replacement ("post") — at a
Distance of 3 Feet from the Windows
Unit
ID
1-01
2-0 1131
3-0 1141
4-01
Data
Representation
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
N
3
3
3
2
2
2
3
2
2
3
3
3
Arithmetic
Mean
324.8
19535.7
19210.8
352.0
129.3
-222.7
6427.0
12325.9
4304.6
14043.7
9200.9
-4842.7
Geometric
Mean
296.5
17469.5
58.9121
324.4
84.9
0.26121
5254.7
10619.4
1.59121
6056.6
9188.0
1.5212'
Log Std.
Dev,1"
0.56
0.57
t
0.58
1.39
0.75
0.79
1.59
0.065
Minimum
Value
155.6
10376.5
9977.8
215.4
31.8
-456.7
3238.4
6068.5
2467.4
1955.5
8581.5
-27343.0
Maximum
Value
420.2
32346.0
31925.8
488.6
226.7
11.3
12441.6
18583.4
6141.8
37114.4
9771.4
7294.4
'" Standard deviation of the log-transformed data (expressed in log measurement units).
121 Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
131 Both the pre- and post-activity samples at this unit were not collected due to space and time constraints.
141 The post-activity sample was not collected. Sampler ran out of sampling bottles.
Table WR-31. Descriptive Statistics (Within Each Unit) of Lead Concentrations (//g/g)
Associated with Vacuum Samples from Floors Collected Before ("pre") and
One Hour Following Completion of Window Replacement ("post") -- at a
Distance of 6 Feet from the Windows'1'
Unit
ID
1-01
2-01
3-01
4-01
Data
Representation
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)14'
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
Pre-Activity
Post-Activity
Paired Difference (Post minus pre)
N
4
4
3
4
4
4
4
4
4
4
4
4
Arithmetic
Mean
953.8
2351.9
1895.0
3289.0
1189.2
-2099.8
5950.0
7909.2
1959.2
6781.0
16142.9
9361.9
Geometric
Mean
641.5
1741.8
4.0'3'
1593.7
1024.0
0.6131
3977.4
6488.8
1.613'
4829.3
13208.1
2.713'
Log Std.
Dev.'21
1.15
0.94
t
1.70
0.61
0.94
0.67
0.94
0.74
Minimum
Value
136.0
688.8
-1284.8
154.0
632.4
-5856.8
2381.0
4318.5
1332.1
1809.6
5691.2
1076.9
Maximum
Value
2160.4
4263.7
4127.8
8161.3
2304.5
484.6
16345.2
17677.3
2483.5
16470.1
31728.1
28448.5
111 Includes QC samples and regular samples.
121 Standard deviation of the log-transformed data (expressed in log measurement units).
131 Geometric mean represents ratio of the post-activity result to the result of the corresponding pre-activity sample. No
measurement units are associated with this value.
141 One pre-activity/post-activity pair is excluded because the pre-activity sample location differed from the post-activity
sample location.
A-28
-------
Table WR-4a. Descriptive Statistics (Across All Units) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from Stainless
Steel Dustfall Collectors at One and Two Hours Following Completion of
Window Replacement
Data
Representation
Distance
(feet)
N
Arithmetic
Mean
Geometric
Mean
Log Std.
Dev.<"
Minimum
Value
Maximum
Value
Sample Weight (g)
One-hour
One-hour12'
Two-hour
Paired
Differences
(2-hr minus
1-hr)
0
6
6
6
12
16
4
4
4.6002
0.2710
0.1048
0.0409
1 .2420
0.0428
0.0522
1.3878131
1 .8070
2.1259
1.5476
0.5332
0.1095
0.0023
0.0097
-0.0040
21.7653
1.6381
0.2056
0.1123
Lead Loading (//g/ft2)
One-hour
One-hour121
Two-hour
Paired
Differences
(2-hr minus
1-hr)
0
6
6
6
12
16
4
4
84604.1
770.6
5252.5
5247.3
24736.3
240.6
519.6
2.9I3)
Lead Concentration (/ig/j
One-hour
One-hour'21
Two-hour
Paired
Differences
(2-hr minus
1-hr)
0
6
6
6
12
16
4
4
27274.5
29669.2
33634.9
23553.2
13826.9
5620.2
9951.8
2.1(3)
1.67
1.75
2.69
0.58
3104.0
15.4
40.4
37.7
331410.0
4155.0
20192.0
20183.7
1)
1.54
1.88
2.15
0.94
489.6
114.4
763.1
-697.9
87222.4
335193.5
98210.1
71879.3
"' Standard deviations of the log-transformed data (expressed in log measurement units).
121 Results for both regular and QC samples.
131 Geometric mean of the ratio of the 2-hour result to the result of the adjoining 1-hour sample. No measurement units are
associated with this value.
A-29
-------
Table WR-4b. Descriptive Statistics (for Each Unit) of Sample Weight (g) Associated with
Vacuum Samples from Stainless Steel Dustfall Collectors at One and Two
Hours Following Completion of Carpet Removal
Unit
ID
1-01
2-01
3-01
4-01
Data
Representation
One-hour
One-hour121
Two-hour
Paired
Differences
(2-hr minus 1-hr)
One-hour
One-hour121
Two-hour
Paired
Differences
(2-hr minus 1-hr)
One-hour
One-hour'2'
Two-hour
Paired
Differences
(2-hr minus 1-hr)
One-hour
One-hour121
Two-hour
Paired
Differences
(2-hr minus 1-hr)
Distance
(feet)
0
6
6
6
0
6
6
6
0
6
6
6
0
6
6
6
L_N
3
4
1
1
3
4
1
1
3
4
1
1
3
4
1
1
Arithmetic
Mean
0.1968
0.0066
0.0097
-0.0040
8.5455
0.9407
0.1836
0.1123
2.8128
0.0233
0.0203
0.0073
6.8456
0.1135
0.2056
0.0478
Geometric
Mean
0.1812
0.0056
0.0097
0.7080131
4.3343
0.5462
0.1836
2.5750'31
2.3217
0.0111
0.0203
1.5615'31
1.3053
0.0992
0.2056
1.3029'31
Log Std.
Dev.">
0.5038
0.6380
1.3981
1 .4693
.
0.7399
1.4289
•
2.5906
0.6714
.
•
Minimum
Vatue
0.1095
0.0031
0.0097
-0.0040
1 .8604
0.0713
0.1836
0.1123
1.2872
0.0023
0.0203
0.0073
0.1107
0.0369
0.2056
0.0478
Maximum
Value
0.2999
0.0137
0.0097
-0.0040
21.7653
1.6381
0.1836
0.1123
5.3255
0.0706
0.0203
0.0073
19.3900
0.1578
0.2056
0.0478
"' Standard deviations of the log-transformed data (expressed in log measurement units).
121 Results for both regular and QC samples.
131 Geometric mean of the ratio of the 2-hour result to the result of the adjoining 1-hour sample. No measurement units are
associated with this value.
A-30
-------
Table WR-4c. Descriptive Statistics (for Each Unit) of Lead Loading (//g/ft2) Associated
with Vacuum Samples from Stainless Steel Dustfall Collectors at One and
Two Hours Following Completion of Carpet Removal
Unit
ID
1-01
2-01
3-01
4-01
Data
Representation
One-hour
One-hour121
Two-hour
Paired
Differences
(2-hr minus
1-hr)
One-hour
One-hour'21
Two-hour
Paired
Differences
(2-hr minus
1-hr)
One-hour
One-hour121
Two-hour
Paired
Differences
(2-hr minus
1-hr)
One-hour
One-hour'21
Two-hour
Paired
Differences
(2-hr minus
1-hr)
Distance
(feet)
0
6
6
6
0
6
6
6
0
6
6
6
0
6
6
6
N
3
4
1
1
3
4
1
1
3
4
1
1
3
4
1
1
Arithmetic
Mean
11908.7
285.8
40.4
37.7
83903.3
840.4
140.1
135.5
129459.7
166.5
637.4
632.4
113144.7
1789.6
20192.0
20183.7
Geometric
Mean
8417.5
70.3
40.4
2.6S39'3'
26367.0
327.8
140.1
1 .3450'31
60404.1
105.2
637.4
4.2953'31
27927.6
1382.7
20192.0
4.8597'31
Log Std.
Dev.'1'
0.98
1.89
1.90
1.48
.
•
1.54
1.31
.
2.32
0.80
•
Minimum
Value
4489.1
15.4
40.4
37.7
7361.3
104.2
140.1
135.5
16387.0
16.7
637.4
632.4
3104.0
619.0
20192.0
20183.7
Maximum
Value
26158.0
1039.1
40.4
37.7
233692.9
2863.4
140.1
135.6
331410.0
365.5
637.4
632.5
313980.0
4155.0
20192.0
20183.7
"' Standard deviations of the log-transformed data
121 Results for both regular and QC samples.
131 Geometric mean of the ratio of the 2-hour result
associated with this value.
(expressed in log measurement units).
to the result of the adjoining 1 -hour sample. No measurement units are
A-31
-------
Table WR-4d. Descriptive Statistics (for Each Unit) of Lead Concentration (//g/g)
Associated with Vacuum Samples from Stainless Steel Dustfall Collectors
at One and Two Hours Following Completion of Carpet Removal
Unit
ID
1-01
2-01
3-01
4-01
Data
Representation
One-hour
One-hour121
Two-hour
Paired
Differences
(2-hr minus
1-hr)
One-hour
One-hour121
Two-hour
Paired
Differences
(2-hr minus
1-hr)
One-hour
One-hour121
Two-hour
Paired
Differences
(2-hr minus
1-hr)
One-hour
One-hour'21
Two-hour
Paired
Differences
(2-hr minus
1-hr)
Distance
(feet)
0
6
6
6
0
6
6
6
0
6
6
6
0
6
6
6
N
3
4
1
1
3
4
1
1
3
4
1
2
3
4
1
1
Arithmetic
Mean
52798.0
88953.5
4168.2
3048.0
1967.8
956.8
763.1
-697.9
32397.0
10727.3
31398.0
19983.4
21935.3
18039.1
98210.1
71879.3
Geometric
Mean
46459.8
12547.8
4168.2
3.7207'3'
1413.2
600.1
763.1
0.5223'31
26017.3
9505.6
31398.0
2.750713'
21396.3
13938.6
98210.1
3.730013'
Log Std.
Dev.'11
0.63
2.43
•
1.05
1.28
•
0.81
0.57
0.27
0.87
•
Minimum
Value
24788.0
1120.3
4168.2
3048.0
489.6
114.4
763.1
-697.9
12730.7
5177.3
31398.0
19983.4
16192.9
5322.2
98210.1
71879.3
Maximum
Value
87222.4
335193.5
4168.2
3048.0
3956.8
1827.1
763.1
-697.9
62230.8
19074.7
31398.0
19983.4
28039.8
32116.5
98210.1
71879.3
111 Standard deviations of the log-transformed data
121 Results for both regular and QC samples.
131 Geometric mean of the ratio of the 2-hour result
associated with this value.
(expressed in log measurement units).
to the result of the adjoining 1-hour sample. No measurement units are
A-32
-------
Table WR-5a. Descriptive Statistics (Across All Units) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from Window
Wells Taken Prior to Window Replacement111
Data
Parameter
Sample Weights
(g)
Loadings
Oug/ft2)
Concentrations
Oug/g)
N
11
(12)
11
(12)
11
(12)
Arithmetic
Mean
4.906
(4.529)
185742.2
(170456.3)
28476.47
(26293.65)
Geometric
Mean
3.029
(2.552)
134531.0
(95884.2)
23482.54
(19336.86)
Log Std.
Dev.121
1.258
(1.338)
0.94
(1.48)
0.653
(0.917)
Minimum
Vaiua
0.295
(0.295)
26786.0
(2311.7)
8458.91
(2282.63)
Maximum
Value
13.576
(13.576)
415342.3
(415342.3)
70067.02
(70067.02)
111 Results in parentheses summarize values with outlier included.
121 Standard deviation of the log-transformed data (expressed in log measurement units).
Table WR-5b. Descriptive Statistics (Within Each Unit) of Sample Weights, Lead Loadings,
and Lead Concentrations Associated with Vacuum Samples from the
Window Wells Taken Prior to Window Replacement111
Unit
ID
N
Arithmetic
Mean
Geometric
Mean
Log Std.
Dev.'21
Minimum
Value
Maximum
Value
Sample Weights (g)
1-01
2-01
3-01
4-01
2
(3)
3
3
3
0.297
(0.327)
2.924
5.525
9.341
0.297
(0.324)
2.913
5.304
8.464
0.0081
(0.1532)
0.1071
0.3411
0.5718
0.295
(0.295)
2.613
4.295
4.483
0.298
(0.387)
3.237
7.861
13.577
Lead Loadings (//g/ft2)
1-01
2-01
3-01
4-01
2
(3)
3
3
3
44014.3
(30113.4)
224769.5
203828.0
223114.3
42687.2
(16150.0)
215579.7
178221.1
136222.9
0.35
(1.70)
0.36
0.65
1.44
33287.6
(2311.7)
145304.8
91663.5
26786.0
54741.0
(54741.0)
296279.6
336049.0
415342.3
Lead Concentration (jug/g)
1-01
2-01
3-01
4-01
2
(3)
3
3
3
56582.26
(38482.65)
20689.59
27467.15
18535.23
54952.43
(19031.35)
18248.23
22990.21
17510.97
0.34
(1.85)
0.67
0.70
0.43
43098.31
(2282.63)
8458.91
15303.16
10623.19
70067.02
(70067.02)
27171.01
51756.13
23130.62
111 Results in parentheses summarize values calculated with outlier included.
121 Standard deviation of the log-transformed data (expressed in log measurement units).
A-33
-------
Table WR-6a. Descriptive Statistics (Across All Units) of Sample Weight, Lead Loadings,
and Lead Concentrations Associated with Interior and Exterior Paint Chip
Samples Collected Following Completion of Window Replacement
Location
N
Arithmetic
Mean
Geometric
Mean
Sample Weights (
Exterior
Interior
13(3I
12
0.5828
0.6335
0.5621
0.6021
Log Std.
Dev.(tl
Minimum
Value
Maximum
Value
a)'21
0.2592
0.3168
0.4951
0.4901
1.0110
1.0327
Loadings (mg/cm2)
Exterior
Interior
13(3>
12
20.566
6.825
11.991
2.190
1.410
2.307
0.339
0.037
49.651
23.270
Concentrations (mg/g)
Exterior
Interior
13.3)
12
135.32
82.66
105.09
33.83
0.958
1.969
7.71
1.21
227.52
190.33
111 Standard deviation of the log-transformed data (expressed in log measurement units).
121 Sample weights represent the weight of subsamples used in the chemical analysis. In several instances, two subsamples
from the same chip were analyzed. In such cases, the results from the two subsamples were combined to form a single
observation.
131 An extra paint chip sample was collected at unit 2-01 from a window where no other samples were collected.
A-34
-------
Table WR-6b. Descriptive Statistics (Within Each Unit) of Sample Weight, Lead Loadings,
and Lead Concentrations Associated with Interior and Exterior Paint Chip
Samples Collected Following Completion of Window Replacement
Unit
ID
Location
N
Arithmetic
Mean
Geometric
Mean
Sample Weights (g)(2)
1-01
2-01
3-01
4-01
Exterior
Interior
Exterior
Interior
Exterior
Interior
Exterior
Interior
3
3
401
3
3
3
3
3
0.6687
0.6765
0.6291
0.5089
0.5025
0.5052
0.5154
0.8434
0.6319
0.6396
0.5978
0.5089
0.5025
0.5052
0.5154
0.7989
Log Std.
Oev.'11
Minimum
Value
Maximum
Value
0.4002
0.3979
0.3504
0.0153
0.0062
0.0062
0.0193
0.4233
0.4951
0.5073
0.4964
0.5021
0.4998
0.5016
0.5040
0.4901
1 .0028
1.0127
1.0110
0.5174
0.5059
0.5072
0.5212
1.0327
Lead Loadings (mg/cm2)
1-01
2-01
3-01
4-01
Exterior
Interior
Exterior
Interior
Exterior
Interior
Exterior
Interior
3
3
401
3
3
3
3
3
9.63
5.897
36.948
0.057
24.911
8.368
5.314
12.976
7.26
4.75
34.33
0.055
23.991
8.339
2.436
10.566
1.029
0.901
0.472
0.336
0.345
0.103
1.815
0.815
2.284
1.6867
17.620
0.037
16.344
7.440
0.339
4.573
16.422
8.619
49.651
0.068
31.917
9.089
12.067
23.270
Lead Concentrations (mg/g)
1-01
2-01
3-01
4-01
Exterior
Interior
Exterior
Interior
Exterior
Interior
Exterior
Interior
3
3
4<3>
3
3
3
3
3
127.11
80.43
139.04
1.46
196.66
122.04
77.23
126.71
94.86
65.48
136.20
1.45
194.83
117.26
44.44
117.48
1.078
0.876
0.243
0.158
0.169
0.361
1.549
0.480
27.41
23.90
95.841
1.21
162.85
77.35
7.71
72.88
189.91
116.99
162.11
1.59
227.52
146.71
146.01
190.33
Standard deviation of the log-transformed data (expressed in log measurement units).
121 Sample weights represent the total weight of the subsamples taken from each paint chip collected. In several instances,
two subsamples from the same chip were analyzed. In such cases, the results from the two subsamples were combined
to form a single observation.
131 An extra paint chip sample was collected at unit 2-01 from a window where no other samples were collected.
A-35
-------
1000000
100000
g- 10000
&
f> 1000
s
| 100
10
1
#
•
•
•
•
Pre-ActMty Pre-ActMty Pre-ActMty Pre-ActMty
window well floor floor floor
D=0 D=3 D=6
Sample Type
D = distance from window (ft.)
Figure WR-1a. Boxplots of Lead Loadings (//g/ft2) for Vacuum Dust Samples From the
Floors and Window Wells Taken Prior to Window Replacement Activities
1000001
10000
1000
100
Pre-ActMty
window well
Pre-ActMty
floor
D=0
Pre-Activity
floor
D=3
Pre-ActMty
floor
D=6
Sample Type
D = distance from window (ft.)
Figure WR-1b. Boxplots of Lead Concentrations (//g/g) for Vacuum Dust Samples From
Floors and Window Wells Taken Prior to Window Replacement Activities
A-36
-------
1000000
100000
10000
1000
100
io ^
Post-Activity Post-Activity Post-Activity Post-Activity Post-Activity
floor floor floor steel steel
D-0 D-3 D-6 D-0 D-6
Sample Type
D = distance from window (ft.)
Figure WR-2a. Boxplots of Lead Loadings (//g/ft2) for Vacuum Dust Samples From Floors
and Stainless Steel Dustfall Collectors Taken One-Hour After Completion of
the Window Replacement Activities
•3
5
1000000
100000
10000
1000
100
10
Post-Activity Post-Activity Post-Activity Post-Activity Post-Activity
floor floor floor steel steel
D-0 D-3 D-6 D-0 D-6
Sample Type
D = distance from window (ft.)
Figure WR-2b. Boxplots of Lead Concentrations (//g/g) for Vacuum Dust Samples From
Floors and Stainless Steel Dustfall Collectors Taken One-Hour After
Completion of the Window Replacement Activities
A-37
-------
100.00 i
^ 10.00
1.00
0.10
0.01^
Exterior
paint chip
Interior
paint chip
Sample Type
Figure WR-3.
Boxplots of Lead Loadings (mg/cm2) for Paint Chip Samples Taken from the
Interior and Exterior of the Window Frames After Completion of the Window
Replacement Activities
A.3 DESCRIPTIVE STATISTICS OF DATA FROM THE CEP PHASE
Table CED-1 presents descriptive statistics on lead levels from personal air monitoring of
R&R workers in the CED phase of the EFSS. Lead levels were expressed in terms of task-length
averages (TLAs, or average exposure over the duration of activity, in ng/m3 of air) for a given
target activity and substrate on which the activity was performed. For each activity/substrate
combination, Table CED-1 presents the number of monitoring results, the arithmetic and
geometric means, the standard deviation of log-transformed data, and the minimum and
maximum observed values. Because these data tended to be lognormally distributed, the
geometric means are better indicators of central tendency within the data distributions than are
the arithmetic means.
One objective of the CED phase was to study the relationship between lead disturbance
generated by performing a target R&R activity and the distance from the activity at which lead
loadings are being measured from settled dust samples. This relationship helps explain potential
lead exposures associated with a given activity and was quantified through statistical modeling
procedures. Exploratory analysis demonstrated that the relationship between settled dust lead
loading and distance from activity was well approximated by the following linear model:
log(loading) = P0 + P^Distance) + Error
(A-l)
A-38
-------
where P0 and p, are parameters which quantify the linear relationship. This model was fitted
separately to data for each experimental unit, or for each combination of individual target activity
applications within a study unit.
From the fitted regression lines determined through Model (A-l), estimates are obtained
for three indicators of lead disturbance (these indicators are discussed further in Section C.5 of
Appendix C):
• lead disturbance within a 6'xl' gradient from the target activity;
• estimated lead loading in dust that settles from zero to one foot from the target
activity; and
• estimated lead loading in dust that settles from five to six feet from the target
activity.
For each model fitting, Table CED-2 presents these estimates, in addition to the estimates of P0
and P, from the model. One row of Table CED-2 exists for each experimental unit.
A series of figures follow Table CED-2 which present predicted lead disturbance (in log
units) as a function of distance from activity, as determined through statistical modeling results.
Each figure contains a number of plots representing a particular target activity and substrate:
Figure CED-1: Drilling into wood;
Figure CED-2: Drilling into plaster;
Figure CED-3: Sawing into wood;
Figure CED-4: Sawing into plaster;
Figure CED-5: Wood door modification;
Figure CED-6: Sanding of painted wood;
Figure CED-7: HVAC ductwork removal;
Figure CED-8: Demolition.
The first plot in each figure contains prediction curves for each fit of Model (A-l) within this
activity/substrate combination (as given by the dashed lines). Separate fits of the model were
made for each occurrence of the activity/substrate in the study (i.e., each experimental unit), as
indicated in Table CED-2. Also in this plot are prediction curves of single fits of the "population
model" (solid line) and the "two-stage" model (thick-dashed line). Both models, explained in
Section C.4 of Appendix C, predict lead disturbance for the activity/substrate group across the
entire CED phase. As both models give nearly equivalent predictions, the solid and thick-dashed
lines are nearly plotted on top of each other.
Subsequent plots in Figures CED-1 to CED-8 illustrate the result of fitting model (A-l) to
each experimental unit in the study by plotting the observed settled dust lead loadings (in the log
domain) versus distance from activity, as well as the fitted prediction curve. These plots are
identified by the building and room in which the activity was monitored.
A-39
-------
Table CED-1. Summary Statistics for Task-Length Average Personal Worker Lead Levels
(/yg/m3) During Each Combination of CED Activity and Substrate
Task
Clean/Plaster
Clean/Wood
Demolition
Door Modification/
Wood
Drill/Plaster
Drill/Wood
HVAC Removal
Component
Removal/Wood
Abrasive Sanding
Saw/Plaster
Saw/Wood
#Data
Points
4
2
20
6
6
7
4
2
9
2
6
Arithmetic
Mean
27.701
103.44
152.63
819.93
6.9075
26.324
50.075
344.01
544.82
145.51
581.85
Geometric
Mean
24.461
102.36
106.81
486.45
6.2510
15.147
49.623
343.84
332.75
109.99
545.84
Log Std Dev
0.5503
0.2050
0.7406
1.1722
0.5181
1.2947
0.1571
0.0451
1.0110
1.1080
0.3801
Minimum
Value
14.592
88.549
33.553
112.03
2.6168
3.3666
40.381
333.05
73.645
50.245
397.48
Maximum
Value
53.283
118.33
947.06
2280.4
11.598
50.235
58.310
354.97
2311.7
240.77
967.99
A-40
-------
O
c
LL
(0
in
to
u>
•o
n
a>
Q
O)
a>
T~
<
"35 o
5! u
^ Q)
E.2
•o o
<
-5
.E n
a>
0) U
EC
M «°
Cjl
6
UJ
O
(0
T
4
•
1
T
L
1
L
i
r>
«
s
x
' S e
S O O "5*^
' ^** O w **
ta r«* £M O *^|
3 ^ (A "I S
5 in g ^
„
i^olf
= V «/> -* js£
U° £
«
V Js C *?*
- X * "S^L
5 ;v -J S 3
2 £°
!•»
flj UJ O
w
-
f
1
0
00
CO
o
«-
^
«^
00
Tfr
0
CO
in
CN
0
00
5
in
in
p»
S
CO
00
p-
CN
^
^.
00
tt
1
in
in
O5
t-
CN
"-
CO
CO
o
CO
CO
r-
CO
8
tv
CO
CO
CO
CO
8
o
*—
<*
00
CM
CO
00
r—
*~
in
in
CN
p~
1
O)
CO
CO
CO
«-
CO
CO
en
cn
CO
0
00
p-
CN
CN
CO
P»
P~
co
00
in
cn
in
CO
cn
o
CO
00
o
1
CO
r-
00
CO
CN
•a
o
0
^
CN
in
rr>
CO
CO
O
in
CO
3
CO
in
CM
p~
in
03
»t
•*
00
CN
00
CO
*"
o
CO
co
1
in
•*
CN
CN
^
CO
1^
CN
CO
CO
O)
in
CO
CO
CN
rv
•*
CO
CO
CN
•
p~
CO
'
in
CM
CO
CO
o
00
in
CO
o
P-.
CO
CO
CO
0
CO
CO
o
y^
m
CO
in
CN
in
00
Px
CO
*"•
^
in
m
CN
1
§
CO
CM
CO
00
in
,_
CO
CN
00
O
*~
in
CO
in
*~
O
CO
CO
p-
o
o
co
1
CO
in
en
<-
CN
cn
m
CO
cn
CO
CO
00
cn
CN
00
00
<*
CO
CM
0
cn
en
00
*~
p-
o
00
CO
1
o
in
O
CM
•-
CO
o
CO
«*
CN
cn
in
CO
j^
m
r»
cn
in
in
P-.
00
^-
CO
o
in
r-
CO
CO
1
CN
CN
•-
CN
(A
ro
61
CN
r—
O
cn
o
CN
CO
CO
CO
in
O
«—
CO
CO
CO
CO
o
in
CN
00
p-
o
cn
o
CN
00
§
CO
CN
CN
CO
cn
m
in
en
00
CM
00
in
0
00
r-
*
in
cn
in
CO
O
on
p-
co
en
in
CO
00
O
CN
in
in
1
P-
cn
CM
CO
«-
00
in
^.
CN
00
00
cn
00
o
in
o
CO
S
cn
CO
CO
CO
00
in
CO
S
p-
CN
*~
in
CN
CO
o
1
cn
cn
CN
CM
«-
CO
o
CN
CO
^.
00
CO
CO
00
CO
00
CO
CM
s
Pv
CM
•*
CO
CO
•*
CO
CM
00
CO
O
in
p-
p-
00
1
O
in
CM
CN
CO
CO
^.
o
CO
T—
S
0
CO
cn
in
cn
cn
CN
cn
CO
in
CM
CO
in
CO
^
in
f~
O
00
O
CO
1
o
CM
CM
CN
CN
•o
0
O
^
5
a
C/J
p-
(T)
in
CO
in
CM
CO
o
in
CO
CO
CO
00
O
CO
CN
in
in
CO
O
in
,—
00
CO
1
0
00
CO
CO
m
i^
CM
O
cn
00
00
in
CN
CO
CO
CO
O)
px
CN
CO
in
CO
o
px
*"
If)
ps.
T—
1
§
in
o
CN
co
m
CO
i^
00
CN
00
S
in
CN
cn
0
00
CN
00
CN
CN
00
CO
o
*~
T—
0
o
00
1
in
CO
o
o
T—
CO
cn
in
px
00
in
cn
o
cn
cn
CO
in
CO
o
CO
CO
in
CN
CM
CM
o
CN
CN
CN
00
o
1
CO
CO
in
CN
CN
to
51
A-41
-------
•o
0)
o
o
UJ
O
.
•g
c
«
•c
S
1
SS ?_
C """""' C/l O "Si
3 ^ (A ^ S
fl£ £
*> S c1
Hlo:s£
g U. Q P
£ w~ j^jfr ° pi
tS *A rt ^
u — • §
•0 S* *,
S - ™ c (*(""
w ^ 5 .* {•
fX O » "•£
— J 5 ??
(0 ,2 j
-
|g j
0 IU D
(A
•D
fl
s w
UJ
SS
*J 4)
C °
il
,2 S
Eil1-
o £ w '2
DC « O -*
> X
O)
•g —
g3
X
a
2
1
w
S>"?
31
CO
CO
CM
O
CO
CO
00
CM
CO
cn
CO
00
o
1—
CO
CO
CM
in
r-
CO
O
CM
CO
t—
CM
CM
cn
CM
*~
rv
in
00
o
rv
rv
en
CM
00
in
CM
CO
en
CO
CM
O
t—
CO
0
CO
00
rv
«—
*~
co
in
00
CM
00
in
O
*~
CM
in
CM
00
CO
00
00
CD
03
CD
cn
in
CM
03
|v
^
o
CO
co
cn
CO
IV
^-
*~
in
CO
cn
CM
CO
rv
CM
O
CM
•<*
CD
in
rv
rv
^
iv
in
iv
CO
cn
i—
^
cn
CO
^
*j-
CD
CO
CO
cn
CM
rv
cn
o
CD
03
CO
'
in
CO
rv
in
CM
•a
o
o
^
o
o
0
t—
CO
03
00
^
CO
CM
O
CO
in
CO
CO
O)
CO
rv
r-
06
in
rv
00
00
CD
CO
CO
^
O
CO
CO
1
CO
00
CO
CO
CO
CO
CD
in
^
CO
cn
co
CO
CM
cn
CO
rv
^
CO
CO
CO
CO
CM
cn
CO
CO
|V
00
CO
o
|v
in
CO
CD
IV
1
^
CM
cn
CO
co
o
CO
CD
CO
in
in
o
CD
CO
CM
*t
CM
CO
^
CO
r—
CO
CO
o
CM
CD
in
*~
in
CO
8
'
CM
**
03
o
*~
o
o
in
CO
0
00
in
CD
CO
CO
cn
CM
^
•*
>*
CM
in
CM
r—
CO
iv
CM
r-
cn
CM
CM
rv
o
co
CO
CM
CO
1
0
CD
o
CO
CM
^
o
o
*
c
(D
w
CO
00
CM
00
rv
00
00
co
rv
«*
cn
o
cn
CO
CO
cn
CM
CO
in
CO
O
rv
o
CO
CO
CM
CO
1
cn
iv
O
CO
CD
CD
|V
CO
CM
00
*fr
rv
CM
CO
o
CM
CO
00
00
00
CD
in
m
*~
00
1
in
'
in
m
CM
O
•~
o
CM
CO
CO
^
O
rv
S
CM
CO
00
rv
in
^*
*—
03
iv
CM
IV
CM
O
in
CO
IV
CO
T—
'
in
CO
CD
CM
D
Q
O
I
00
in
O
rv
5
CM
O
CM
0
CO
00
CO
in
00
0
o
CO
CM
o
in
co
CO
o
0
in
^f
*~
cn
rv
0
CO
CO
CO
«—
00
CM
in
cn
-fr
00
O
CO
CO
CO
o
CM
CO
IV
CM
'
rv
IV
CO
CO
*~
CO
CO
CO
CO
03
CM
03
CM
CO
»t
O
rv
O
CO
^
CO
CO
CO
rv
CO
o
|
CM
'
in
CD
in
in
CM
CO
CM
cn
rv
CO
CO
00
CO
CD
cn
CO
rv
rv
00
0
CO
CO
o
rv
CO
i—
'
CM
03
03
CM
ai
(D
°-
Tfr
CM
CD
CO
|v
CM
00
CO
•
iv
CO
CM
CO
CM
CM
CM
CD
CO
CO
CO
^_
00
|V
CO
00
0
CM
O
*"
O
CD
^
l£>
*~
CM
CM
CO
CO
CO
CO
CO
CO
•*
CM
CO
00
rv
CM
CO
CM
^T
CM
00
CD
CM
CO
CO
in
CO
o
CO
o
'
CO
CM
CO
in
CM
«
(U
a.
•5
0
o
*•
A-42
-------
^ m
O
I
II
I
I
CO
Q
•
I
L CO
1 CM
"_ O
- O)
9JBnbs/6rf)6o| ISHQ
° I
r» 3
^ 2
- co
8
10 g
CO
CM
I- o
Q
111
o
0)
3
D>
iZ
A-43
-------
co
s
A
s
!
(joo»
(B
S
9
V
•!
•!
s
-^»
!
CM
S
'^
!
•I
«nnb«^)n)0oi IBRQ
•I
o
o
Q
Lit
u
§
O)
e«nb6(/Cn)Oon9na
vng
A-44
-------
CM
I
'^
I
CM
9
9
(|oo» UBntyOnXSoi tana
5
!
(toot ajanbtyOnXSai JWQ
9
V
'I
O
O
a
in
o
a>
3
o>
A-45
-------
8
Q_
II
i
£
o
II
_ CO
- CM
CO
CM
h O
in
o
|
o>
ftooj 9JBnbs/6r/)Bo|
A-46
-------
s
a«nb«/0n)0q pna
CM
I
"^
!
CM
\l
(pq 8JBnbe/On)Oo( isna popes
•^
I
-I
s —
TJ
Q)
C
o
o
Q
UJ
O
a>
3
O)
(>oo»
A-47
-------
T- W
CO
CM
h
-------
CM
I
•!
ftooj uantyOnXtoi JSPQ
-^
!
i
^^
!
CM
•!
•^
: !
9
9
*
*
P
•!
1
I
o
a
CO
a
Ul
o
a>
o>
P
-------
CM
I
*
9
9
'I
•JVt»j/Bn)60|
9
V
!
O
o
CO
Q
111
o
£
O)
IT
(jooi
PW9S
A-50
-------
00
h*
CO
IO
^r
CO
CM
J- O
9JBnbs/6ry)6o| JSIIQ
Q
UJ
O
£
O)
A-51
-------
^^
!
IB
*
9
•!
(>oo» ««nbei/0^eo| tana pen>9S
•^
!
9
*
•I
+3
O
O
i
Q
Ul
O
3
O)
(topi euenbei/O^eq
A-52
-------
T- O
HI
O
OJ
ftooj aienbs/6r/)6o| ISFIQ
A-53
-------
IB
*
9
V
f
9
"I
£
8
aienbs^8n)(5o| jsna
«JBnb8i/Bn)Oo|
•I
(pen ejenb8/Cn)8onana
ejBnbe/On)Bon8na
Q)
C
o
O
in
a
UJ
O
£
A-54
-------
CM
!
•I
I
•o
V
O
O
in
6
at
O
i
D)
A-55
-------
s
J- o
o
HI
O
£
3
ftooj 9jenbs/6r/)6o| isna
A-56
-------
'^
!
Qoat Maflbeyenytai pna
-^
I
•• i ••
: i
'^
!
9
tt
0
"^
o
O
CO
Q
01
O
-------
Q
II
I
S
i
in
•- £
. £
" $2
CM
i o
- o)
r °°
r |s"
- CO
"^
: *"
: °°
- CM
J- o
^- o
ftooj. 9jenbs/6r/)6o|
Q
01
o
a>
•3
D)
il
A-58
-------
1
I
^^
!
•I
ajenbq/6n)eo| jsna
•I
0)
u
O
o
a
UJ
o
O)
jsna
A-59
-------
Q_
II
1
•>- m
00
CM
- co
CO
CM
J- o
op
Q
111
O
o>
3
CD
ftooj 9JBnbs/6T/)6o| isna
A-60
-------
ftoq «JBnb8,Cn)Oonsna
-^
!
IP
*
P
t*
ftoo» flJBnbej/CnXJq isna
•I
(joq euenbe^flnXSoi tsnQ
•o
0)
o
o
op
6
UJ
o
0)
3
O)
A-61
-------
co
1
p
V
•!
ftooj uanbty6n)Bo|
i
CM
*
P
P
'I
11 u
Is p jt p p F
+3
o
U
00
Q
Ul
O
0)
O)
ftooj
A-62
-------
*
*
^ '
fP3H.»^«i
ftooiwrt
«j8nbe/On)Co(
*
*
9
V
*
9
T3
a>
O
o
00
a
UJ
O
0)
3
O)
t»*»!
QOOt
JWQ
A-63
-------
A.4 DESCRIPTIVE STATISTICS OF DATA FROM THE CLEANUP INVESTIGATION
Section 7D presents design and results of an investigation of the effects of cleanup
procedures on lead levels in settled dust that remain for occupants to encounter. For each
combination of R&R activity and cleanup method, the following tables describe the collected
data:
Table CI-1: Lead loadings in paint chip samples taken from surfaces disturbed by R&R
activity
Table CI-2: Lead loadings in post-activity settled dust samples taken prior to cleanup
Table CI-3: Lead loadings in post-cleanup settled dust samples
Table CI-4: Lead loadings in next-day SSDC vacuum-dust samples.
Table CI-1. Descriptive Statistics for Measured Lead Loadings (mg/cm2) in Paint Chip
Samples Taken from Surfaces Disturbed by R&R Activity in the Cleanup
Investigation
R&R
Activity
Drilling
Drilling
Abrasive
Sanding
Abrasive
Sanding
Cleanup
Method
Broom
Vacuum
Broom
Vacuum
Nm
3
3
3
3
Arithmetic
Mean
1.77
4.72
7.19
14.27
Geometric
Mean
0.32
1.31
1.91
14.17
Log
Std. Dev,
2.42
2.86
2.97
0.14
Minimum
0.08
0.05
0.06
12.17
Maximum
5.15
9.17
12.60
15.84
One result exists for each experimental unit (i.e., each time the activity/cleanup method combination was performed).
A-64
-------
Table CI-2. Descriptive Statistics for Measured Lead Loadings (//g/ft2) in Post-Activity
Settled Dust Samples Taken Prior to Cleanup in the Cleanup Investigation
R&R
Activity
Drilling
Drilling
Abrasive
Sanding
Abrasive
Sanding
Cleanup
Method
Broom
Vacuum
Broom
Vacuum
Distance
from
Activity
(ft.)
0
3
6
0
3
6
0
3
6
0
3
6
Nd»
3
3
3
3
3
3
3
3
3
3
3
3
Arithmetic
Mean
141773
425
203
253005
471
287
3021341
3177
17405
371105
8443
796
Geometric
Mean
26739
176
65
73508
312
146
653201
715
1376
202540
5277
491
Log
std.
Dev.
2.38
1.93
2.60
2.81
1.08
1.83
3.34
2.26
3.71
1.60
1.37
1.43
Mm.
5332
22.3
3.3
2869
144
17.7
13831
121
30.6
34421
1150
96.9
Max.
411270
1013
392
444986
1074
463
5096250
9075
50500
762190
16374
1452
111 One result exists for each experimental unit (i.e., each time the activity/cleanup method
combination was performed).
A-65
-------
Table CI-3. Descriptive Statistics for Measured Lead Loadings (//g/ft2) in Post-Cleanup
Settled Dust Samples Taken in the Cleanup Investigation
R&R
Activity
Cleanup
Method
Distance
from
Activity
(ft.)
N'11
Arithmetic
Mean
Geometric
Mean
Log
Std.
Dev.
Min.
Max.
LEAD LOADINGS IN DUST-VACUUM SAMPLES
Drilling
Drilling
Abrasive
Sanding
Abrasive
Sanding
Broom
Vacuum
Broom
Vacuum
0
3
6
0
3
6
0
3
6
0
3
6
3
6
3
3
6
3
3
6
3
3
6
3
93
66
22
217
89
40
3941
2936
1794
3731
114
306
42
24
12
52
40
27
235
232
336
388
62
161
1.79
1.72
1.40
2.52
1.52
1.13
3.53
2.29
2.38
2.92
1.32
1.36
6.2
3.3
3.3
3.8
4.5
9.0
12
33
82
53
7.6
65
219
256
53
582
318
85
11720
16918
5212
11041
373
769
LEAD LOADINGS IN DUST-WIPE SAMPLES
Drilling
Drilling
Abrasive
Sanding
Abrasive
Sanding
Broom
Vacuum
Broom
Vacuum
0
6
0
6
0
6
0
6
3
3
3
3
3
3
3
3
266
202
805
166
2995
2830
905
432
166
124
360
123
1074
829
808
303
1.27
1.33
1.84
1.03
1.80
2.04
0.60
1.07
46
30
50
41
255
149
425
108
579
427
1857
312
8131
7854
1402
904
111 At zero and six feet from activity, one result exists for each experimental unit (i.e., each time
the activity/cleanup method combination was performed). At three feet from activity, two
dust-vacuum results and one dust-wipe result exist for each experimental unit.
A-66
-------
Table CI-4. Descriptive Statistics for Measured Lead Loadings (//g/ft2) in Next-Day SSDC
Vacuum-Dust Samples Taken in the Cleanup Investigation
R&R
Activity
Drilling
Drilling
Abrasive
Sanding
Abrasive
Sanding
Cleanup
Method
Broom
Vacuum
Broom
Vacuum
Distance
from
Activity
IfU
0
6
0
6
0
6
0
6
N(D
3
2
3
3
3
3
3
3
Arithmetic
Mean
92
48
119
42
79
261
28
13
Geometric
Mean
57
46
95
40
60
155
27
12
Log
Std.
Dev.
1.16
0.47
0.79
0.37
0.99
1.33
0.34
0.60
Min.
25
33
53
31
20
41
19
5.9
Max.
216
64
234
61
138
588
36
17
111 One result exists for each experimental unit (i.e., each time the activity/cleanup method
combination was performed). For drilling/broom cleanup, results for only two experimental
units were available.
A-67
-------
This page intentionally left blank.
-------
APPENDIX B
DATA PROCESSING AND OUTLIER DETECTION
B-1
-------
B.O DATA PROCESSING AND OUTLIER DETECTION
B.1 DATA PROCESSING
The following subsections discuss how the requirements for data processing were met in
the EFSS. Specifically, these discussions address data storage, data transfer, data tracking,
database verification, and necessary data manipulations prior to the statistical analysis.
B.1.1 Data Storage. Transfer, and Tracking
All requirements concerning the storage, tracking, and transfer of the study data were
followed in the EFSS as specified in the Quality Assurance Project Plan (QAPjP). Within
several overnight shipments, MRI provided Battelle with the analytical study data in hard-copy
listings and on diskette in Lotus spreadsheet format. These listings represented results for
specific analytical batches. Battelle has stored the original data listings and diskettes in a locked
file cabinet, along with the original completed field data logs for which Battelle is responsible.
Battelle has made backup diskettes of the data spreadsheet files; these backups are currently
stored in the office of the database manager. The study database created by Battelle from the
MRI data files is stored on the hard disk of the database manager, and relevant data for statistical
summary and analysis are located on a network hard disk accessible by Battelle statisticians.
Battelle was responsible for entering data from the field sample logs into the study
database. After verifying the accuracy of the data entry (Section B.I .2), Battelle merged these
data with the analytical data sent by MRI in creating the study database.
Data tracking has been done in accordance with the QAPjP. All dates corresponding to
significant events in the collection, laboratory analysis, and transfer of the data are recorded in
the Battelle study database. Tracking programs confirmed that all data corresponding to
collected field samples and laboratory samples were included in the study database.
B.1.2 Data Verification
Prior to entering data into the study database, Battelle verified the correctness and accuracy
of study data in the following ways:
• Battelle staff hand-entered data from the field sampling logs, using double data entry
techniques and hand-checking to ensure that accurate data were entered.
• When a batch of analytical data was received from MRI, Battelle compared the
submitted hard-copy data listings with listings made of the contents of the Lotus
spreadsheets. This verification procedure included spot-checking of calculations.
• Information used to identify the records in the electronic database were compared
with similar information from the sample field logs, to verify overall data
consistency.
B-2
-------
Battelle staff members reviewed any discrepancies found in the data verification procedure,
making any necessary corrections to the database as a result. In this review, Battelle notified
MRI of any data issues that needed to be brought to their attention.
B.1.3 Data Manipulation
The QAPjP for the EFSS notes that data values for variables to be included in the
statistical analysis may need to be transformed based on their underlying distributional forms.
Statistical analyses and/or data summary were performed on three data variables: physical
sample weight, lead concentration, and lead loadings. Battelle statisticians examined the
distribution of these variables and concluded that within each study phase, statistical analyses
were more appropriately applied to the log-transformed data values. The data transformation
issue is further discussed in Section 5.2.
B.2 OUTLIER DETECTION
Outliers are loosely defined as data values which do not coincide with preconceived
assumptions on the data distribution for the given parameter. The assumptions are usually
functions of the observed distribution of the parameter data, given the underlying distribution has
some known form (e.g., lognormality). When data values are identified as outliers, they are
generally reviewed for accuracy, and any erroneous data are corrected or are omitted from
statistical analyses. Numerous outliers can imply that the assumptions on the data distribution
are not appropriate, and special care should be taken in applying appropriate methodologies in
the statistical analyses.
In each phase of the R&R study, outlier analyses were conducted using three methods:
• logic checks,
• formal statistical tests, and
• graphical review.
The outlier analyses were applied to sample weight, lead concentration, and lead loading data.
The above methods flagged data values for individual samples, as well as pairs of samples (i.e.,
samples taken from adjoining areas and having results with some expected intrinsic relationship)
whose results were inconsistent in some way. The findings of the outlier analyses are presented
below for each study phase.
B.2.1 Logic Checks
As a first step in the outlier detection process, logic checks were performed to flag results
which ran contrary to intuition or some underlying criterion. One check flagged those individual
sample results having non-positive lead loadings. Within each phase, no results were flagged for
this reason. Another set of checks within the window replacement and carpet removal phases
flagged sample pairs in the following way:
B-3
-------
• One hour/two hour settled dust sample pairs: Flag those pairs where a stainless steel
dustfall collector (SSDC) sample collected two hours following the activity had a
lower lead loading than the adjoining SSDC sample collected one hour following the
activity.
• Pre-activity/post-activity floor dust sample pairs: Flag those pairs where a floor dust
sample collected post-activity had a lower lead loading than the adjoining floor dust
sample collected prior to start of the activity.
Note that only lead loadings were included in the above two types of logic checks. Lead
concentrations were not considered, as it is possible that R&R activity could result in either
higher or lower lead concentrations in settled dust over time, according to how dust is distributed
as a result of the activity.
Results
Table B-l contains a list of sample pairs which failed one of the two logic checks in the
carpet removal phase. In this phase, five of the 24 pre-/post-activity sample pairs resulted in a
higher sample lead loading for the pre-activity sample than for the post-activity sample. Two of
these five samples had the post-activity sample loading decrease by more than 50%. It is
suspected that these five results are due to normal spatial variability, as the field sample logs
revealed no unusual circumstances with sample collection for these pairs.
Seven of the 24 one-/two-hour post-activity sample pairs in the carpet removal phase
resulted in a higher sample lead loading for the one-hour sample than for the two-hour sample.
Of these seven samples, five had the one-hour sample location closer to the activity than the
adjoining two-hour sample location. This finding supports the notion that these violations to the
logic checks may be the result of spatial variability. The largest deviation between one- and two-
hour results was found with a sample pair from unit 2-01, where the one-hour result (216.0
ug/ft2) was nearly ten times larger than the two-hour result (23.0 ng/ft2). This deviation was
nearly five times greater than the next largest deviation and therefore was considered a candidate
for review.
Table B-l also contains a list of sample pairs which failed one of the two logic checks in
the window replacement phase. None of the four one-/two-hour post-activity SSDC sample pairs
resulted in a higher sample lead loading for the one-hour sample than for the two-hour sample.
However, 9 of the 34 pre-/post-activity sample pairs resulted in a higher sample lead loading for
the pre-activity sample than for the post-activity sample. Of these, the two samples with the
largest negative difference were reported to MRI for further investigation but no problem was
uncovered in the laboratory analysis. These sample pairs were both taken from unit 4-01. It is
suspected that all-other results are due to normal spatial variability, as the field sample logs
revealed no unusually circumstances with sample collection for these pairs.
B-4
-------
Table B-1. Sample Pairs in the Carpet Removal and Window Replacement Phases Whose
Lead Loadings Were Flagged by Logic Checks
1. Pre-Activity/Post-Activity Floor Dust Sample Pairs:
Pro-activity lead loading larger than post-activity lead loading
Unit
ID
Location
Distance
from
Activity
Pre-Activity Sample
Sample
ID
Lead
Loading
toflffi2)
Post-Activity Sample
Sample
ID
Lead
Loading
(PBffi2)
(Post-activity
result) - (pro-
activity
result}
Ratio of
Pro-activity
to post-
activity
results
Carpet Removal Phase
1-02
2-01
2-03
2-05
2-05
Hall (L2)
Hall (L2)
Kitchen (L1)
Foyer (L1)
Kitchen (L3)
60058
62091
60411
60701
60746
18.78
226.0
35.04
267.7
447.5
60056
62031
62221
60691
60796
16.69
54.87
30.92
125.8
252.4
-2.09
-171.1
-4.12
-141.9
-195.2
0.89
0.24
0.88
0.47
0.56
Window Replacement Phase
2-01
2-01
2-01
2-01
3-01
3-01
4-01
4-01
4-01
1 IBED2)
2 (BED2)
1 (HAL)
1 (HAL)
1
10
13(BED1)
13IBED1)
17(LVG1)
6
3
0
6
3
6
0
3
0
60482
60477
60487
60522
60302
60612
60494
60569
60559
51.46
157.95
28.39
50.60
11027.00
3703.20
439845.0
29402.00
145420.0
60332
60432
60362
60402
60577
60322
60504
60424
60439
50.97
107.62
19.0
30.42
10429.00
3162.90
85685.11
10342.00
38826.0
-0.49
-50.33
-9.39
-20.18
-598.00
-540.30
-354159.89
-19060.00
-106594.00
0.99
0.68
0.67
0.60
0.95
0.85
0.19
0.35
0.27
2. One-hour/Two-hour Post-Activity SSDC Sample Pairs:
One-hour lead loading larger than two-hour lead loading
Unit
ID
Location
1 -hr, Post-Activity Sample
Sample ID
Lead Loading
OfSfff)
2-hr. Post-Activity Sample
Sample ID
Lead
Loading
0*g/ftz)
(1-hr, result)
- (2-hr.
result)
Ratio of 1*
hr . to 2-hr,
results
Carpet Removal Phase
1-01
1-01
1-02
1-02
1-03
2-01
2-05
Bathroom (L2)
Kitchen (L3)
Hall (L3)
Hall(L1)
Kitchen (L3)
Hall (L2)
Foyer (L1)
60017
60010
60064
60061
60101
62006
60771
205.7
92.61
36.50
14.37
13.35
216.0
6.78
60006
60002
60070
60054
60099
62116
60751
163.2
59.94
22.39
12.89
10.61
23.00
5.95
-42.5
-32.7
-14.1
-1.48
-2.74
-193.0
-0.83
0.79
0.65
0.61
0.90
0.79
0.11
0.88
B-5
-------
B.2.2 Formal Statistical Tests
The primary statistical approach to outlier detection was the fitting of simple linear
models to estimate how individual data points deviate from the central tendency of the observed
data distribution. While the specific methods differed across the three study phases, this basic
approach was taken in each phase.
Methodology
In the carpet removal phase, the following analysis of variance model was fitted to the
log-transformed data Y, (i=l,...,n):
Y, = fJ + e,. , (B-1)
where jo. represents an overall mean value and e{ represents deviation from the mean. For each
model fit, the absolute value of each studentized residual was compared to the upper (1-
a/2n)*100 percentile of the Student-t distribution, where n is the number of residuals and a is the
overall significance level. Those observations whose studentized residuals exceeded this
percentile (in absolute value) were declared outliers at the a level. This test takes a Bonferroni
approach, implying that the overall error rate in falsely identifying an observation as an outlier is
no higher than a.
In the window replacement phase, the Grubbs outlier test was used. In the Grubbs test,
the most extreme of the log-transformed data (maximum or minimum) was subtracted from the
mean value, and the corresponding absolute difference was divided by the standard deviation of
the log-transformed data. If the absolute value of this statistic exceeded the critical value
tabulated in Grubbs (1950), the data point was declared an outlier. This procedure yields similar
results to the method used in the carpet removal phase.
In the CED phase, the Grubbs test was applied to the residuals in some modelled
relationship. The primary step in the outlier detection process consisted of flagging data points
that differ statistically from some underlying statistical relationship with a series of covariates.
Two models from Section C.4 of Appendix C were considered in this outlier test according to the
type of data considered: Model (CED-1) for personal air lead concentrations, and Model (CED-
3) for SSDC sample lead loadings. The minimum and maximum residuals (divided by their
standard error) from the model fit were compared against critical values tabulated by Grubbs
(1950). A data point was flagged as an outlier if its residual exceeded the critical values at either
the cc=0.05 or 0.10 level.
Each formal outlier test was run at the 0.05 and 0.10 significance levels. Significance at a
lower a level denotes a more severe outlier.
B-6
-------
Carpet Removal Phase Results
In the carpet removal phase, the study data were partitioned into several data
classifications, and Model (B-l) was fitted separately to data within each classification. These
classifications and their associated samples sizes are as follows:
Vacuum samples (for sample weight, loading, and concentration):
Regular pre-activity floor dust samples (n=40);
QC (side-by-side) pre-activity floor dust samples (n=8);
Pre-activity carpet dust samples (n=24);
Regular 1-hr, post-activity samples (n=64);
Regular 2-hr, post-activity samples (n=24);
QC (side-by-side) 1-hr, post-activity samples (n=16);
QC (side-by-side) 2-hr, post-activity samples (n=8).
Vacuum samples (for sample weight and concentration onM:
• Field blanks (n=8 for sample weight, n=7 for concentration).
Wipe samples (for sample loading):
• Regular 1-hr, post-activity samples (n=16);
• QC (side-by-side) 1-hr, post-activity samples (n=8).
Air samples (for sample loading"):
• Pre-activity ambient air samples (n=8);
• During-activity ambient air samples (n=16);
• During-activity personal air samples (n=14).
Table B-2 indicates that only one data point was flagged as an outlier at the cc=0.10 level:
the lead concentration in a 2-hour post-activity vacuum sample taken from a stainless steel
dustfall collector at unit 1-04. Its value of 29,867 (jg/g was over four times higher than the next
largest lead concentration among the 2-hour samples. In addition, four data points were declared
outliers at the cc=0.20 level:
• The vacuum field blank taken at unit 1-02 had a lead concentration of 173.2 ng/g,
compared to a range of 509-1376 ug/g among the other seven vacuum field blanks in
the study.
• The ambient air sample taken during the activity in unit 1-04 was 13.38 ug/m3, while
the highest reading among the other units was 5.36 ng/m3. The pre-activity (baseline)
ambient air sample result for this unit was not larger than that for the other units.
However, the two personal air sample results for this unit were the highest in the
study, both falling above the OSHA worker action level of 30 ug/m3.
B-7
-------
Table B-2. Data Values in the Carpet Removal Phase Identified as Outliers by Formal
Statistical Tests
Unit
ID
1-02
1-04
1-04
2-01
2-01
Instr.
Batch
E10193A
V11013A
E10213A
E10213A
E11303A
Sample
Prep
Batch
602
605
604
604
607
Sample
ID
60089
60106
60143
62011
60991
Sample
Medium
Vacuum
Ambient
air
Vacuum
Vacuum
Vacuum
Sample
Type
Field blank
During
activity
2-hr post-
activity
(SSDC)
1 -hr post-
activity
(SSDC)
QC side-by-
side samples
only)
1 -hr post-
activity
(window sill)
Location
...
Bedroom #3
Bathroom #1
(L1)
Hall (L3)
Bedroom #1
Parameter
Lead
cone.
Sample
loading
Lead
cone.
Sample
weight
Lead
cone.
Sample
value
1 73.2 vglg
13.38//g/m3
29867 fjg/g
0.0009 g
66776 fjglg
o
Level
0.20
0.20
0.10
0.20
0.20
Hi/Lo
Low
High
High
Low
High
• The sample weight for a one-hour post-activity QC side-by-side sample, taken from
a stainless steel dustfall collector in unit 2-01, where the weight of 0.0009 g fell
below the next lowest weight of 0.0058 g among these QC side-by-side samples.
• A lead concentration of 66,776 ug/g for a one-hour post-activity vacuum sample
from a window sill, compared to a range of 71.86-23,687 ug/g for the other post-
activity window sill dust samples.
The field sample logs indicated no special citations which would indicate why these samples had
unusually low or high results, and MRI reported no problems in the laboratory analysis after
further investigation. Thus, they were included in the statistical analysis.
Window Replacement Phase Results
Outliers detected in the formal statistical tests on window replacement data are presented
in Table B-3. As in the carpet removal phase, the study data were initially partitioned into
several data classifications. Grubbs test was run separately on data within each classification.
These classifications and their associated sample sizes were as follows:
B-8
-------
Table B-3. Data Values in the Window Replacement Phase Identified as Outliers by Formal
Statistical Tests
Unit ID
1-01
4-01
1-01
1-01
Instr.
Batch
E11153A
E02104A
E11173A
E11153A
Samp.
Prop.
Batch
614
620
616
614
Sample
ID
60230
60359
60257
60237
Sample
Medium
Vacuum
Paint chip
Paint chip
Vacuum
Sample
Type
Window Well
Exterior
Exterior
Pre-activity
floor at 3
feet
Parameter
Loading
Lead Cone.
Loading
Lead Cone.
Lead Cone.
Sample
Weight
Sample
Value
2311.65
2282.63
7.711
7.711
27.409
0.0055
a
Level
0.10
0.05
0.05
0.05
0.05
0.10
Hi/La
Low
Low
Low
Low
Vacuum floor dust samples (for sample weight, loading, and concentration'):
Regular pre-activity samples at 0 feet (n=12);
Regular pre-activity samples at 3 feet (n=12);
Regular pre-activity samples at 6 feet including QC (side-by-side) samples (n=16);
Regular 1-hr, post-activity samples at 0 feet (n=12);
Regular 1-hr, post-activity samples at 3 feet (n=12);
Regular 1-hr, post-activity samples at 6 feet including QC (side-by-side) samples
(n=16);
Vacuum Stainless Steel Dust Collector (SSDC) samples (for sample weight, loading, and
concentration):
• Regular 1-hr, post-activity samples at 0 feet (n=12);
• Regular 1-hr, post-activity samples at 6 feet including QC (side-by-side) samples
(n=16);
• Regular 2-hr, post-activity samples at 6 feet (n=4);
Vacuum Window Well dust samples (for sample weight, loading, and concentration):
• Pre-activity window well samples (n=12);
Vacuum samples ffor sample weight and concentration only):
• Field blanks (n=8 for sample weight, n=7 for concentration).
B-9
-------
Paint chip samples (for sample weight, loading, and concentrationV.
• Interior (n=l 2);
• Exterior (n=l 2);
Air samples (for sample loading):
• Pre-activity ambient air samples (n=4);
• During-activity ambient air samples (n=8);
• During-activity personal air samples (n=8).
Table B-3 presents those window replacement data that were identified as outliers by the
formal statistical test. Only three samples were flagged as outliers at the
-------
B.2.3 Graphical Review
The aim of graphical review of data for outlier detection was to visually observe those
data points which deviated from the general distribution of all data points. Two plotting
procedures were used to check for outliers in each of the three phases:
• the exploratory plotting and regression fitting procedure, and
• the single sample and simultaneous plotting procedure.
Exploratory plotting and regression fitting was applied to the carpet removal and window
replacement phases, while single sample and simultaneous plotting were used in the window
replacement and CED phases. A description of each of these procedures follows.
Exploratory Plotting and Regression Fitting: The exploratory plotting procedure
characterizes the overall relationship between the results of settled dust samples that are paired in
some way (e.g., wipe/vacuum sample pairs), flagging those samples whose results deviate
substantially from this observed relationship.
The relationship between paired sample results is observed by creating a scatterplot of the
log-transformed results of one sampling approach versus the other (e.g., wipe versus vacuum),
fitting a linear regression equation to the points in the plot, and flagging those points which
deviate substantially from the fitted line. Data points are flagged if the studentized residual
exceeds a value of two in absolute value. The results of one or both samples represented by a
given flagged point in the plot are then labeled as potential outliers.
Single Sample and Simultaneous Plotting: For single (i.e., unpaired) sample results,
descriptive plots, including lognormal probability plots of lead concentration, lead loading and
sample weight, and scatterplots were used to identify unusual observations.
Scatterplots were created of log-transformed loadings versus log-transformed weights or
concentrations grouped into appropriate categories (e.g., by distance from activity). Possible
confounding factors which could explain extreme results, such as dwelling unit, were then
examined. In cases where such a factor was deemed to explain deviations, the observation was
not included as an outlier. An example of a simultaneous scatterplot used to identify outliers in
the single sample case is presented in Figure B-l for paint chip sample results in the CED study.
Note the high variability among Baltimore(l) samples, which excludes the observation
associated with a low loading and low concentration from consideration. Dwelling unit in this
case is a confounding variable. In addition, dilution factors could explain some extreme values.
For example, an observation appearing as an outlier in a plot of lead concentration versus weight
may not appear as an outlier in a plot of lead loading versus weight.
B-11
-------
evP
I
B
nj
s
1
100.0000
10.0000
1.0000
0.1000
0.0100
0.0010
0.0001-
,*.:
0
o
o
o
o
0 ft
o
o
-6 -5 -4 -3-2-10 1 2
Paintchip Log Sample Concentration (log mg/g)
000 Baltimore Unit 1 « ° ° Baltimore Unit 2 * * * Denver Units
Figure B-1. Scatterplot of Paint Chip Log Lead Loadings versus Paint Chip Log
Lead Concentrations for the Demolition CED Activities
Only data associated with "regular" samples were included in the exploratory plotting
procedure for the 1-hour vs. 2-hour and vacuum vs. wipe comparisons (i.e., results for samples
with a field QC purpose, such as side-by-side samples, were not included). Some of the
scatterplots reviewed in this outlier analysis are found throughout this report. The regression
fittings were conducted separately within the same data categories that were considered in the
formal statistical tests procedure.
Carpet Removal Phase Results
Table B-5 presents those paired samples which have been flagged in the exploratory
plotting and regression fitting procedure for the carpet removal phase. The pre-activity/post-
activity pairs that were flagged had much larger post-activity results relative to their pre-activity
results. One pair occurred in unit 1-04, which had high lead levels in all samples. In the
wipe/vacuum pair that was flagged (unit 2-01, location L2) the vacuum sample loading was
higher than the wipe sample loading. Note that the loading and concentration associated with the
1-hr post-activity vacuum sample were also much higher than that for the 2-hr post-activity
sample at the same location. This unexpected result of observing higher loadings for a 1-hr
sample compared to a 2-hr sample was flagged in the logic check procedure above. The other
two l-hr/2-hr sample pairs that were flagged had much higher results for the 2-hour sample.
B-12
-------
Table B-5. Sample Pairs in the Carpet Removal Phase Identified as Outliers by Graphical
Review and Regression Analysis
1. Pre-Activitv/Post-Activitv Floor Dust Sample Pairs
Unit
ID
2-05
1-04
Location
Bathroom (L2)
Bathroom #1
(L1)
Parameter
Sample wgt.
Lead loading
Lead cone.
Pre-Activity Sample
Sample
ID
60666
60128
60128
Result
0.1629 g
2.61 //g/ft2
90.78//g/g
Post-Activity Sample
Sample
10
60676
60148
60148
Result
0.6285 g
6, 135 //g/ft2
20,662 //g/g
2. One-hour/Two-hour Post-Activity SSDC Sample Pairs
Unit
IP
2-03
2-01
2-02
Location
Bathroom (L2)
Hall (L2)
Sun room (L3)
Parameter
Sample wgt.
Lead loading
Lead cone.
Lead loading
Lead cone.
Pre-Activity Sample
Sample ID
62216
62006
62006
60486
60486
Result
0.01 58 g
21 6.0 //g/ft2
6990 //g/g
2.85 //g/ft2
370.2 //g/g
Post-Activity Sample
Sample
ID
60436
62116
62116
60466
60466
Result
0.2245 g
23.0 //g/ft2
746.7 //g/g
54.82 //g/ft2
3807 //g/g
3. Wipe/Vacuum SSDC Sample Pairs
Unit
L ID
2-01
Location
Hallway (L2)
Parameter
Lead loading
Wipe
Sample ID
62186
Sample
Result
23.62 //g/ft2
Vacuum Sample j
Sample ID
62006
Result I
21 6.0 //g/ft-2 I
B-13
-------
Window Replacement Phase Results
Table B-6 presents those results in the window replacement phase which have been
flagged in the single sample and simultaneous plotting procedure.
None of the paired samples for the 2-hour versus 1-hour comparisons were identified as
outliers. In addition, those paired samples for the pre-activity versus post-activity floor dust
comparisons identified as outliers did not differ from outliers detected in the logic check.
Therefore, paired samples are not included in Table B-6.
Table B-6. Sample Results in the Window Replacement Phase Identified as Outliers by
Graphical Review and Regression Analysis
Unit ID
1-01
1-01
1-01
4-01
4-01
4-01
4-01
Instr.
Batch
E11153A
E11153A
E11153A
E11173A
E02104A
E02104A
E02104A
E02104A
Sample
Prep.
Batch
614
614
614
616
620
620
620
620
Sample
ID
60230
60233
60184
60257
60344
60359
60339
60359
Sample
Medium
Window
Well
Window
Well
Paint chip
Paint chip
Sample
Type
Vacuum
Vacuum
Exterior
Exterior
Component
ID
2 (DEN)
2 (KIT)
2 (BAT)
2 (DEN)
17 (LVG1)
13 (BED1)
20 (LVG1)
13 (BED1)
Parameter
Loading
Cone.
Loading
Cone.
Weight
Loading
Cone.
Weight
Loading
Loading
Loading
Weight
Cone.
Weight
Sample
Value
2311.65
2282.63
33287.56
43098.31
.2950
54740.95
70067.02
.2984
2.28
3.54
0.34
0.5211
146.01
0.5212
Comment
Low value for
loading,
concentration.
Unusually high
concentrations
and loadings for
a low sample
weight.
Low values for
concentrations
and loadings.
High weights:
60339 has an
unusually low
concentration
compared to a
high weight.
CED Phase Results
Although several observations appeared at the extremes of these plots for SSDC samples,
distance from activity provided reasonable evidence for such deviations. For paint chip samples
several observations also appeared at the extremes of these plots. However, dwelling unit
provided reasonable evidence for such deviations. No observations for either the SSDC or paint
chip samples deviated substantially from the linear relation expected between log transformed
concentrations and loadings.
Paired samples included the nine vacuum/vacuum side-by-side pairs and the twelve
vacuum/wipe side-by-side pairs. No outliers were detected when considering the relationship
between paired samples.
B-14
-------
B.2.4 Outlier Summary
Outliers were considered serious contenders for rejection from statistical analyses when
they either failed two or more of the outlier checking procedures (logic checks, graphical review
or formal statistical tests) or Battelle's field log sheets revealed some sampling protocol
deviation. Only one outlier in the three phases of the R&R study was deleted from the statistical
analyses. A breakdown of outlier analysis results within each phase is as follows:
Carpet Removal
10 sample pairs and 5 additional single samples were flagged by at least one of the above
three analysis methods. Examination of the sample log sheets revealed no immediate
explanation for any of the results, and no problems were uncovered with these results as a
result of laboratory analysis. Therefore, none of these flagged results were removed from the
statistical analysis.
Window Replacement
9 sample pairs and 8 additional single samples were flagged by at least one of the above three
analysis methods. Examination of the sample log sheets and review of laboratory analysis
documentation revealed no immediate explanation for any of the extreme results. However,
one of these flagged results has been removed from the statistical analysis. The window well
sample (MRI ID 60230) was flagged in both the graphical review and formal statistical test.
In addition, this result appeared inconsistent with other measurements within the same unit
and the same window. Thus, this sample was removed from the analyses.
CEP
Three samples were identified as possible outliers by formal statistical tests. No outliers were
flagged by either graphical review or logic checks. Examination of the sample log sheets
revealed no immediate explanation for these three results, and no results were excluded from
statistical analysis.
B-15
-------
This page intentionally left blank.
-------
APPENDIX C
STA TISTICAL METHODS AND MODELS IN THE EFSS
C-1
-------
C.O STATISTICAL METHODS AND MODELS IN THE EFSS
In this appendix, the approaches to statistical analysis of the EFSS data are presented in
detail. Section C.I presents statistical models used in the carpet removal and window
replacement phases to characterize the statistical relationship between lead exposure and
predictor variables. The approach to estimating components of variation on the lead exposure
data within these two phases is presented in Section C.2. Section C.3 contains the method used
in the window replacement phase to estimate the effect of distance from activity on settled dust
sample lead loadings. Section C.4 presents statistical methods used to characterize lead
disturbance associated with various R&R activities in the CED phase. Section C.5 presents the
approach to estimating lead disturbance within a 6* x 1' dustfall gradient in an effort to compare
potential lead hazards in settled dust across CED activities. Section C.6 presents the
methodology used to calculate 95% confidence intervals for the 50th, 75th, and 95th percentile of
worker personal exposure results and for the percentage of results above the OSHA permissible
exposure limit of 50 ug/m3. Finally, Section C.7 presents the meta-analysis methodology for
combining the surface preparation results with results obtained from sources independent from
the R&R study.
C.1 MODELS IN THE CARPET REMOVAL AND WINDOW REPLACEMENT
PHASES TO EVALUATE FACTORS OR MEASUREMENTS IN
RELATION TO LEAD EXPOSURE
Methods
In the carpet removal and window replacement phases, a common statistical modeling
approach was used to express the distribution of lead in dust and air as a function of pre-activity
lead levels and/or other activity characteristics. Lead measurements were initially classified as to
the type of exposure represented and the approach to collecting the sample (personal worker
exposures, ambient air exposures, loadings from SSDC dust samples, etc.). A set of covariates
were identified as potential predictors of the lead measurements within each classification. A
statistical model evaluated the significance of the association between lead measurements and the
covariates. A reduced form of the model was used to characterize variability (Section C.2).
For the carpet removal data, the common model form across data categories was
generally loglinear, characterizing a linear relationship between the natural logarithm lead
exposure measurement and a series of p (^ 1) log-transformed covariates and duration of activity.
The model form was
+ £ P
log{Y..) = \og(fj) + £ Pklog(Xkij) + YT, + H. + Lj(i) , (C-1)
k = 1
where Yy is the lead exposure measurement for the sample taken at the jth location within the ith
study unit, |j, represents the (unknown) geometric mean lead measurement across the study phase,
XklJ represents the value associated with the j* location within the ith study unit of covariate Xk, Pk
is the (unknown) multiplicative effect of covariate Xk on the lead loading, T, is the duration of
C-2
-------
carpet removal (minutes) in the ith unit, y is the (unknown) additive effect of activity duration on
the log lead measurement, H; represents the random effect of the ith study unit (normally
distributed with mean zero and variance ou2), and Lj(i) represents the random effect of the jth
sample location within the ith study unit (normally distributed with mean zero and variance oL2).
Only data corresponding to "regular" (i.e., not side-by-side) samples were included in these
model fits. The MIXED procedure in the SAS® System was used to fit model (C-l) to each
category of lead loading data, resulting in estimates for the model parameters u, P,,...,PP, y, ou2,
and oL2 for a given category.
For the carpet removal phase, Table C-l lists the data categories and the covariate group
considered within each category. Each data category considered a specific group of covariates Xk
in model (C-l). According to the model, each covariate Xk has a multiplicative effect on the
(untransformed) lead loading, with the parameter Pk representing the extent of the effect
associated with covariate Xk. Thus, a test of whether pk is significantly different from zero
indicates whether the covariate Xk is significantly associated with the lead exposure measurement
in the given data category. Similarly, a test of whether y is significantly different from zero
indicates whether the duration of carpet removal activity is significantly associated with lead
exposure.
In the window replacement phase, a model similar to Model (C-l) was fitted to lead
measurements. Due to the small numbers of data points within each category of lead
measurement considered, each fit of the model included only one covariate. Therefore, the
model took the form
logfY,,) = logO/) + Pklog(Xkij) + H. + Lj(i) , {C-1a)
where the notation is interpreted in the same manner as for Model (C-l). The only exception was
for ambient air lead concentrations, where the pre-activity ambient air lead concentration for the
study unit was included as a covariate within each fit of model (WR-2). Multiple covariates were
considered within each data category; the model was fitted a number of times equal to the
number of covariates considered for the data category. Table C-2 lists the data categories and the
covariates considered in the various model fits within the category. The MIXED procedure in
the SAS® System was used to fit each model.
Results
Results of fitting the various models in Table C-l on carpet removal data (models
(CPT-1) to (CPT-4)) are presented in Table C-3. This table presents estimates and associated
standard errors for model parameters u, Bi,...,Bp, and y. The estimated parameters Pk and y were
generally not significantly different from zero at the 0.05 level, implying that their associated
covariates were not statistically associated with the lead loading of the sample type of interest.
However, the models were fit to a small number of data points that generally had high variability,
resulting in statistical tests with low power to detect differences from zero.
c-3
-------
Results of fitting the models in Table C-2 on window removal data (models (WR-1) to
(WR-5)) are presented in Tables C-4a through C-4d. These table presents model parameter
estimates and their standard errors. Table C-4a presents results from fitting model (WR-1) with
different covariates to personal exposure data. Table C-4b contains results from fitting model
(WR-2) with different covariates to area air data. Results from fitting models (WR-3) and
(WR-4) with different covariates to SSDC settled dust lead loading data are found in Table C-4c.
Finally, Table C-4d presents results from fitting model (WR-5) with different covariates to lead
amounts disturbed in a 6'xl' dustfall gradient. The estimated parameters Pk were generally not
significantly different from zero at the 0.05 level.
Table C-1. Data Categories and Covariates Considered in Fitting Model Form (C-1) to
Lead Loading Data in the Carpet Removal Phase
Model ID
Number
Data Category
Covariates Included in the Model
(CPT-1)
Personal Worker
Exposures (//g/m3)
X,,:
X2l:
Geometric mean lead loading in carpet in the ith
unit.
Geometric mean lead loading in pre-activity floor
dust samples in the ith unit.
Geometric mean lead loading in pre-activity window
sill dust samples in the ith unit.
(CPT-2)
Area Airborne Exposures
(//g/m3)
X,,:
X2l:
X3l:
Geometric mean lead loading in carpet in the ith
unit.
Geometric mean lead loading in pre-activity floor
dust samples in the ith unit.
Geometric mean lead loading in pre-activity window
sill dust samples in the ith unit.
Lead loading in the pre-activity area air sample in
the ith unit.
(CPT-3)
1-hour post-activity lead
loadings from floors or
SSDCs (//g/ft2)
X,,:
X2ij:
Geometric mean lead loading in carpet in the ith
unit.
Lead loading in the adjacent pre-activity floor dust
sample (jth location within ith unit).
Geometric mean lead loading in pre-activity window
sill dust samples in the ith unit.
(CPT-4)
1-hour post-activity lead
loadings from window
sills U/g/ft2)
X,,: Geometric mean lead loading in carpet in the ith
unit.
X2l: Geometric mean lead loading in pre-activity floor
dust samples in the ith unit.
X3l|: Lead loading in the adjacent pre-activity window sill
dust sample (jth location in the ith unit).
C-4
-------
Table C-2. Data Categories and Covariates Considered in Fitting Model Form (C-1a) to
Lead Loading Data in the Window Replacement Phase
Model «>
Number
(WR-1)
(WR-2)
(WR-3)
(WR-4)
(WR-5)
Data Category
Personal Worker
Exposures Oug/m3)
Area Airborne Exposures
Cug/m3)
1 -hour post-activity lead
loadings from SSDCs at
zero feet from window
(//g/ft2)
1 -hour post-activity lead
loadings from SSDCs at
six feet from window
(//g/ft2)
Estimated lead amount
in 6'x1 ' gradient from
window
Number of
Model Fits
7
6
6
6
7
Covariates Considered Among the Model Fits
(For a given data category, each covariate was included in
only one model fit, with the exception of area airborne
exposures, where covariate XT, was included In every fit
of Model (WR-2)}
Xly: Pre-act. floor dust lead loading at 0 ft. from window.
X2l|: Pre-act. floor dust lead loading at 3 ft. from window.
X3li: Pre-act. floor dust lead loading at 6 ft. from window.
X4I|: Pre-act. window well dust lead loading.
X5ij: Lead content of paint chips collected from the
interior sash/frame of the window.
X6I|: Lead content of paint chips collected from the
exterior sash/frame of the window.
X7,: Pre-activity ambient air lead loading in the ith unit.
X,,,: Pre-act. floor dust lead loading at 0 ft. from window.
X2l|: Pre-act. floor dust lead loading at 3 ft. from window.
X3i|: Pre-act. floor dust lead loading at 6 ft. from window.
X4I|: Pre-act. window well dust lead loading.
X5ij: Lead content of paint chips collected from the
interior sash/frame of the window.
Xei|: Lead content of paint chips collected from the
exterior sash/frame of the window.
X7I: Pre-activity ambient air lead loading in the ith unit
(included in every model fit to area airborne exposure
data).
X,, Pre-act. floor dust lead loading at 0 ft. from window.
X2ll Pre-act. floor dust lead loading at 3 ft. from window.
X3l| Pre-act. window well dust lead loading.
X4| Lead content of paint chips collected from the
interior sash/frame of the window.
X5i Lead content of paint chips collected from the
exterior sash/frame of the window.
XBI Pre-activity ambient air lead loading in the ith unit.
X,, Pre-act. floor dust lead loading at 3 ft. from window.
X2, Pre-act. floor dust lead loading at 6 ft. from window.
X3, Pre-act. window well dust lead loading.
X4i Lead content of paint chips collected from the
interior sash/frame of the window.
X6i Lead content of paint chips collected from the
exterior sash/frame of the window.
X6, Pre-activity ambient air lead loading in the ith unit.
X,| Pre-act. floor dust lead loading at 0 ft. from window.
X2i Pre-act. floor dust lead loading at 3 ft. from window.
X3I Pre-act. floor dust lead loading at 6 ft. from window.
X4| Pre-act. window well dust lead loading.
X5I Lead content of paint chips collected from the
interior sash/frame of the window.
X6i Lead content of paint chips collected from the
exterior sash/frame of the window.
X7| Pre-activity ambient air lead loading in the ith unit.
C-5
-------
Table C-3. Parameter Estimates (and Standard Errors) from Fitting Statistical Models to
Evaluate Factors Relating to Lead Disturbance in Carpet Removal Data
Standard errors of model parameter estimates are in parentheses.
Model
Parameter'11
fJ
P,
P2
P3
04
Y
Estimates from
Modeling
Personal-Air
Lead Cone.
(Model CPT-1)
1.521
(5.104)
0.219
(0.679)
-0.831
(0.487)
-0.120
(0.492)
0.020
(0.024)
Estimates from
Modeling Ambient
Air Lead Cone.
(Model CPT-2)
-3.666
(6.591)
0.089
(0.497)
-0.472
(0.373)
0.010
(0.421)
0.021
(0.018)
-0.421
(1.666)
Estimates from
Modeling Post-
Activity Floor
Dust Lead
Loading
(Model CPT-3)
5.623 *
(2.383)
-0.139
(0.314)
0.289
(0.209)
-0.463
(0.231)
0.021
(0.011)
Estimates from
Modeling 1-hr.
Post-Activity
SSDC Oust Lead
Loading
(Model CPT-3)
1.761
(2.586)
0.078
(0.339)
-0.198
(0.216)
-0.099
(0.251)
0.020
(0.012)
Estimates from
Modeling Post-
Activity Window
Sill Dust Lead
Loading
(Model CPT-4)
3.552
(2.890)
-0.328
(0.392)
0.173
(0.281)
0.591 *
(0.241)
0.009
(0.013)
* Estimate is significantly different from zero at the a = 0.05 level.
111 Model parameters Pk correspond to covariates in the order given in Table C-1.
Table C-4a. Parameter Estimates (and Standard Errors) from Fitting Model (WR-1) to
Evaluate Factors Relating to Lead Disturbance for Personal Exposure Samples
in the Window Replacement Phase111
Covariate
Pre-activity Floor Lead Loading (0 feet)
Pre-activity Floor Lead Loading (3 feet)
Pre-activity Floor Lead Loading (6 feet)
Window Well Lead Loading
Interior Paint Chip Lead Loading
Exterior Paint Chip Lead Loading
Pre-activity Ambient Air Lead Concentration
Model Parameter
P
-0.0372 (1.07)
0.122 (0.833)
0.264 (0.953)
-6.64 (12.1)
1.81 (0.656)
1.75 (1.81)
2.36 (0.899)
3,
0.271 (0.130)
0.313 (0.125)
0.298 (0.147)
0.739 (1.03)
0.253 (0.286)
0.109 (0.695)
0.284 (0.492)
111 Standard errors of model parameter estimates are in parentheses. See Model (C-1 a) for parameter
interpretation.
Note: None of the estimates are significantly different from zero at the 0.05 level.
C-6
-------
Table C-4b. Parameter Estimates (and Standard Errors) from Fitting Model (WR-2) to
Evaluate Factors Relating to Lead Disturbance for During-Activity Area Air
Samples in the Window Replacement Phase'11
Covariate
Window Well Lead Loading
Exterior Paint Chip Lead Loading
Interior Paint Chip Lead Loading
Pre-activity Floor Lead Loading (0 feet)
Pre-activity Floor Lead Loading (3 feet)
Pre-activity Floor Lead Loading (6 feet)
Model Parameter
I*
-8.12 (9.45)
-1.07 (0.984)
0.234(1.11)
-3.16'41 (0.819)
-2.40'41 (0.652)
-2.88'41 (0.756)
py2'
0.0708 (0.367)
0.610 (0.388)
0.0978 (0.580)
-0.51 8(4> (0.186)
-0.394 (0.164)
-5.44141 (0.190)
P2131
0.713 (0.797)
0.821 (0.510)
0.0458 (0.368)
0.355"" (0.0810)
0.344'41 (0.0784)
0.404141 (0.0921)
111 Standard errors of model parameter estimates are in parentheses. See Model (C-1a) for parameter
interpretation.
121 Parameter estimate for pre-activity area air.
131 Parameter estimate for listed covariate.
141 Estimate is significantly different from zero at the a =0.05 level.
Table C-4c. Parameter Estimates (and Standard Errors) from Fitting Models (WR-3) and
(WR-4) to Evaluate Factors Relating to Lead Disturbance for SSDCs at Zero
and Six Feet from Windows Being Removed'11
Covariate
Mode! Parameter
M
e,
Zero feet
Window Well Lead Loading
Interior Paint Chip Lead Loading
Exterior Paint Chip Lead Loading
Pre-activity Floor Lead Loading (0 feet)
Pre-activity Floor Lead Loading (3 feet)
Pre-activity Ambient Air Lead Concentration
-1.82 (5.61)
10.1 (0.533)
10.8(1.05)
8.69 (1.10)
7.90 (0.851)
10.3 (0.668)
1.02 (0.474)
0.0785 (0.226)
-0.289 (0.374)
0.188 (0.132)
0.325'21 (0.123)
0.169 (0.365)
Six feet
Window Well Lead Loading
Interior Paint Chip Lead Loading
Exterior Paint Chip Lead Loading
Pre-activity Floor Lead Loading (3 feet)
Pre-activity Floor Lead Loading (6 feet)
Pre-activity Ambient Air Lead Concentration
4.35 (2.45)
5.45 (0.891)
6.44 (1.22)
3.64 (1.52)
4.39 (1.52)
6.33 (0.750)
0.120 (0.630)
0.0923 (0.361)
-0.382 (0.427)
0.270 (0.219)
0.195 (0.229)
0.659 (0.410)
111 Standard errors of model parameter estimates are in parentheses. See Model (C-1a) for parameter
interpretation.
121 Estimate is significantly different from zero at the a = 0.05 level.
C-7
-------
Table C-4d. Parameter Estimates (and Standard Errors) from Fitting Model (WR-5)
Evaluate Factors Relating to Estimated Total Lead Disturbance in a Six-foot
by One-foot Gradient Perpendicular to Windows Being Removed'11
Covariate
Model Parameter
ft
Total Lead
Window Well Lead Loading
Interior Paint Chip Lead Loading
Exterior Paint Chip Lead Loading
Pre-activity Floor Lead Loading
(0 feet)
Pre-activity Floor Lead Loading
(3 feet)
Pre-activity Floor Lead Loading
(6 feet)
Pre-activity Ambient Air Lead
Concentration
0.276
(5.22)
1 1 .4(2)
(0.984)
10.7'21
(0.528)
9.38'21
(1.02)
8.59(2)
(0.750)
9.47(2)
(0.933)
11.1'2'
(0.617)
ft
0.906
(0.441)
-0.269
(0.351)
0.0871
(0.225)
0.189
(0.122)
0.326'2'
(0.109)
0.228
(0.142)
0.249
(0.337)
111 Standard errors of model parameter estimates are in parentheses. See Model (C-1a) for parameter
interpretation.
121 Estimate is significantly different from zero at the a = 0.05 level.
C.2 ESTIMATING COMPONENTS TO TOTAL VARIABILITY IN LOG-
TRANSFORMED LEAD LOADING DATA IIM THE CARPET REMOVAL AND
WINDOW REPLACEMENT PHASES
The models in Section C.I, used to evaluate the relationship of pre-activity lead
measurements and activity characteristics to lead exposures in a given data category, contained
random effects to characterize unit-to-unit variability (denoted by ou2) and within-unit (i.e.,
location-to-location) variability (denoted by oL2) in the response variable (log-transformed lead
loadings). In addition, a third variance component can be characterized when results from side-
by-side QC samples within a sampling location are available for analysis. This third component,
denoted by oR2, represents variability present in results within a sampling location at a given
study unit, and could be estimated for floor dust and SSDC samples. Estimating these three
variance components provides information on how one variance component dominates another
relative to the total variability in the response.
C-8
-------
The covariates within the models in Section C.I generally were not significantly
associated with lead measurements. As a result, reported estimates of the three variance
components were obtained after fitting the following model to data within each of the data
categories in Tables C-l and C-2:
H. + Lm + e^ , {C.2)
.
where Yjjk is the kth replicate lead loading at the jth location in the ith unit, \i is an estimate of the
study- wide geometric mean, H, is a random effect of the ith unit (normally distributed with mean
zero and variance ou2), LJ(i) is a random effect of the jth location in the ith unit (normally
distributed with mean zero and variance oL2), and ek(lj) is a random effect of the kth replicate
within the jth location in the ith unit (normally distributed with mean zero and variance oR2). Note
that for data categories without side-by-side QC sample data (e.g., personal and ambient air
exposures, window-sill dust lead loadings), the variance component oR2 cannot be estimated.
Restricted maximum likelihood estimation techniques were applied to estimate the
variance components ou2, oL2, and oR2 (where appropriate) using the MIXED procedure in the
SAS® System. For the sample types where no side-by-side QC samples were collected, oR2 was
set to zero. The total variability in log-transformed lead loadings within the data category was
then assumed to be ou2+oL2+oR2. To determine the extent to which a single variance component
dominated total variability, ratios of each variance component to total variability were expressed
in percentage terms.
The results of fitting the variance component Model (C-2) to lead measurements are
presented in Section 7A-2.2.4 for the carpet removal phase and Section 7B-2.2.4 for the window
replacement phase.
C.3 MODELING THE RELATIONSHIP BETWEEN LEAD EXPOSURE AND DISTANCE
FROM ACTIVITY AREA IN THE WINDOW REPLACEMENT PHASE
The sampling design of the window replacement phase specified that for each window
being removed, a settled dust sample from a SSDC be collected at one hour after completion of
the activity at two distances from the window (zero feet and six feet). It was desired to determine
whether distance from the activity could explain the lead loadings found in the SSDC dustfall
samples. The statistical model used to evaluate this relationship was
log(Y,.) = logOt;) + (3d,, + H, + L](i) , (C-3)
where d^ represents the distance from the window of the jth location within the 1th study unit, B is
the (unknown) additive effect of dy on the log lead measurement log(Y,j), and the remaining
model parameters are interpreted in the same manner as model (C-la) in Section C.I. Thus, a
test of whether the parameter P is significantly different from zero indicates whether the distance
is significantly associated with the SSDC lead loadings.
c-9
-------
Results of fitting model (C-3) to the SSDC lead loadings in the window replacement phase
are presented and discussed in Section 7B-2.2.3.
C.4 MODELS IN THE CEP PHASE TO EVALUATE FACTORS OR MEASUREMENTS
IN RELATION TO LEAD EXPOSURE
Section 7C presents the results of several statistical models for lead exposure data from
the CED phase of the EFSS. These models were used to:
• characterize the statistical relationship between personal exposure to airborne lead
and predictor variables such as target activity, substrate, and pre-activity lead levels;
and
• explain the lead loading in settled dust as a function of distance, target activity,
substrate, and pre-activity lead levels.
This section presents the statistical methodology upon which the results of Section 7C were
based.
Like the models in Section C.I, the characterization of the distributions of lead in air,
settled dust, and paint in the CED phase are based on linear models that describe the natural
logarithm of the observed lead loadings as a function of one or more covariates, or predictor
variables. The models also characterize total variability in the response as a function of several
sources of variability, such as worker-to-worker variability and unit-to-unit variability.
The MIXED procedure in the S AS® System was used to fit the models to relate the
measured lead loadings to appropriate explanatory variables. This procedure accommodates both
fixed and random effects factors, as well as a variety of statistical dependence structures.
In describing the model fitting procedure, the following terms are used:
Subunit - a portion of a dwelling unit where a specific CED activity was performed. This
can be an entire room or a portion of a room.
Experimental unit ~ the occurrence of a specific CED activity (i) within a subunit (k) of a
dwelling unit (j). This is generally represented by EUijk.
Table C-5 provides a summary of the statistical models applied to CED data. These
models were generally fit separately to each combination of target activity and substrate. The
models appear in the order that they are referenced in Section 7C.
C-10
-------
Table C-5. Statistical Models Considered in the CED Phase to Characterize Lead Loadings
Model
CED-1
CED-2
CED-3
CED-4
CED-5
CED-6
CED-7
CED-8
Response
Variable11'
log(PEM,jkl)
log(PEMijkl)
log(Dustljk|)
log(Dust,)kl)
log(Dust,Jk,)
log(PEM,Jkl)
log(Dustljkl)
Model Rt
Separately By
Activity/Substrate
Activity
Exper. Unit
Activity/Substrate
Activity
Activity
Activity
Predictor
Variables
None
Substrate
Distance
Distance
Distance, Substrate
Substrate, Paint
Distance,
Substrate, Paint
Variance
Components
"unit "worker
"Task(Umt) "error
"worker "error
"intercept "slope "error
"intercept "slope "error
"intercept "slope "error
"worker "error
"intercept "slope "error
«> PEM =
Dust =
personal exposure monitor result (//g/m3).
dust lead loading (//g/ft2).
Each model in Table C-5 is presented in further detail in the following subsections, along with
estimates of model parameters obtained from fitting the models.
C.4.1 Variance Components Associated with Worker Personal Exposure Levels
Model
To characterize the components of variability in personal exposure measurements, the
following model was fit separately for each combination of target activity and substrate (denoted
by subscript i):
log(PEMijkl) = fJ-t + Unit + Taskk(Unit) + Worker, + e,,.
ijkl
(CED-1)
where
PEM,jkl is the personal exposure measurement (ng/m3) for the 1th worker within subunit k
of the jth dwelling unit;
u., is the mean of the log(PEM) responses over the activity;
Unitj is a random effect attributable to the jth dwelling unit, having standard deviation
°~Umt'
Taskk(Unitj) is a random effect attributable to the occurrence of subunit k within dwelling
unit j, having standard deviation oTask(Umt);
C-11
-------
Worker, is a random effect attributable to the 1th worker performing the activity, having
standard deviation oWorker; and
eijk, is a random term representing replicate variability and measurement error, having
standard deviation oError.
The four random effects are each assumed to be independent from each other and have a normal
distribution with mean zero. In addition, the random effect Worker, is assumed to be crossed
with Unit, and Taskk(Unit,).
In the above model, oUnit is interpreted as a measure of heterogeneity that exists between
different dwelling units and is estimable only if multiple dwelling units were considered. The
component oTask(Unit) is interpreted as a measure of the variability that exists between occurrences
of activity (i) in multiple subunits (k) within dwelling unit (j), and is estimable only if data for
multiple subunits exist. The component oWorker is interpreted as a measure of heterogeneity
between different workers performing the same activity, and is estimable only when data for
multiple workers at a given activity are present.
Results
For each activity, Table C-6 presents estimates of the geometric mean and those variance
components which were estimable as a result of fitting model (CED-1) to log-transformed
personal exposure data.
Table C-6. Estimated Variance Components and Geometric Mean Resulting from Fitting
Model (CED-1) to Personal Exposure Results for a Given CED Activity/Substrate
Combination
Model (CED-1): log(PEMijkl) = //, + Unitj + Taskk(Unitj) + Worker, + eijkl
CED Activity
Drilling
Sawing
Sanding (hand)
Sanding (power)
Door Modification
Component Removal
HVAC
Demolition
Cleanup
Substrate
Plaster
Wood
Plaster
Wood
Wood
Wood
Wood
Wood
Duct
Plaster
Plaster
Wood _^
Model parameter estimates
Geometric
Mean
6.76
15.15
109.99
545.84
254.06
570.82
590.29
343.84
49.62
107.83
24.46
102.36
°Work«i
0.434
0.000
0.000
0.000
0.896
°Unlt
0.860
0.147
0.463
°TuMUl«>
0.451
°&««
0.313
1.295
1.108
0.380
0.589
0.847
0.045
0.101
0.475
0.550
0.205
C-12
-------
C.4.2 Differences in Worker Personal Exposure Lead Levels that are
Attributable to Substrate (Plaster Versus Wood)
Model (CED-1) in the previous subsection was fitted to personal exposure lead levels for
a given combination of CED activity and substrate to estimate variance components in these
levels. For three of the activities (drilling, sawing, and cleanup), separate modeling results were
obtained for wood and plaster substrates (Table C-6). For these activities, it is desired to
estimate differences in modeled worker exposure lead levels that are attributable to wood versus
plaster substrates.
Model
The following model was fit separately for drilling, sawing, and cleanup activities to
characterize differences due to substrate:
log(PEMijkl) = fj, + Pi-Plaster^ + Worker, + ei)kl, (CED-2)
where
PEMykl is the personal exposure lead concentration (ug/m3) for the 1th worker within
subunit k of the jth dwelling unit;
Uj is the mean of the log(PEM) responses over the activity;
P, measures the additive effect on the log(PEM) response that results from performing the
ith CED activity on plaster;
Plasterjjk =1 if EUijk corresponds to performing an activity that disturbed a painted plaster
surface, and equals zero otherwise;
Worker, is a random effect attributable to the 1th worker performing the activity, having
standard deviation oWorker(i); and
eljk) is a random term representing replicate variability and measurement error, having
standard deviation oError(l).
The two random effects are each assumed to be independent from each other and have a normal
distribution with mean zero. Note that those random effects included in model (CED-1) but not
in model (CED-2) were associated with variance components which were not estimable for any
of these three CED activities.
C-13
-------
In model (CED-2), aWorker(l) is interpreted as a measure of heterogeneity between different
workers performing the ith activity, and the error term oError(i) represents replication and
measurement error.
Results
For drilling, sawing, and cleanup activities, Table C-7a presents estimates of model
parameters resulting from fitting model (CED-2) to log-transformed personal exposure data.
Table C-7a. Parameter Estimates from Fitting Model (CED-2) to Personal Exposure Results
for Drilling, Sawing, and Cleanup Activities
Model (CED-2): log(PEMijkl) = jj, + ft-Plaster^ + Worker, + eijkl
CED Activity (i)
Drilling
Sawing
Cleanup
Parameter estimates (standard errors in parentheses)
M
2.696
(0.416)
6.302
(0.232)
4.991
(0.391)
ft
-0.798
(0.529)
-1.602
(0.465)
-1.703
(0.248)
°Work8f (I)
0.428
0.000
0.538
°Erf<» W
0.938
0.570
0.252
Note that model (CED-1) provides separate estimates of oWorker and oError for plaster and
wood substrates, while model (CED-2) provides a common estimate for these two variance
components over both substrates. As a result, model (CED-2) can be considered a reduced
version of model (CED-1), as it has two fewer parameters. A likelihood ratio testing procedure
has been applied to assess the adequacy of the model (CED-2) over model (CED-1). The results
of this testing procedure are presented in Table C-7b.
Table C-7b. Results of the Likelihood Ratio Testing Procedure for Comparing Model
(CED-2) Relative to Model (CED-1)
CED
Activity
Drilling
Sawing
Cleanup
-2*log(U
Model
(CED-1):
Plaster
8.2874
3.7361
5.6944
Model
(CED-1):
Wood
22.0721
6.3087
0.3614
Model
(CED-2)
35.0591
12.7690
6.1779
Likelihood
Ratio Test
Statistic
4.6996
2.7242
0.1221
Degrees of
Freedom
2
2
2
P-Value
0.095
0.256
0.941
C-14
-------
The results of the likelihood ratio testing procedure demonstrate no statistical differences
between the two models at the 0.05 level of significance for any of the three CED activities.
Therefore, model (CED-2) provides an adequate description of the data while estimating the
substrate effect on worker personal exposures to airborne lead.
C.4.3 Association Between Potential Lead Hazard to Occupants and CED
Activity/Substrate Combinations
The goal of the statistical analysis of lead loadings in settled dust samples from stainless
steel dustfall collectors (SSDCs) is to characterize the potential lead hazard to occupants that
may result from the dust and debris generated by a given CED activity. SSDCs were placed at
varying distances from a surface being disrupted by an activity, in order to measure the amount
of lead generated by each CED activity. As a result, the effect that distance has on potential lead
exposure to occupants that results from a specific CED activity can be estimated.
Model
For a given experimental unit EUljk (i.e., the occurrence of activity i within subunit k of
dwelling unit j), the relationship between lead loadings from SSDC samples and distance from
activity was expressed by the following linear regression model:
log(Dustijkl) = P*0(ijk) + P"1(ijk)-Distanceijkl + Errorijkl, (CED-3)
where
log(Dustijkl) is the log-transformed dust lead loading (ug/ft2) associated with the 1th SSDC
sample occurring at EUljk;
P*o(yk) is tne unknown intercept term;
P'i(ijk) istne unknown slope term, relating the change in log(Dustijld) associated with a unit
change in distance;
Distanceljkl is the distance that the 1th SSDC has been placed from the surface being
disrupted by the activity in EUljk; and
Errorjjk, is a random term representing replicate variability and measurement error, having
standard deviation oError(ljk).
Note that model (CED-3) expresses the untransformed lead loading data as an exponential
function:
Dustijkl = exp{P*0(ijk)}-exp{p*1(ijk)-Distanceijk,}-eijkl (CED-4)
where eijkl represents multiplicative error in the model.
C-15
-------
Although model (CED-3) implies separate fits to data for each experimental unit EUijk,
our goal is to determine an overall relationship between lead loadings and distance from activity
for an entire CED activity/substrate combination. Two approaches were used to achieve this
goal:
1 . Two Stage Model Approach. Fit model (CED-3) independently to data for each
experimental unit EUijk. We then assume that the vectors of parameter estimates
(3*o(ijk)> $*i(ijk)) obtained from each fit are a set of independent and identically
distributed observations from a multivariate normal distribution with mean (P*oi, P*,j)
and covariance matrix Sp.;. The mean and covariance matrix are estimated based on
the observed parameter estimates.
2. Population Model Approach. The following random effects model is fit separately
for each combination of CED activity and substrate (denoted by i):
log(Dustijkl) = Poi + Pn-Distanceyu + R0jjk + R1ijk-Distanceijkl + eijkl, (CED-5)
where, in addition to the terms in model (CED-3),
Po, is the population average intercept for activity i;
Pu is the population average slope relating log(Dustijld) to Distanceijkl for activity i,
Roijk = (P*o(ijk)~Poi) is a random effect representing the difference between P0i and P*o(ijk)
(the intercept for EUijk under model (CED-3));
R-iijk = (P*i(ijkrPii) is a random effect representing the difference between pu and P*i(ijk)
(the slope for EUijk under model (CED-3)); and
ejjkl is the error term associated with this model.
Across experimental units, the vectors (R^, Ri,jk) are assumed to be independent and bivariate-
normally distributed with mean (0,0) and covariance SRi. In addition, the error term eijkl is
assumed to be normally distributed with mean zero and standard deviation oError(i).
Results
Both approaches were applied to the settled dust data for each CED activity/substrate
combination to obtain an overall estimate of the linear relationship between log-transformed lead
loading and distance from activity. The results are given in Table C-8.
Table C-8 shows that the slope and intercept estimates and their associated standard
errors are very consistent between the two statistical approaches for each CED activity/substrate
combination. As a result, the population model approach (i.e., fitting model (CED-5)) was taken
in the final characterization.
C-16
-------
The estimated relationship between lead loading and distance for a given CED activity
was later used to quantify the amount of lead disturbed in a 6'xl1 dustfall gradient from the
surface being disrupted. The dustfall gradient approach is presented in Section C.5.
Table C-8. Estimates of the Intercept and Slope (and Associated Standard Errors) from the
Two Stage and the Population Models of Settled Dust Lead Loading as a
Function of Distance
Target
Activity
Drilling
Drilling
Sawing
Sawing
Abrasive
Sanding
Door
Modification
HVAC Removal
Demolition
Substrate
Plaster
Wood
Plaster
Wood
Wood
Wood
Duct
Plaster
Number of
Etys
4
7
2
6
3
6
2
9
Two Stage Approach
r«
(se(p0]»
10.596
(1.133)
12.921
(0.497)
11.286
(1.250)
12.631
(0.472)
11.576
(0.757)
11.126
(0.795)
7.951
(2.304)
9.218
(1.062)
r,,
(sefpj)
-1.718
(0.368)
-1.486
(0.083)
-0.941
(0.141)
-0.668
(0.117)
-0.350
(0.025)
-0.422
(0.098)
-0.351
(0.214)
-0.320
(0.114)
Population Model
Approach
h
(se(p0i)>
10.484
(1.096)
12.877
(0.602)
11.249
(1.203)
12.631
(0.474)
11.522
(0.737)
1 1 .099
(0.707)
8.080
(2.207)
8.775
(0.839)
PL
(sefpj)
-1.671
(0.344)
-1.472
(0.121)
-0.929
(0.247)
-0.668
(0.117)
-0.342
(0.047)
-0.417
(0.094)
-0.375
(0.226)
-0.263
(0.071)
C.4.4 Differences in the Relationship Between Lead in Settled Dust and Distance
that can be Attributed to Substrate (Plaster Versus Wood)
Model (CED-5) in the previous subsection was fitted to SSDC lead loadings for a given
combination of CED activity and substrate to characterize the relationship between lead loading
and distance from the activity. For two of the activities (drilling and sawing), separate modeling
results were obtained for wood and plaster substrates. For these activities, it is desired to
estimate differences in the relationship with distance that is attributable to wood versus plaster
substrates.
C-17
-------
Model
The following model was fit separately for drilling and sawing activities to characterize
differences due to substrate:
log(Dustijkl) = P(W)0i + p'^-Distance^, +Plasterijk [p(p-W)0i +
+ [1-Plasterijk] [R(W)0ijk + R(W)1ijk-Distanceijkl] + Plasterijk [R(p)0ijk CED-6
+ R(p)1ijk-Distanceijkl] + eijkl,
where
log(Dustijkl) is the log-transformed dust lead loading (ug/ft2) associated with the 1th SSDC
sample occurring at EUijk;
p(W)0l is the population average intercept for activity i when performed on wood surfaces;
P(w),, is the population average slope relating log(Dustijkl) to Distance^ for activity i when
performed on wood surfaces;
Distanceljki is the distance that the 1th SSDC has been placed from the surface being
disrupted by the activity in EUijk;
Plasterijk =1 if EUljk corresponds to performing an activity that disturbed a painted plaster
surface, and equals zero otherwise;
P(P~W)OI is the population average difference in intercept attributable to substrate (Plaster-
Wood) for activity i;
P(p"W)h is the population average difference in slope relating log(Dust,Jkl) to Distance^,
attributable to substrate (Plaster- Wood) for activity i;
R(W)oyk = (P*oojk)~P(W)oi) is a random effect which explains the difference between p(W)0l and
P*o(yk) (me intercept in model (CED-3) for EUijk);
R(W)oijk = (P*i(ijk)'P(W)ii) i§ a random effect which explains the difference between P(W)h and
P*1(ljk) (the slope in model (CED-3) for EUljk);
R(P)oijk = (P*o(uk)~P(P)oi) i§ a random effect which explains the difference between P(p)oi and
P 0(ijk)>
R(P)orjk ~ (P*i(yk)-P(P)ii) is a random effect which explains the difference between P(P),J and
P); and
eijk, is the error term.
C-18
-------
The vectors of random effects (R(W)0yk» R(W)ujk)are assumed to be independent and bivariate-
normally distributed with mean (0,0) and covariance matrix SR(W)j. The vector random effects
(R^oijkJ R(P)iijk) arQ assumed to be independent and bivariate-normally distributed with mean (0,0)
and covariance matrix SR(P)i. The error term eljkl is assumed to be normally distributed with mean
zero and standard deviation o.
Error (i)'
Results
Table C-9a presents the parameter estimates that result from fitting model (CED-6) to lead
loading data separately for drilling and sawing activities.
Table C-9a. Parameter Estimates from Fitting Model (CED-6) to Settled Dust Lead
Loadings for Drilling and Sawing Activities
Model CED-6:
log(Dustljkl) = P'w>0i + P'W|
Distanceijkl]
Distanceijkl
Plaster
{1-Plaster,,) [R(W)0ijk
ijk
(p"W|
. +
ii]k- Distanceijk|] + Plaster
Distanceijk|]
ijk
RlP1ii-k
CED
Activity
Drilling
Sawing
ROW
P «
12.884
(0.558)
12.631
(0.553)
P^n
-1.474
(0.110)
-0.668
(0.127)
n(P-W)
P w
-2.536
(1.175)
-1.350
(1.342)
rw',«
-0.133
(0.321)
-0.271
(0.254)
R%,k
Std Dev.
0.543
0.501
RIWI
T(k
Std Dev.
0.000
0.179
R'%
Std Dev.
1.490
1.282
R 1Jk
Std Dev
0.422
0.000
"error (I)
1.402
1.253
Note that model (CED-5) provides separate estimates of oError for plaster and wood
substrates, while model (CED-6) provides a common estimate over both substrates. As a result,
model (CED-6) can be considered a reduced version of model (CED-5), as it has one fewer
parameter. A likelihood ratio testing procedure has been applied to assess the adequacy of the
model (CED-6) over model (CED-5). The results of this testing procedure are presented in Table
C-9b.
Table C-9b. Results of the Likelihood Ratio Testing Procedure for Comparing Model
(CED-6) to Model (CED-5)
ceo
Activity
Drilling
Sawing
-2«lofl{LJ
Model (CED-
6):
Plaster
69.8120
36.6159
Model (CED-
6):
Wood
135.3740
102.2258
Model (CED-
5)
208.0132
139.8868
Likelihood
Ratio Test
Statistic
2.8272
1.0451
Degrees of
Freedom
1
1
P-Value
0.093
0.307
C-19
-------
The results of the likelihood ratio testing procedure demonstrate no statistical differences
between the two models at the 0.05 level of significance for either CED activity. As a result,
model (CED-6) provides an adequate description of the data while estimating the substrate effect
on potential occupant exposures to lead in settled dust.
C.4.5 Investigating the Substrate Effect on Worker Personal Exposure Monitoring
Results after Adjusting for the Effects of Pre-activity Paint Lead Loading
For three CED activities (drilling, sawing, and cleanup), model (CED-2) presented in
Section C.4.2 investigated the difference in worker exposure to airborne lead that is attributable
to substrate effects (wood versus plaster). The results of that analysis demonstrated a potential
substrate effect, although it was not statistically significant at the 0.05 level. One cause for the
substrate effect was higher paint lead loadings on wood surfaces in comparison to plaster
surfaces. By adjusting model (CED-2) to include an effect of paint lead loading, we can
investigate whether this cause is appropriate.
Model
The following random effects model, fit separately to each of the three activities (denoted
by i), was used to measure differences in the response of worker exposure to airborne lead that
are attributable to substrate after adjusting for the effect of paint lead loading:
log(PEMijkl) = //| + Pa-Paint,^ + ft-Plaster^ + Worker, + eijkl/
(CED-7)
where the model terms have the same interpretation as in model (CED-2), with the following
addition:
Poi measures the effect of paint lead loading on the log(PEM) response for activity i.
Results
Table C-10 presents the parameter estimates and associated standard errors from model
(CED-7).
Table C-10. Parameter Estimates from Fitting Model (CED-7) to Personal Exposure Results
for Drilling, Sawing, and Cleanup Activities
Model CED-7: log(PEMijU) = # + Pa-Paint^ + P^PIaster^ + Worker, + e.
ijkl
CID Activity
Drilling
Sawing
Parameter estimate? Standard errors in parentheses) \
H
1.916(0.593)
6.238 (0.508)
'I fe,
0.109 (0.064)
0.009(0.061)
I '•&• I
-0.374 (0.602)
-1.556(0.597)
°W«*«rffl
0.079
0.000
:,*W*
0.938
0.623
C-20
-------
C.4.6 Investigating the Substrate Effect on the Relationship Between Lead Loading
in Settled Dust and Distance After Adjusting for the Effects of Pre-Activity
Paint Lead Loading
For two CED activities (drilling and sawing), model (CED-6) presented in section C.4.4
investigated the difference in lead loadings from SSDC samples that is attributable to substrate
(wood versus plaster). The results of that analysis demonstrated a potential substrate effect,
although it was not statistically significant at the 0.05 level, indicating higher lead loadings in
settled dust for activities which disturbed a lead-painted wood surface. One cause for the
substrate effect was higher paint lead loadings on wood surfaces in comparison to plaster
surfaces. By adjusting model (CED-2) to include an effect of paint lead loading, we can
investigate whether this cause is appropriate.
Model
The following model was used to measure differences in lead loadings from SSDCs that
are attributable to substrate within drilling and sawing activities (denoted by i), after adjusting for
the effect of paint lead loading:
log(Dustijkl) = 3(W)0i + 0™,, • Distanceiikl + Plasterijk [p(p-W)0i + p'™"1,, • (CED-8)
Distancejjk|] + ovPaintijk + [1-Plasterijk] [R1^ + R(W|1ijk •
Distanceijkl] + Plasterijk [R|p|0ijk + R(p|1iik-Distanceijkl] + eijkl ,
where the model terms have the same interpretation as in model (CED-6), with the following
addition:
a, measures the effect of paint lead loading on the log-transformed lead loading for
activity i.
Results
Table C-l 1 presents the parameter estimates that result from fitting model (CED-8) to
drilling and sawing activity data.
C-21
-------
Table C-11. Parameter Estimates from Fitting Model (CED-8) to SSDC Lead Loading Data
for Drilling and Sawing Activities
Model CED-8:
log(DustijH) =
^. Djstance.j|ti + p|asterijk
(1-Plasterljl()
R
-------
error) for each CED activity we used the Multivariate Delta Method (Bishop, Fienberg and
Holland 1975, page 492):
If (X,,X2) follow a multivariate normal distribution with mean (65,63) and variance S, then
a continuous and differentiate function of X,and X2, f(Xl5X2) has the following
asymptotic distribution:
f(Xj,X2) is asymptotically distributed with mean f(6,,62) and variance
In our estimates of the 6'xl' gradient lead-loading (and associated standard error) for each
CED activity, f(00i, 0 u) has the following form:
6
o
and [df/d(poi, pn)] has the following form:
exp((50,.+6-(51/)-exp(|30,.)^ ( explp^+e-fl^-explp1,,,.) e-exptpVB-p1,,)
The estimated 6'*r gradient lead-loadings for each CED activity/subtask combination was
estimated as f($Qi, 0 }i) and their associated standard deviations were estimated as the square root
of [df/dtfJoj.fi^lStdf/d^oj,^)]. The estimated [0-1] and [5-6] foot SSDC and their associated
standard deviations were estimated by adjusting the limits of integration on the function f(P0l,Pi,)
and adjusting df/d(p0i,ph) accordingly.
C.6 METHODOLOGY UNDERLYING NORMAL DISTRIBUTION THEORY
CONFIDENCE INTERVALS. CONFIDENCE INTERVAL FOR A PERCENTILE.
AND CONFIDENCE INTERVAL FOR THE PROPORTION LESS (GREATER)
THAN A SPECIFIED VALUE
This section discusses the methodology underlying the calculation of confidence intervals
on distribution percentiles and on the probability of being less (greater) than a specified level.
This methodology is applied to the (natural) logarithms of personal exposure lead concentrations
(Hg/m3 of air). Since air volume is proportional to sampling time, these loadings are effectively
adjusted for sampling time. Confidence intervals are constructed for each R&R activity.
The confidence interval procedures are based on components of variance models having
either one, two, or three variance components. The appropriate model was determined by those
variance components that could be estimated from the available data. The variance components
within these models are as follows:
C-23
-------
Model #1
(one variance component): Worker-to-worker
Model #2
(two variance components): Unit-to-unit
Worker-to-worker within units
Model #3
(three variance components): Unit-to-unit
Worker-to-worker within units
Replicate variability within workers
The only activity where three variance components can be estimated is demolition.
For each activity where the two- and three-variance component models are appropriate, it
is assumed that the data sets are (approximately) balanced or nearly balanced (i.e., the number of
workers within units is (approximately) the same across units and the number of replicate
determinations within workers is (approximately) the same across workers within units). If the
data sets are only nearly balanced, we use the harmonic mean of levels (e.g. workers within units)
in the confidence interval calculations.
Models
The forms of the variance component models are as follows:
One Variance Component Mode/:
Yk= H +ek k=l,...,n Yk ~ ind N(n, oe2)
Two Variance Components Model:
Yy = u + hj + wj(i) i=l,...,I; j=l,...,J
hf ~ ind N(0, oh2); wj(i) ~ ind N(0, ow2)
Three Variance Components Model:
Yjjk = u + hj + wj(i) + eijk i=l,-,I; j=l,.»J; k=l,...,n
wj(i)~indN(0,ow2);
eijk~indN(0,oe2)
C-24
-------
C.6.1 Normal Theory Two-Sided 100(1 -a) Percent Confidence Interval on the
p-th Distribution Percentile
One objective of the analysis of personal exposure data is to compare (across activities)
potential lead exposures associated with each R&R activity. This is done by summarizing the
distribution of personal exposure data for each activity by a single summary statistic, and
estimating the error associated with the statistic. This statistic has been taken to be the 75th
percentile of the distribution. The same methods were used to estimate the 50th and 95th
percentiles as well. This section presents the statistical approach taken to calculate confidence
intervals on these estimated distribution percentiles.
The (1-a)* 100% two-sided confidence intervals calculated in this section are based on the
noncentral Student-t distribution. The methods for calculating these confidence intervals are as
follows, according to the variance component model applied to the personal exposure data for a
given R&R activity:
A. One Variance Component
Yk (k=1,...,n) ~ ind
where v=n-l d.f.
Let u + £poe denote the p-th percentile of the normal distribution, Q<, pz 1; £p = $"'(p).
We wish to place a 100(1-a) percent two-sided confidence interval on (j. + £poe. Determine kl5 k2
such that
PlY.+k^^+^o^Y.+k^)^! -a
Determine k,, k2 such that
C-25
-------
Rearranging these inequalities implies that
is the lower 100a/2 percentile of the noncentral t distribution with u d.f. and noncentrality £p\/n.
Similarly
k2=-kv;i-
vn
is the upper 100a/2 percentile of the noncentral t distribution with u d.f. and noncentrality £p
A 100(1 -a) percent two-sided confidence interval on \i + £poe is thus
r7 s)
where t,,(£p\/n) represents the noncentral t-distribution with u=n-l d.f and noncentrality parameter
£pv/n, and where ip=&l(p).
B. Two Variance Components
h, ~ ind N(0, oh2); wj(l) ~ N(0, ow2); h,, wj(l) are independent
ft = Y.. = -\ E E Y, /> - N +
The mean square among units is MSH, with 1-1 df.
i _ _
MSW = |Y - ')2 MSW ~
MSH = -^- £ (Yr-Y..)2 MSH ~ (Jo2 + c
The mean square among workers within units is MSW, with I(J-l)d.f.
1 y* 5
KJ-D M f.
Let
C-26
-------
r2 - o2 + a2
Cp2 B 02/l + 02/IJ
Then t2,
-------
the lower 100 a/2 percentile of the noncentral t distribution with 1-1 d.f. and noncentrality
parameter E — . Estimate T/
-------
1 ' J n I O2 O2 O2
*• \y I \"^ *V^ \"^ \f M h il n w e
" • Y- • IS 5 £ £ Y"1 " "TT + IT * uS.
The mean square among units is MSH, with 1-1 d.f.
MSH = -^1 E (Y, - V...)2 MSH ~ (Jno2 + no2 + o2) X2y(l-1)
1-1 1=1
The mean square among workers within units is MSW, with I(J-l)d.f.
MSW = ITTTT % (9 ' 9)2 MSW
KJ-I) 1=1 j=
The mean square among replicate determinations within workers is MSB, with IJ(n-l) d.f.
MSB = ^^ E E E (Y.jk - Y,.)2 MSB - oe2 Xj^^Jtn-l)
I J n
MSB = ! ^
Let
T2 s o2 + o2 + of
q>2 s o2/! + 02/U + of/Un
T2,
-------
. .; a/2 E -
M ^
k2 = TTT t'i-1 '• 1-«/2
Estimate T/al
The theory presented in this section is used to calculate a (l-a)*100% two-sided
confidence interval on the probability that a given random variable Y exceeds a pre-specified
threshold value (a), based on assumptions on its underlying distribution. In the EFSS, this
method is applied to obtaining confidence intervals on the proportion of workers whose personal
exposure sample results exceed the OSHA permissible exposure limit of 50 |J.g/m3.
Assume that
and let
Then
This implies that
where E = a) = 1-<
p = 1 -q =
-------
For each value of p, 0
-------
q, = 1 -pu
qu = 1 -PI
C.7 MODELS USED TO ESTIMATE MEAN EXPOSURE LEVELS ACROSS
STUDIES IN COMBINING OTHER SOURCES OF SURFACE
PREPARATION DATA
The individual data points were available for all surface preparation studies located in the
search for other sources of data. Therefore within-study and between-study variability could be
estimated directly from the data rather than approximated by statistical meta-analytic techniques.
To characterize the components of variability in personal worker exposures the following model
was fit separately for interior and exterior dry surface preparation:
ij) = fj + Study, + Worker^ (OS-1)
where
PEM,j is the personal exposure lead concentration (|ag/m3) for the jth worker in the ith
study
(4, is the mean of the log (PEM) responses
Study; is the random effect of the ith study (normally distributed with mean zero and
n
variance os)
Worker, is a random effect of the jth Worker (normally distributed with mean zero and
variance o^)-
Parameter estimates for each model fit are presented in Table 7E-4 of Volume 1 . Estimates of
the 75th and 95th percentiles of the distribution and confidence intervals for the geometric mean,
75th and 95th percentile were calculated according to the methodology presented in Section C.6
of Appendix C. Likewise, estimates of the percent of workers exceeding the OSHA PEL and
their associated 95th confidence interval were calculated according to the methodology presented
in Section C.6 of Appendix C.
C-32
-------
APPENDIX D
QUALITY CONTROL
D-1
-------
D.O QUALITY CONTROL
To ensure that the sampling and analysis protocols employed in the EFSS produced data of
sufficient quality, a number of different quality control (QC) samples were included in the study
design for each study phase. Field QC samples were intended to help assess variability
introduced by the sampling method and to detect potential biases from field sources such as
sample transfer and handling. The types of field QC samples collected in this study were:
• Field blanks: Lead-free samples prepared in the laboratory and transported to the
field, to assess potential bias or contamination. Field blanks consisted of a sample
collection medium (dust bottle, wipe, filter cassette, paint chip collection vial)
removed from packaging, connected to any necessary sampling device (vacuum, air
pump) and immediately removed, then packaged without actually taking a dust
sample. At a given study unit, one field blank of each sampling media to be used on
that day was taken prior to R&R activities and field sampling.
• Field side-by-side samples: Vacuum and wipe samples collected in areas adjoining
"regular" sample areas, to determine variability due to the sample collection process.
In addition, laboratory QA/QC measures were implemented during the analysis of the field
samples.
Section D.I of this appendix presents a statistical summary of the field blank data. Tables
of the side-by-side sample results, along with results for adjoining regular samples, are presented
in Section D.2 (a discussion of side-by-side sample results is found in Section 6.4). Results of
laboratory QA/QC sample analysis that deviated from data quality objectives are summarized in
Section D.3.
D.1 FIELD BLANKS
The following field blanks were collected within each phase of the EFSS:
• Carpet removal: Four field blank samples were collected at each of the eight study
units prior to carpet removal activities, one for each sample type considered in this
phase (vacuum, wipe, personal air, and ambient air);
• Window replacement: Three field blank samples were collected at each of the four
study units prior to window replacement activities, one for each sample type
considered in this phase (vacuum, personal air, and ambient air);
• CED phase: at a given study unit, one field blank sample was collected for each
sample medium to be employed on that day (vacuum, wipe, personal air, paint chip).
Field blank results for all sample types were reported in terms of lead content (i.e., ^g lead per
sample). In addition, vacuum field blank samples were reported in ^g lead per gram of dust in
the sample.
D-2
-------
Tables D-la through D-lc present the results of analysis on the field blank samples for the
carpet removal, window replacement, and CED phases, respectively. The results (jig/sample) for
all three phases were generally close to the analytical detection limit associated with the given
sample medium, despite the frequency to which detected results were observed within the field
blanks. Detected results are primarily the result of instrument sensitivity. These tables indicate
that no apparent bias in sample collection and handling was observed as a result of reviewing the
field blank data.
Table D-1a. Descriptive Summaries of Field Blank Sample Results Within
the Carpet Removal Phase
Arithmetic
Mean (S.E.)
Minimum/
Maximum
Result
% Not
Detected
Results
Sample Types
Vacuum
//g/sampfe
2.19
(1.09)
0.450/
3.12
62.5%
pate
679.5
(374)
173. 21
1,376
57.1%
(n = 7)
Wipe
/yg/sample
2.72
(0.58)
2.34/
3.94
50%
Personal Air
_/yg/sample
0.215
(0.250)
0.093/
0.831
0%
Ambient Air
//g/sample
0.209
(0.073)
0.1 107
0.312
0%
Note: This summary is based on analysis of n = 8 field blank samples.
Table D-1b. Descriptive Summaries of Field Blank Sample Results
Within the Window Replacement Phase
Arithmetic
Mean (S.E.)
Minimum/
Maximum
Result
% Not
Detected
Results
Sample Types
Vacuum
//g/sample
2.09
(0.58)
1.72/
2.94
50%
//g/a
1259.81
(1422.82)
45.008/
3263.33
50%
Paint Chip
//g/sample
6.08
(0.01)
6.07/
6.09
100%
Personal Air
//g/sample
0.080
(0.006)
0.075/
0.089
0%
Ambient Air
//g/sample
0.102
(0.050)
0.047/
0.151
0%
Note: Summaries for vacuum, personal air, and ambient air are based on analysis of n = 4
field blanks. Summaries for paint chip are based on analysis of n = 2 field blanks
(at units 1-01 and 4-01).
D-3
-------
Table D-1c. Descriptive Summaries of Field Blank Sample Results Within the CED Phase
Arithmetic
Mean (S.E.)
Minimum/
Maximum
Result
% Not
Detected
Results
Sample Types
(Number of Samples)
Dust Vacuum
-------
a
•o
a
a.
E
a
(0
a
_«
I
QC
(0
•o
a>
I
0 «
.i w
a n
a. £
a a
"
o
•0 .
» 0)
w ^
c °-
o E
+3 n
2 w
c O
u a
8 .
o 2
O o
O)-1?
C i
u
OOOOO'-'-O
CSIDO'COCOOO-
80 *~ *~ o w O) ^
OOOCSCMOO
(D (O (O tO (O tO CO CO
o^^o^^
OOOCNOOO
CO CO CD
CO CO CO CO
oooooooo 00900000
coooor^ooo^'-
OOCOCOCOIVOOCN
pppppppp
dddddddd
I'
mcoo«-Ococoo
-
dddddd
1
COCOCOLOtvCOCOCN
• ^^CNCNTtCOOJCN
OO^o^c^rN.
OOOCMOOiO
CO (D (O CO (D (O tO
— iva>o>r--coco
PppppppO
dddddddd
to
r-COOPJ'fOO'-CO
§O »~ *~ o ^ *~ r^>
O O O CM O
-------
^
0)
3
o
0
CNJ
Q
2
ra
1
1
1
1
si
|i
g
o
.s
1
i
5
I
1
£
1
w
e
1
i
w
VI
tf
1
£
1
^
§
!
i
u
I
«
P
|i
*°5
1
u
s
u
«£
o i
m 5
fs
-;qqoqoq
dqdddddd
cncg^^cn^coco
<£^r^3gWcrig
5ST'fi- CM
Qcoo)co(¥J^'(o*~f^
t(Y)^(Q^OO^OCO
3 QO CO r- O ^ 00 ^ O)
3 r>.coro»-^-
o
CQ
ScoS^SSSS
O)CMr>jO) •*• r*«" o o
CM g «> 5 oi
i— i— CM ^"
_
1
3
i
^
3
o
C-
§
|s^coe0rS5cM
O^TtOOcOgOOcM
•- *" CO
1
to
OO^fCMCMCM^1
fl) «— CM «• ft 03 Ol *~
w
1 . co „ . . , ,
'So^«-«-CDCMr-
XOOOCMOCMCM
c
I
a
V)
«
3
Q
0)
a
CM °, 00 CO »- «- 00
m ^ *f m m *i i-
CM *** CM CD r^ r-
§00 UJ O i- 05 CO
in «fr CM CM •- *
•- ^ ^ «t 00 O) <-
CO i- CO CO CO CO CO
r>- i- CM oo en i^ o
O «- •- •- «- CM 00
O O O CM CM CM O
CO CO CO CO CO CO CC
CM co re T- CM co in
0000000
«- »- t- CM) CM CM CM
D-6
-------
CO
•o
(0
Q.
E
(0
CO
to
Q
_w
3
Q)
DC
(0
•i
9
•O
0) 0)
c in
*: ra
« £
O Q.
-
E -
a a
U
(0
UJ
j= K
f?
c .£
*
« 0)
c 9-
O
c w
o S
O CO
o>-9
c o>
•5 2
1:1
to 5
co <
CNI
6
.2
to
rv m o
t- O CO
O in co
00-
O r^ O
in rv Q
oS * cn
co cn r-
00 ^ ^
K " £
^ CO «-
«M 10 2
CM
cn *f oo
r- o CM
§CM O
in co
d ^ ^
CM cn
co rv
"t CM
CM CO
in o
"*. "*•.
CO CM
Tf rv
rv CM
co *
co co cn
000
CM CM cn
rv m co
't CM
*)• CM T-
Scn in
cn in
q fv m
d d CM
rv f
CM ,J
d «>
^ « -
£ CM ^
3.3
CM cn in
d £ S
CM fn CO
oo in TV
00 O i-
8 CM in
cn rv
d ~ d
cn cn cn
'- in o
in o CM
*"• 9 •"
TOO
m " CO
co O. ^
o § g
S S In
,_ (O cn
t- cn cn
oi CM -.
r-: in m
co co co
g> cn Tt- o
55 •* t
CM 00
rv en o
co r- in
co co cn
o o o
CM
co
o
d
S
CM
't
rv co
cn co
— O
cn in
co rv
•* «-
O in
rv in
i- in
?O
CM
m ^t
co co
O •<*•
CM r-
? § ?
r** co
p* PS 01
in co tj>
^ in co
SO o
(O (D
9 9
CO *
999
CM CO *
O
CM
D-7
-------
Table D-2c. Loadings for Each Sample Pair Defined by a Regular Dust Sample and an
Adjoining Side-by-Side QC Sample, Both Collected by Vacuum Techniques, in
the R&R CED Phase
1 Sample Taken First in the Pair 1 Sample Taken Second in the Pair I
MR) ID
Time
Collected
Loading 1
Oug/ft2) 1 MRI ID
Time Collected
Loading 1 Difference in
(//a/ft') 1 Loadings fo/g/ft'J
1372 N. Carey Street, Baltimore, MD
Wall Demolition,
2nd Floor
Bathroom
Wall Demolition,
3rd Floor
Bathroom
Wall Demolition,
Kitchen
61159
61799
61849
10:15
15:08
13:27
2671.70
817.26
3438.80
60969
61669
61834
10:22
15:15
13:30
1365.40
1297.50
2521.80
1306.3
-480.24
917.00
960 Lipan Street, Denver, CO
Door, Baseboard,
Frame Removal,
Bedrooms
Wall Demolition,
Kitchen
Window Sanding,
Dining Room
Wall Demolition,
Dining Room
60448
60483
60418
60633
60878
61213
15:06
15:50
17:01
15:47
15:56
11:06
48944.5
75693
138631
918.29
1187.70
45.25
60403
60413
60443
60848
60843
61198
15:18
15:55
17:07
15:51
16:00
11:14
23794
25652
11968
1140.2
1241.8
98.64
25150.5
50041
126663
-221.91
-54.10
-53.39
D.3 SUMMARY OF LABORATORY QA/QC FINDINGS
In their data reports, MRI summarized the results of analyzing laboratory QA/QC samples
within each instrumental and sample preparation batch. In the carpet removal phase,
environmental samples were analyzed across eight instrumental analysis batches and nine sample
preparation batches Samples in the window replacement samples were analyzed across nine
instrumental analysis batches and thirteen sample preparation batches. Samples in the CED
phase were analyzed across 16 instrumental analysis batches and 13 sample preparation batches.
In all three phases, the results of analyzing initial calibration verification and continuing
calibration verification samples in each instrumental analysis batch were within the protocol
criteria of ±10%. This indicates that the analytical instrument was properly calibrated for all of
the sample analyses.
Table D-3a through D-3c report the status of meeting data quality objectives within
batches in the carpet removal, window replacement, and CED phases, respectively. These tables
also show the number of field samples, duplicate samples, and field blanks analyzed in each of
the instruments analysis batches. Most of the incidents where data quality objectives were not
met occurred in analysis of NIST standard reference material (SRM) 1646.
D-8
-------
Table D-3a. Summary of Conclusions Made in Analysis of Laboratory QA/QC Samples
in the Carpet Removal Phase
instrument
Analysis Batch
Total Number of Field
Samples Analyzed
Sample
Preparation
Batch
Nates
Personal Air (MCE Filter) Cassettes
V10293A
V11013A111
V10283B
20 Field Samples
7 Field Blanks
3 Field Samples
0 Field Blank
19 Field Samples
8 Field Blanks
605
606
Percent recovery for NIST SRM 1646 was 71%, which
was below the lower control limits. Historical data
continued to be monitored throughout the study. Percent
recovery for NIST SRM 2704 was 90%, meeting the
data quality objectives.
Percent recovery for NIST SRM 1 646 was 77.5%, which
falls below the lower control limit as published in the
QAPjP. Historical data continued to be monitored
through the study, and no further corrective action was
deemed necessary. Percent recovery for NIST SRM
2704 was 105%, which meets the data quality
objectives.
Dust Wipe Samples
E07223A
E10193A
24 Field Samples
8 Field Blanks
32 Field Samples
2 Field Blanks
601
602
Percent recovery for NIST SRM 2704 and NIST SRM
1646 were 93.5% and 102.2%, respectively, meeting
the data quality objectives.
Percent recovery for NIST SRM 1646 was 75.3% for
batch 602, which falls between the lower warning limit
and the lower control limit as published in the QAPJP.
Historical data continued to be monitored through the
study, and no further corrective action was deemed
necessary. Percent recovery for NIST SRM 2704 was
94.6%, meeting data quality objectives.
Dust Vacuum Samples
E10213A
E10263A
E11303A121
36 Field Samples
1 Field Blank
36 Field Samples
2 Field Blanks
34 Field Samples 2
Field Blanks
38 Field Samples
2 Field Blanks
5 Field Samples
0 Field Blanks
2 Field Samples
0 Field Blanks
603
604
607
608
617
607'3'
Percent recoveries for the NIST SRM 1646 were 71.7%
for batch 603 and 75.0% for batch 604, both of which
fall below the lower control limits as published in the
QAPjP. Historical data continued to be monitored
through the study, and no further corrective action was
deemed necessary. Percent recoveries for the NIST SRM
2704 were 97.1 % for batch 603 and 86.9% for batch
604, both of which meet data quality objectives.
Percent recoveries for NIST SRM 1 646 were 94% for
batch 607 and 86% for batch 608. Percent recoveries
for NIST SRM 2704 were 104% for batch 697 and 98%
for batch 608. All of these results meet data quality
objectives.
Percent recoveries in batch 617 were 87.3% for the
NIST SRM 1646 and 87.4% for NIST SRM 2704. Both
of these results meet data quality objectives.
111 One sample preparation batch was divided into two instrument analysis batches.
This instrument analysis batch consisted of five carpet samples and four window samples.
Two samples from sample prep batch 607 were reanalyzed due to dilution problems in the original analysis.
D-9
-------
Table D-3b. Summary of Conclusions Made in Analysis of Laboratory QA/QC Samples
in the Window Replacement Phase
Instrument
Analysis Batch
Total Number of Field
Samples Analyzed
Sample
Preparation
Batch
Notes
Personal and Ambient Air (MCE Filter) Cassettes
V11223A
V11233A111
V02034A
26 Field Samples
4 Field Blanks
1 Field Sample
10 Field Samples
0 Field Blanks
610
621
The percent recovery for NIST SRM 1 646 was 78.1 %,
which falls between the lower warning and control
limits. For NIST SRM 2704 the percent recovery was
104.1 %, which meets the data quality objectives.
Percent recovery for NIST SRM 1646 was 78.2%, which
fall between the lower warning and control limits. For
NIST SRM 2704 the percent recovery was 87.6%,
which meets the data quality requirements.
Paint Chip Samples
E11173A
1 9 Field Samples
3 Duplicates
1 Field Blank
615
Percent recoveries for NIST SRM 1 579 were 97.6, 93.4,
and 99.7%, meeting the data quality objectives. Percent
recoveries for AIHA SRMs were 103, 95, and 105% for
AIHA 1 and 101, 93, and 102% for AIHA 2. All of
these results meet the data quality objectives.
Oust Vacuum Samples
E11153A
E11163A121
E11303A131
E02254A
49 Field Samples
3 Field Blanks
39 Field Samples
0 Field Blank
8 Field Samples
1 Duplicate
31 Field Samples 1
Field Blank
612
614
613
614
617
618
619
622
Percent recovery for NIST SRM 1646 was 87.3% for
batch 612 and 78.1% for sample batch 614. The result
for bath 614 fall between the lower warning and control
limits. Percent recoveries for NIST SRM 2704 was
97.6% for batch 612 and 88.22% for batch 614, both
of which meet data quality objectives.
The percent recovery for the NIST SRM 1646 was
100% for batch 613, and for NIST SRM 2704 it was
102%. Both of these results meet the data quality
objectives.
The percent recovery for the NIST SRM 1 646 was
87.3% for batch 617 and 83.6% for batch 618. The
recoveries for NIST SRM 2704 were 87.4 and 93.7% for
batches 617 and 168, respectively. All of these results
meet the data quality objectives.
The result for the NIST SRM 1646 was 99.0% for batch
619 and 96.7% for batch 622. The percent recoveries
for NIST SRM 2704 were 102 and 95.3% for batches
619 and 622, respectively. All of these results meet the
data quality objectives.
"' The samples from instrument batches V11223A and V11233A were combined when assessing data quality objectives.
121 Instrument analysis batch E11163A included the reanalysis of one sample from sample prep batch 614.
131 Instrument analysis batch E11303A included two samples from the carpet removal phase.
D-10
-------
Table D-3c. Summary of Conclusions Made in Analysis of Laboratory QA/QC Samples in
the CED Phase
Instrument
Analysis Batch
Total Number of Reid
Samples Analyzed
V03314A
V041 34A
V05264A
39 Field Samples
2 Field Blanks
22 Field Samples
1 Field Blank
29 Field Samples
2 Field Blanks
Sample
Preparation Batch
Notes
Personal Air (MCE Filter) Cassettes
629
630
635
Data quality objectives met.
Percent recovery for NIST SRM 1646 was 75.4%, which
falls between the lower warning limit and lower control
limit as published in the QAPjP.
Percent recovery for NIST SRM 1646 was 72.7%, which
falls below the lower control limit as published in the
QAPjP. Historical data continued to be monitored through
the study, and no further corrective action was deemed
necessary.
Dust Wipe Samples
V04264A
E04254A
E04214A
E05254A
4 Field Samples
9 Field Samples
6 Field Samples
2 Field Blanks
14 Field Samples
2 Field Blanks
631
638
Percent recovery for NIST SRM was 77.2%, which falls
between the lower warning limits and the lower control
limit as published in the QAPjP.
Percent recovery for both NIST SRM 2704 and NIST SRM
1646 were 63.2% and 50.7%, respectively, which does
not meet the data quality objectives. Historical data
continued to be monitored throughout the study, and no
further corrective action was deemed necessary.
Paint Chip Samples
E03294A
V04264B111
E05264A
39 Field Samples
4 Duplicates
2 Field Blanks
2 Field Samples
22 Field Samples
3 Duplicates
628
636
Data quality objectives met.
Data quality objectives met.
Dust Vacuum Samples
E04044A
E04134A
E041 84B121
V04264A12'
E05264B
E06014B131
79 Field Samples
1 3 Duplicates
83 Field Samples
12 Duplicates
1 Field Blank
1 Field Sample
1 Duplicate
1 Field Sample
1 Duplicate
75 Field Samples 5
Duplicates
4 Field Blanks
2 Field Samples
624
625
626
627
633
634
Percent recovery for NIST SRM 1646 was 78.5% for batch
624, which falls between the lower warning limit and the
lower control limit as published in the QAPjP. The result
for batch 625 was 65.6%, which does not meet the data
quality objectives. Historical data continued to be
monitored through the study, and no further corrective
action was deemed necessary.
The result for the NIST SRM 1646 was 67.3% for batch
626, which falls below the lower control limits as published
in the QAPjP. Historical data continued to be monitored
through the study, and no further corrective action was
deemed necessary. The result for the NIST SRM 2704
was 75.3% for batch 626, which falls between the lower
warning limit and the lower control limit as published in the
QAPjP.
The result for the NIST SRM 1646 was 78.4% for batch
634, which falls between the lower warning limit and the
lower control limit as published in the QAPjP.
"' Rerun samples from Instrumental Analysis Batch No. E03294A (over range).
121 Rerun samples from Instrumental Analysis Batch No. E04134A.
Samples moved from Instrumental Analysis Batch No. E05264B.
D-11
-------
References for Volume II
Bishop, Y.M.M., Feinberg, S.E., and Holland, P.W. (1975).
Discrete Multivariate Analysis: Theory and Practice. Cambridge, MA: The MIT Press.
Grubbs, F.E. (1950). Sample criteria for testing outlying
observations. Annals of Mathematical Statistics. 21:27-58.
Hahn, GJ. and Meeker, W.Q. (1991). Statistical Intervals. A
Guide for Practitioners. New York: John Wiley and Sons, Inc.
Odeh, R.E., and Owen, D.B. (1980). Tables for Normal Tolerance
Limits. Sampling Plans, and Screening. New York: Marcel Dekker, Inc.
Quality Assurance Project Plan for the Window Replacement Pilot
of the Renovation and Remodeling Lead Exposure Study (September 30,1993). Prepared
by Battelle and Midwest Research Institute for the U.S. EPA.
SAS Institute Inc., SAS® Technical Report p-229, SAS/STAT®
Software: Changes and Enhancements. Release 6.07. Gary, NC: SAS Institute Inc., 1992,
pp. 287-368.
D-12
-------
50272-101
REPORT DOCUMENTATION
PAGE
1. REPORT NO.
EPA 747-R-96-008
3. Recipient's Accession No.
4. Title and Subtitle
Lead Exposure Associated with Renovation and Remodeling Activities: Environmental Field
Sampling Study Volume II: Appendices
5. Report Date
May 1997
6.
7. Author(s)
Menkedick, J.R., Menton, R.G., Constant, P., Lordo, R.A., Strauss, W.J.
8. Performing Organization Rept. No.
9. Performing Organization Name and Address
10. Project/Task/Work Unit No.
Batlelle Memorial Institute
505 King Avenue
Columbus, Ohio 43201-2693
and
Midwest Research Institute
425 Volker Boulevard
Kansas City, Missouri 64110
11. Contract(C) or Grant(G) No.
(C) 68-D5-0008
(G)
12. Sponsoring Organization Name and Address
U.S. Environmental Protection Agency
Office of Pollution Prevention and Toxics
401 M Street, S.W.
Washington, D.C. 20460
1 3. Type of Report & Period Covered
Technical Report - Appendices
14.
15. Supplementary Notes
In addition to the authors listed above, the following key staff members were major contributors to the study: Halsey Boyd, David
Burgoon, Beth Burkhart, Paul Feder, Pam Hartford, Mary Kayser, Steve Naber, Nick Sasso, and Shawn Shumaker of Battelle; and Jack
Balsinger, Derrick Bradley, John Jones, and Gary Wester of MRI.
16. Abstract (Limit 200 words)
The U.S. Environmental Protection Agency, in response to the Residential Lead Based Paint Hazard Reduction Act of 1992 (Title X),
conducted a study of lead exposure associated with renovation and remodeling (R&R) activities. This report presents the results of a
literature review and one of the principle data collection efforts of the study: the Environmental Field Sampling Study (EFSS). The EFSS
collected 90 personal air samples and 556 settled dust samples to assess potential exposure to workers and occupants from selected R&R
activities. Task length average exposures measured by personal air samplers on R&R workers were greater than 100 A/g/m3 for paint
removal, interior demolition, and sawing, and greater than 49 yt/g/m3 for interior surface preparation and central heating system
maintenance/repair. Lead loadings from stainless steel dust collectors were measured as indicators of the amount of lead disturbed and
made available by the R&R activity for exposure to occupants. With the exception of carpet removal and drilling into plaster, all activities
monitored in the EFSS deposited significant amounts of lead, ranging from 218 /jg/ft* for sawing lead-painted plaster to 42,900 //g/ft2 for
paint removal. Other exposure modifiers, as well as sampling methodology issues, are discussed in the report.
17. Document Analysis
c.
Descriptors
Lead-based paint, lead hazards, renovation and remodeling, field study, wipe and vacuum dust-lead sampling, dust-lead, personal
exposure samples, worker certification, dustfall sampling
Identifiers/open-ended Terms
Lead, renovation and remodeling, worker exposure, Title X, dustfall
COSATI Field/Group
18. Availability Statement
Release Unlimited
19. Security Class (This Report)
Unclassified
20. Security Class. (This Page)
Unclassified
21. No. of Pages
135
22. Price
(SeeANSI-239.18)
OPTIONAL FORM 272 (4-77)
(Formerly NTIS-35)
Department of Commerce
------- |