EPA 402-R-08-003
March 21, 1989
RESEARCH TRIANGLE INSTITUTE
March 21, 1989 RTI/4240/02-03D
FINAL DRAFT
(Third Revision)
THE NATIONAL RESIDENTIAL RADON SURVEY
DESIGN REPORT
by
Jane W. Bergsten
Nicholas A. Hclt
Robert M. Lucas
Martha L. Smith
Research Triangle Institute
Research Triangle Park, North Carolina 27709
Prepared for:
SC & A, Incorporated
8200 Riding Ridge Place
McLean, Virginia 22102
POST OFFICE BOX 12194 RESEARCH TRIANGLE PARK, NORTH CAROLINA 27709-2194
-------
-------
TABLE OF CONTENTS
1.0 Survey Design 1-1
1.1 Purpose of the National Residential Radon Survey ..1-1
1.2 Summary of the National Residential Radon Survey Design 1-2
1.3 Definition of Survey Population 1-3
2.0 Sample Design 2-1
2.1 Introduction 2-1
2.2 Optimization Procedure .2-2
2.3 Optimization Undertaken 2-4
2.4 Recommended Design 2-15
2.5 Comparing the Recommended Design with the Minimal Design 2-18
2.6 Procedures for Selecting the Sample 2-24
3.0 Survey Imp! ementation 3-1
3.1 Data Collection Instrument 3-1
3.2 Training Materials 3-4
3.3 Field Staff Selection, Training, and Responsibilities 3-6
3.4 Field Quality Assurance 3-13
4.0 Data Reduction Procedures 4-1
4.1 Survey Monitoring System 4-1
4.2 Questionnaire Processing 4-2
4.3 Data File Preparation .4-4
4.4 Constructing the Statistical Analysis File 4-4
5.0 Quality Assurance Plan 5-1
Appendix A Glossary of Terms Used in Chapter 2 A-l
Appendix B National Residential Radon Survey Questionnaire B-l
Appendix C Control /Screeni ng Form C-l
-------
1.0 SURVEY DESIGN
1.1 Purpose of the National Residential Radon Survey
In the Superfund Amendments and Reauthorizatlon Act of 1986 (SARA, P.L.
99-499), Congress required the Administrator of the Environmental
Protection Agency (EPA) to conduct a national assessment of radon gas
"found in structures where people normally live and work, including
educational institutions." One component of this national assessment is
the National Residential Radon Survey (NRRS), which will investigate radon
concentrations in occupied housing units.
The results of this survey will be used by the EPA to assess the
effectiveness of various radon program strategies. In particular, the EPA
will use the information obtained from the survey to answer the following
questions:
1. What is the frequency distribution of annual average radon
concentration in occupied housing units?
2. What are the relationships between specific housing
construction types and concentration of radon in occupied
housing units?
In addition to characterizing the national distribution, we also plan
to study subsets of the surveyed homes to learn how radon levels vary among
smaller segments of the population. To ensure that we can make
statistically accurate conclusions about the country's geographic regions,
we will stratify our sample by the 10 EPA Regions and select a
representative sample from each. In this way, we will be able to estimate
the radon distribution for each region and also be able to highlight
differences in radon levels across various parts of the country.
-------
Comparisons of other subgroups of the population will also be made.
For example, we will compare the radon distribution of single family homes
with that for homes in multi-unit structures and compare the distribution
for rental units with that for owner-occupied units. In addition, we will
compare radon levels of different floors within the same residence. This
will show, for example, how radon concentrations in the basement relate to
concentrations in other parts of the home.
The data gathered from this survey will also provide information on how
much radon people are actually exposed to in their homes and what the
resulting increased risks of lung cancer are. For example, a high radon
level in a rarely used storage basement is less of a health risk than a
moderate radon level in a living room or bedroom where people spend hours
every day. Although residential exposure is only part of a person's total
exposure, EPA should be able to estimate the fraction of lung cancers that
could be averted by achieving various reductions of radon concentrations in
homes.
1.2 Summary of the National Residential Radon Survey Design
The National Residential Radon Survey will obtain personal interviews
and 12-month measurements of radon concentrations on each level of a
nationwide random sample of approximately 5,000 residences.
The sample will be a three-stage area probability sample of housing
units. A nationwide probability sample of approximately 125 primary
sampling units (PSUs), each consisting of one or more counties or county
equivalents, will be selected, and one or more interviewers will be hired
in each sample PSU. A probability sample of second stage units (SSUs),
1-2
-------
each consisting of a small geographic area, for example, a Census-defined
city block, will be selected within each sample PSU. The sample SSUs will
be visited and the addresses of each housing unit (HU) will be listed. A
probability sample of addresses will be selected from these, listings to
obtain the sample of HUs for the survey.
Each sample home will be visited and screened for survey eligibility.
(The household must have no firm plans to move within the following
12 months to be eligible for the survey. A personal interview will be
attempted in each survey eligible sample home to obtain information about
the structural, heating and ventilation characteristics of the home and to
obtain information about the residents. In addition to demographic
characteristics, we will request information about smoking characteristics
of the residents and about the amount of time each resident spends on the
different levels of the home. Finally, an alpha track detector will be
placed on each level in the home and will be left there for 12 months.
The design calls for maintaining contact with the sample homes during
the 12 month radon measurement period and for the collection of the
detectors at the end of the 12 months, or beforehand 1f the household has
an unexpected move from the sample housing unit.
1.3 Definition of Survey Population
Because the congressional mandate calls for the measurement of radon
concentrations in places "where people normally live and work, including
educational institutions," a decision must be made about how each different
type of living place should be handled. Our proposal assumed that the
National Residential Survey would cover only the household population, that
1-3
-------
is, the population living in housing units. However, we will identify the
remaining segments of the population living in other types of quarters and
recommend a procedure for handling each. The possible options for handling
these special groups are:
Include the group in the National Residential Radon Survey
Include the group in the school survey
Include the group in the place of work survey
Include the group in a special survey
Exclude the group altogether.
The Census Bureau classifies the population into the following categories.
The household population:
• Persons living in housing units (comprising about 97.5 percent of
the nation's population in 1980.)
—A housing unit is typically a house, apartment, or mobile home,
but may also be one or more rooms occupied as separate living
quarters. The latter are quarters in which the occupants live and
eat separately from others in the building and which have either
direct access to the outside or access throught a common hall.
The nonhousehold population:
• Inmates of institutions (comprising about 1.1 percent of the
nation's population in 1980.)
—Inmates of mental institutions,
such as mental hospitals, psychiatric wards in general
hospitals or veteran's hospitals, alcohol or drug treatment
centers, etc.
—Inmates of homes for the aged,
such as nursing homes, county homes for the aged and
dependent, etc.
— Inmates of correctional institutions,
such as reformatories, local jails, and work houses
— Inmates of other institutions,
such as hospitals or wards for tuberculosis, homes, schools,
hospitals, or wards for the mentally or physically
handicapped, orphanages, residential treatment centers for
emotionally disturbed children, homes for unwed mothers, etc.
• Other persons in group quarters (comprising about 1.4 percent of
the population of the nation 1n 1980.)
—Military personnel living in military barracks or on ships.
(Residents of housing units on military bases are counted as
part of the household population.)
1-4
-------
—Residents of college dormitories, fraternity or sorority houses,
or rooming houses exclusively for college students, provided
there are 10 or more unrelated students or 9 or more unrelated
to the resident who operates the place.
—Residents of rooming houses with 10 or more unrelated
persons or 9 or more unrelated to the resident in charge.
Also included in this category are people living in hotels,
motels, Y's and residential clubs who have no permanent
resident elsewhere.
—Persons living in other group quarters such as convents,
monasteries, halfway houses, communes, low-cost transient
quarters. This category also includes crews of commercial
ships, institutional staff residing in group quarters, persons
residing in parks, campsites, etc.
Our recommendation is that the National Residential Radon Survey
exclude a) housing units that are located on military bases, b) all
institutional residences of the institutionalized population, and c) all
"other" types of living quarters. The target population for the NRRS would
therefore be a) all housing units that are continously occupied for 12
months that are not located on military installations and b) the permanent
residents of these housing units.
Our reasons for excluding certain groups from the National Residential
Radon Survey and our recommendations for handling these groups are given
below:
1. Housing units and other places of residence on military bases or
installations; We recommend the exclusion of these groups from
the NRRS because of difficulty of access. We do not recommend
that they be covered in any other EPA sponsored survey, for the
same reason. EPA should, however, provide the Department of
Defense (DOD) with copies of survey design reports and survey
materials so that DOD can implement comparable surveys, if they so
choose.
1-5
-------
2. Inmates of institutions; We recommend the exclusion of the
institutionalized population from the NRRS for two reasons,
a) because the need for separate sampling and data collection
procedures would make this a very costly and time consuming
addition, and b) because of access difficulty. Typically,
whenever the institutionalized population of the nation is
included in a survey research project, it is covered using
procedures that are separate from those used for the household
population survey, for these same two reasons. If EPA does wish
to measure the residential exposure of the institutionalized, we
recommend that a survey covering this population segment be made a
component of the survey of places of work. This would be a
realistic supplement to the latter survey, because institutions
are places of work that would be sampled and measured in that
survey.
3. Other persons living in group quarters; We recommend the
exclusion of this group from the NRRS. As with the
institutionalized population, including this group would require
different sampling and data collection procedures and would be
costly and time consuming.
—Students; We do recommend that residents of college
dormitories, fraternities or sorority houses and rooming
houses exclusively for college students be covered in the
survey of schools. Coverage of these types of living
quarters for all types of residential schools (excluding the
institutionalized discussed above) should be relatively
1-6
-------
easily fit into the school survey design. The students'
school exposure and residential exposure could be measured
using the same sample schools that were selected from a
stratum of schools having residential students.
—Military: We recommend excluding all military, as
explained above.
—Residents of large rooming houses, monasteries, transient
quarters, and persons living in parks, on boats, etc.; We
recommend not obtaining residential radon exposure for this
very small group of people. Their residences are
sufficiently different to make sampling and data collection
•ore difficult. They are also different enough to call for
analysis as a separate group, thus requiring substantial
oversampUng and the associated increase in cost.
1-7
-------
2.0 SAMPLE DESIGN
2.1 Introduction
The general goals of the survey, which were described in Chapter 1,
need to be reformulated into explicit statistical objectives for the key
population parameters. The statistical objectives, including the
associated precision constraints, can then be put in priority order and
used as the basis for developing a sample design. EPA reformulated and
prioritized the study objectives as follows:
(1) First priority; The survey should provide a scientifically sound
estimate of the frequency distribution of annual-average radon
concentrations in occupied residences nationwide.
• Precision constraint; The national estimate of the percent
of homes with radon concentration over 10 pCi/L should have
a relative standard error of no more than 0.5, if the
estimate is in the neighborhood of 0.5 percent.
(2) Second priority; The survey should provide estimates of the
frequency distributions of annual-average radon concentrations in
occupied residences for subgroups such as the 10 EPA Regions and
for subgroups defined along other lines, such as whether the home
is rented or is owner-occupied. The survey should also permit us
to assess the relationship between indoor radon concentration and
house construction and heating, ventilation, and air conditioning
(HVAC) characteristics.
• Precision constraint; The estimate of the percent of homes
with radon concentration over 4 pCi/L, for an EPA Region,
should have a relative standard error of no more than 0.5,
if the estimate is 1n the neighborhood of 7 percent.
These two precision constraints will be used in developing an optimal
sample design. Then, assuming the implementation of the optimal sample
design, the likely precision of other domain estimates will be assessed.
The setting of expected percentages at 0.5 percent of homes with radon
levels over 10 pCi/L and 7 percent with radon levels over 4 pCi/L was based
on work by Nero, et al.
INero, A.V., Schwehr, M.B., Nazaroff, W.W., Revgon, K.L. (1986).
Distribution of Airborne Radon-222 Concentrations in U.S. Homes. Science,
Vol. 234, pp. 992-997.
-------
2.2 Optimization Procedure
The optimization procedure that was used develops an optimal sample
design that satisfies the specified precision constraints for minimal cost.
Two models had to be specified for use in the optimization, a variance
model and a cost model. A three stage area probability sample of
residences, which was described in Chapter 1, is assumed for each model.
The variance model for this three stage design is given by the
following equation:
V{p) = V{p) [RH01]/n(l) + V{p}(l-[RH01])[RH02]/n(l)n(2)
+ V{p}(l-[RH01] + [RH01][RH02] - [RH02])/n(l)n(2)n(3) (1)
where V{p) = p(l-p), the population variance for the proportion p,
RH01 = the intracluster correlation among secondary sampling units
(SSUs) within primary sampling units (PSUs).
RH02 = the intracluster correlation among housing units (HUs)
within SSUs.
n(l) = the number of PSUs selected into the sample.
n(2) = the average number of SSUs selected per PSU.
n(3) = the average number of HUs selected per SSU.
The variance model in equation (1) is appropriate for an unstratified
design. For a stratified design, which will be employed for this survey,
we would use instead:
V{p] = E W(h)2V{p,h),
/\ •*
where V{p,h} is given by equation (1) and is the variance of p in stratum
h, W(h) is the stratum weight, and the sum is taken over all strata.
2-2
-------
Precision constraints can be expressed as upper bounds on the relative
standard error of estimated parameters. The relative standard error (RSE)
^
of p is given by the equation
RSE(p) = sqr(V{p})/(p),
where sqr indicates the square root. Multiple precision constraints can be
imposed by establishing upper bounds for acceptable relative standard
errors for total population estimates and for domain estimates.
The cost model for the three stage sample design is given by the
following equation:
Total cost - CO + Cl*n(l) + C2*n(l)n(2) + C3*n(l)n(2)n(3), (2)
where CO « fixed cost, that is costs that are not affected by the size
or distribution of the sample,
Cl • cost associated with the number of PSUs in the sample, on a
cost per PSU basis,
C2 « cost associated with the number of SSUs in the sample, on a
cost per SSU basis,
C3 • cost associated with the number of HUs in the sample, on a
cost per HU basis.
The values of the cost components used in developing the sample designs are
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71.
The cost cowponents were developed by partitioning the total estimated cost
of conducting all phases of the National Residential Radon Survey into
subparts and assigning each part either to the fixed cost component or to
one of the three stages of the sample. The total cost was made up of
2-3
-------
a) all costs for Phase I of the survey, which includes design, data
collection, and detector placement, b) all costs for Phase II of the
survey, which includes panel maintenance, detector retrieval, and data
analysis, and c) $300,000 for the purchase and reading of the radon
detectors. (The $300,000 cost for detectors was divided by 5,000, which
was the number of responding sample HUs that was budgeted, and assigned to
C3 as $60 per participating HU.)
The sample designs and allocations discussed in section 2.3 were
2
determined using the techniques described in Chromy 1981.
2.3 Optimizations Undertaken
In our attempt to develop a sample design that will produce the most
precise survey estimates within the allocated budget, we will proceed in
a stepwise fashion, examining for each of the following three cases the
effect of meeting two different precision constraints:
• Case 1: We assume that the percentage of residences with radon
concentration above 4 pCi/L is the same for each of the
10 EPA regions and that the percentage of residences
with radon concentration above 10 pC1/L is the same for
each of the 10 EPA regions.
• Case 2: We assume that the percentage of homes above each of the
two cut-off points varies across EPA regions.
• Case 3: We assume that the percentage of homes above each of the
two cut-off points varies across EPA regions and that
they can also vary across major geographic areas within
an EPA region, as well.
The two precision constraints, which were specified in Section 2.1, are:
2Chromy, J.R. (1981). Variance Estimators for a Sequential Sample
Selection Procedure. Current Topics 1n Survey Sampling, edited by
Krewski, D., R. Platek, and J.N.K. Rao, Academic Press, New York.
2-4
-------
(1) The national estimate of the percent of homes with radon
concentration over 10 pCi/L should have a relative standard
error of no more than 0.5, if the estimate is in the
neighborhood of 0.5 percent.
(2) The estimate of the percent of homes with radon concentration
over 4 pCi/L, for an EPA Region, should have a relative
standard error of no more than 0.5, if the estimate is in the
neighborhood of 7 percent.
In order to develop the sample allocations shown in the exhibits
that are presented, we first set up a simple model for a three stage
stratified sample design, which was described in Chapter 1. For this
hierarchical design, our primary sampling units (PSUs) were defined as
counties, county equivalents, or combinations of these entities. The
second stage units (SSUs), to be selected within the selected PSUs, were
defined as smaller identifiable compact geographic areas, such as census-
defined blocks or enumeration districts (EDs). The third stage units, to
be selected within sample SSUs, were the individual housing units (HUs),
a random portion of which will be included in the National Residential
Radon Survey.
Three sets of assumptions were made in producing the calculations
presented in the exhibits provided in this report. The first set
consists of the partitioning of the survey costs into the four cost
components, CO, Cl, C2, and C3, which were described in the previous
section.
The second set of assumptions that were made relate to the
homogeneity of the radon characteristic of interest within PSUs and SSUs.
Two radon characteristics were considered, the proportion of housing
units with radon readings greater than 4 pCi/L, and the proportion of
housing units with radon readings greater than 10 pCi/L. All of the
calculations presented were made assuming that RH01, which is a measure
2-5
-------
of the intraclass correlation of the radon characteristic "between" SSUs
within PSUs, was equal to .05. RH02, which Is a measure of the
correlation "between" HUs within SSUs, was assumed to be .10. These
assumptions were made after investigating the homogeneity of radon
concentration among the homes measured in the EPA/State radon surveys.
(Note that the intraclass correlation is a measure of homogeneity within
clusters. For example, if the HUs within each SSU were identical on a
given variable, but different from the HUs in other SSUs, the intraclass
correlation would be equal to 1. If the HUs within an SSU are no more
alike than HUs In all the SSUs combined, the intraclass correlation would
be 0. We are assuming some but not perfect homogeneity among the HUs in
an SSU and among SSUs within a PSU.)
The third set of assumptions involves the distribution of radon
levels across strata, which were defined as the 10 EPA Regions. For
Case 1 we assuaed that 1n each of the 10 strata, 7 percent of the housing
units will have radon readings over 4 pC1/L and 0.5 percent will have
radon readings over 10 pC1/L. For Case 2, these two percentages are
allowed to vary fro« stratum to stratum. For Case 3, the two percentages
are allowed to vary from substratum to substratum as well as from stratum
to stratua.
Because of the great number of exhibits that will be presented 1n
this Chapter, we have chosen to place them all at the end of the Chapter.
A description of each exhibit Is provided 1n Figure 1, which immediately
precedes the exhibits, to give the reader a quick reference for the
exhibit comparisons.
2-6
-------
2.3.1 Case 1
For Case 1, we assumed that in each of the 10 strata, each
representing one of the 10 EPA regions, 7 percent of the homes will have
radon readings over 4 pCi/L and 0.5 percent will have readings over 10
pCi/L.
Exhibit 1 shows the sample allocation when the only restriction
placed on the sample is precision constraint (1), that a national
estimate of the magnitude of 0.5 percent have a relative standard error
no larger than 0.5. This would mean that the actual standard error would
be no larger than (0.5)(0.5 percent) = 0.25 percentage points. To obtain
this level of precision, the optimal design calls for a total of 89 PSUs,
398 SSUs, and 1,879 HUs.
Notice that in computing the optimum sample allocation, the cost
components and intraclass correlations used are those described above and
are the same as those used for all of the other optimizations presented.
Eleven estimates are described, the national estimate of the
proportion of homes with radon concentration over 10 pCi/L and, for each
of the 10 EPA Regions, an estimate of the proportion of homes with radon
concentrations over 4 pC1/L. Both variables, the proportion over 10
pC1/L and the proportion over 4 pC1/L, are set uniformly across the 10
strata, defined by EPA Region, at .005 for the former and .07 for the
latter.
The exhibit column headed "Desired Maximum RSE" shows the precision
requirements that we used in the calculations. For Exhibit 1 the only
precision requirement that we used 1n the optimization procedure was that
a national estimate of .005 have a relative standard error of 0.5, which
is shown by the 0.500 entry in this column for the U.S. Total. The other
2-7
-------
entries in this column were set to 9.000, a number so large that the
sample allocation would not be affected by regional constraints. The
actual relative standard errors that should result from the sample
allocation shown in the lower portion of the table are shown in the
column headed "Achieved RSE." For the national estimate of 0.5 percent,
the relative standard error will be about 0.500, as shown in the top
entry of this column. For regional estimates of about 7 percent, the
relative standard errors range from a low of .298 for EPA Region 5 to a
high of .728 for EPA Region 8.
Notice that the percentage distributions of the population of HUs
and of the sample of HUs are the same, reflecting the fact that an equal
probability design is optimal under the set of assumptions employed in
the calculations for Exhibit 1. An equal probability design results in a
proportional allocation of the sample across strata. For the cost
components assumed, the cost of the survey using this allocation would be
about $1,717,000.
Exhibit 2 reflects the setting of precision requirements on regional
estimates as well as on the national estimates. If for each of the 10
EPA regions we require the estimated proportion of homes with radon
measurements over 4 pCi/L to have a relative standard error of no more
than 0.5, we would need about 93 PSUs, 415 SSUs, and 1,959 HUs. This
allocation increases the sample size over that shown in Exhibit 1 for
each of 4 strata, identified as Regions 1, 7, 8 and 10, all of which had
relative standard errors larger than 0.5 for the first sample allocation.
The somewhat larger sample shown in Exhibit 2 also has a somewhat higher
estimated total cost, about $1,756,000 as compared to $1,717,000 for the
sample described in Exhibit 1.
2-8
-------
If the only purpose of the survey were to generate 11 estimates
meeting the stated precision, the sample described in Exhibit 2 would be
sufficient. A larger sample size was invisioned, however, because
estimates for other domains were also desired, even though precision
constraints for these other estimates were not explicitly incorporated
into the optimization procedure. Exhibit 3 shows the sample allocation
described in Exhibit 2, expanded to a sample size of 5,000 HUs, which was
the size that was specified in EPA's Request for Proposal and the size
used for estimating costs for the survey. Keeping the same relative
allocation of housing units to strata, as was shown in Exhibit 2, each
stratum allocation was Increased by a constant factor to force the total
number of HUs to 5,000. Similarly, each stratum allocation of segments
was increased by a constant factor to force the total number of segments
to 1,000 and each stratum allocation of PSUs was increased by a constant
factor to Increase the total number of PSUs to 125. The total estimated
cost now equals that budgeted for the all phases of survey, namely about
$2,268,000. This type of adjustment will be made for each of the three
cases examined, permitting easy comparisons of the different sample
allocations.
2.3.2 Case 2
For Case 2, we continue to assume that, nationwide, the proportion
of homes with radon concentration over 4 pC1/L is 7 percent and the
percent over 10 pC1/L Is 0.5 percent, but we allow these percentages to
vary from region to region. Using results of short term measurements
taken 1n the EPA/State Radon Surveys that have been conducted and also
long tern measurements taken by firms that produce alpha track detectors,
2-9
-------
we classified each region as High, Medium or Low, and assigned
approximate percentages to each category. These percentages were then
adjusted so that they would produce national estimates of 7 percent and
0.5 percent for the two radon levels, as we have assumed. We arrived at
the following set of assumptions:
Assumed percent Assumed percent
Area over 4 pCi/L over 10 pCi/L
USA
"High" strata
"Medium" strata
"Low" strata
7
13
7
1
0.5
1.0
0.2
0.1
Exhibit 4 shows the optimal sample allocation when the only
constraint placed on the sample is that for the national estimate of the
percentage of homes having a radon concentration over 10 pCi/L. Assuming
the estimate is approximately 0.5 percent, the relative standard error
should be no larger than 0.5. This constraint can be satisfied, given
the set of assumptions for Case 2, with a sample of 70 PSUs, 313 SSUs,
and 1,480 HUs, and a total price tag of $1,525,000. By comparing the two
columns "Pop % of HUs" and "Samp % of HUs" you can see that the sample is
now more highly concentrated than the population in regions classified as
having high radon potential.
Exhibit 5 shows the optimal sample allocation when constraints are
placed on regional estimates as well as on the national estimate. The
additional constraints that were set are as follows:
2-10
-------
• For a regional estimate in the neighborhood of 7 to 13 percent of
the homes having radon concentrations over 4 pCi/L, the relative
standard error should be no more than 0.5.
• For a regional estimate of 1 percent of the homes having radon
concentrations over 4 pCi/L, the relative standard error should be
no more than 2.
The constraint on regional estimates of small percentages was
relaxed for two reasons. First, from a decision-making standpoint, one
does not need as great relative precision for estimates of very small
percentages as for estimates of larger percentages. For example, the
development of action programs would dictate that when estimating the
percent of homes having very high radon concentrations, it is more
important that an estimate of 13 percent have a standard error of no more
than 6.5 percentage points than it is for an estimate of 1 percent to
have a standard error of no more than 0.5 percentage points. One would
not, therefore, require that the relative standard error be no greater
than 0.5 for estimates of magnitude 1 percent as well as for estimates
near 13 percent.
The second reason for relaxing the precision constraints for small
estimates is that, in order to achieve a very small relative standard
error for a small estimated percentage, a very large proportion of the
sample must be concentrated in strata classified as having a low
residential radon potential. Because such strata have smaller variances,
increasing the sample there reduces the sample size in strata with larger
variances causing the precision of the two major national estimates to be
reduced.
To achieve the precision constraints listed above for the national
estimate of the percent of homes with radon concentration above 10 pC1/L
2-11
-------
and for the regional estimates of the percent of homes with radon
concentration above 4 pCi/L, we would need 77 PSUs, 344 SSUs, and 1,625
HUs, for a total survey cost of $1,595,000. (Exhibit 5.)
By keeping the same relative sample allocation across strata but
expanding the sample to include 125 PSUs, 1,000 SSUs, and 5,000 HUs, we
have a survey that is expected to cost $2,688,000, the amount budgeted.
(See Exhibit 6.)
Our attention so far has been entirely focused on estimates of two
characteristics, the percentage of homes with radon concentration over 4
pCi/L and percentage with concentrations over 10 pCi/L. There are of
course other characteristics that will be estimated from the survey data.
For estimating characteristics having variances that do not parallel
those assumed for the two radon variables on which we are optimizing, our
"optimal" design will not produce the greatest precision. For
characteristics whose variance is the same from stratum to stratum, for
example, a proportional allocation, similar to that shown in Exhibit 1,
would produce the greatest precision. One of the trade-offs associated
with using a design that is optimum for certain estimates is that one may
well substantially decrease the precision of other estimates.
In order to assess the potential loss in precision for estimates for
which a proportional allocation of the sample to our defined strata is
optimal, we have determined the relative standard error for such
estimates, assuming we used the design shown in Exhibit 6. Exhibit 7
shows the sample allocation described in Exhibit 6 but presents for the
relative standard errors those that we would obtain 1f the characteristic
being estimated actually had the same distribution from stratum to
2-12
-------
stratum. As can be seen in the column headed "Achieved RSE," a national
estimate of 0.5 percent would have a relative standard error of 0.431 and
all regional level estimates of the magnitude of 7 percent would have a
relative standard error no larger than 0.455, if the population
proportions were in fact as given in the second and third columns, namely
7 percent and 0.5 percent.
Another way of looking at the information in Exhibit 7 is to see
what the precision of our most important estimates would be if we were
all wrong in the way we classified regions into high, medium and low
radon potential. If we based our design on the assumption that these
classifications were correct and later found that they were incorrect,
that in fact the values were the same in all strata, we would,
nevertheless, achieve the desired precision. The national estimate of
0.5 percent would have a relative standard error of 0.431 and the
regional estimates of 7 percent would have a relative standard error of
no more than 0.455. (Exhibit 7.)
2.3.3 Case 3
For Case 3 we continue to assume that, nationwide, the proportion of
homes with radon concentration over 4 pCi/L is 7.0 percent and that the
proportion of homes with radon concentration over 10 pCi/L is 0.5
percent, but we allow these percentages to vary, not only from one EPA
region to another, but also from one state to another within a single EPA
region. Using the results of short term measurements taken in EPA/State
Radon Surveys and also the results of long term measurements taken by
commercial firms that produce alpha track detectors, we classified states
or major portions of states within each EPA region into High, Medium, and
2-13
-------
Low radon-potential substrata. These classifications are shown in
Exhibit 8.
Exhibit 9 shows the optimal sample design for achieving the relative
standard error listed in the "Desired Maximum RSE" column and shows the
actual relative standard errors that would result from the design in the
"Achieved RSE" column. The allocations to the substrata are presented on
the second page of Exhibit 9. The design calls for 71 PSUs, 317 SSUs and
1,498 HUs.
Exhibit 10 Inflates the allocation shown in Exhibit 9 so that we have
the proposed 125 PSUs, 1,000 SSUs, and 5,000 HUs. Note that for both
Exhibit 9 and Exhibit 10, the precision constraints were set at 0.5 for
the maximum acceptable relative standard error for the national estimate
of 0.5 percent of the homes with readings over 10 pCi/L. The constraints
on the precision of the regional level estimates for the percent of homes
with readings over 4 pCi/L were a maximum relative standard error of 0.5
for estimates of 5.0 percent or larger, a maximum relative standard error
of 1.0 for estimates of 2.0 percent or more but less than 5.0 percent and
a maximum relative standard error of 2.0 for estimated percentages less
than 2.0. With the Exhibit 10 design, the national estimate of 0.5
percent has a relative standard error of 0.325. The precision of the
regional estimates varies from a low relative standard error of 0.133 for
an estimate in the neighborhood of 11.3 percent for Region 5 to a high
of 0.718 for the relative standard error of an estimate of 1.4 percent
for Region 9.
Exhibit 11 shows the relative standard errors that would be achieved
if we used the allocation described in Exhibit 10, when in reality the
2-14
-------
substrata did not vary on residential radon concentration as we have
assumed, but instead were constant across substrata. The requirement for
the national estimate is still met, with a relative standard error of
0.415 for an estimate of 0.5 percent. The precision requirements for the
regional estimates are met in all regions except Region 10, which would
yield a relative standard error of 0.580 for an estimate of 7.0 percent.
We feel that, even if the substrata values assumed in the
optimization are not exactly correct, the worst-case actual occurrence
would be that assumed in Exhibit 11. We are not, therefore, presenting a
table showing the relative standard errors if the substrata values were
the reverse of what we assumed, that is, were actually high in substrata
we called low and low in substrata that we called high.
2.4 Recommended Design
The sample design that we recommend is that presented in Exhibit 12.
This is essentially the same design as shown 1n Exhibit 10, with the
following adjustments:
• The number of PSUs in a substratum was rounded to a whole number
and set at a minimum of 2 PSUs.
• The numbers of SSUs and HUs 1n a stratum were also rounded to
whole numbers. The number of SSUs per PSU was set at 8, and the
number of participating HUs per SSU was set at 5, for staffing and
work-load reasons. (The actual number of HUs that will be
selected into the sample will of course be enough larger than
5 per SSU to yield an average of 5 participating HUs per SSU.)
The distribution of the population of HUs and the proposed sample of
HUs is shown on the second page of Exhibit 12. The Low substratum of
Region 5 will have the lowest sampling rate, with only 1.6 percent of the
sample coming from that substratum, which contains 3.7 percent of the
2-15
-------
population of HUs. The High substratum of Region 9 will be sampled at the
highest rate, with 1.6 percent of the sample coming from this substratum,
which contains only 0.5 percent of the population of HUs. The ratio of the
lowest to the highest sampling rate is about 7.8. Note that, if the
substratum values are as we have assumed here, we will identify about 36
HUs with radon levels over 10 pCi/L using the recommended compared to only
about 25 using an equal probability design.
Exhibit 13 contains additional values for the expected precision of
estimates of the two key parameters, percent >4 pCi/L (Column 5) and
percent >10 pCi/L (Column 6). The precision of domain estimates is given
for domains ranging in size from 50 percent to 5 percent of the HUs in the
nation. This generalized precision table can be used to evaluate the
anticipated precision of estimates for domains not explicitly included in
the design criteria, for example, domains based on HU construction or HVAC
characteristics.
Row 1 of the table represents the precision for national estimates
(100 percent domain). Rows 2 and 3 give the range of the expected
precision for 50 percent domain estimates, reflecting two extremes of
clustering. Row 2 assumes that half of the HUs in each sample SSU fall in
a particular domain. This is the "best case" scenario. Row 3 assumes all
of the HUs in half of the sample SSUs are in the domain and none of the HUs
in the remaining half of the sample SSUs are in that domain. These two
extremes represent a lower bound and a likely upperbound for the RSE of an
estimate for a 50 percent domain that covers all of the PSUs. All domains
described in Exhibit 13 are assumed to be represented 1n all 125 sample
PSUs, which would be typical of domains defined along other than geographic
lines.
2-16
-------
Rows 4 and 5 give the likely precision of estimates for domains
comprising 25 percent of the population. Rows 6 and 7 and Rows 8 and 9
provide the same information for domains representing 10 percent and 5
percent domains of the population of HUs, respectively.
Inspection of the table reveals that good precision (RSE<0.25) can be
achieved for estimating the percent >4 pC1/L for domains as small as 5
percent. Acceptable precision (RSEs) can be achieved for estimating the
percent >10 pCi/L for domains as small as 25 percent.
Exhibit 14 presents precision estimates in a format identical to that
used in Exhibit 13, but with one important modification. In Exhibit 14 we
assume that the key parameters were equal in all of the design strata.
This exhibit can be used to evaluate how "robust" the recommended design is
to errors in our assumed radon distribution.
As one would expect* the precision of the domain estimates in
Exhibit 14 are inferior to those in Exhibit 13. However, the precision
constraint for the national estimate of percent >10 pCi/L is still attained
and acceptable precision (RSE < 0.50) is still attained for the percent
>4 pCi/L for even the 5 percent domain.
In addition to the precision constraints used in developing the
optimal sample design, EPA has a number of other domain estimates which
they hope will have small relative standard errors. Some of these domains
are described in Exhibit 15. If we wished to estimate the proportion of
single family homes with radon concentration over 4 pCi/L, for example, we
would expect that an estimate of 7 percent would hae a relative standard
error of 0.09, as can be seen in the top row of Exhibit 15. In determining
the relative standard error that we would expect, we had to make some
2-17
-------
assumption about the geographic dispersion of single family homes. We
found that about 65 percent of housing units in the population are single
family homes and we expect about that same percentage of single family
homes in the sample (column 1). We also expect that they will be
distributed over 100 percent of the 125 PSUs (column 2), 80 percent of the
SSUs (column 3), and 80 percent of the HUs within the SSUs in which they
are located. Note that all of the entries in column 5 of Exhibit 5 are
less than 0.5. This means that for each of the subgroups described,
estimates in the neighborhood of 7 percent would have relative standard
errors that we expect to be less than 0.5.
The recommended sample design should, therefore, yield estimates of
the percent of homes having high radon concentration with precision
sufficient to satisfy EPA's needs. We expect both national estimates and
estimates for important domains to have greater than the minimum required
precision.
2.5 Comparing the Recommended Design with the Minimal Design
The recommended design shown in Exhibit 12 satisfies a) EPA's two
explicitly stated precision requirements and b) the 5,000 sample size
specified in EPA's Request for Proposal. Recall that the recommended
design was adapted from the design shown in Exhibit 9 by first expanding
the sample size to 5,000, using certain guidelines Imposed by practical
field work constraints, then incorporating certain statistical
restrictions, such as rounding the number of PSUs to be selected from a
stratum to a whole number and requiring at least two PSUs to be selected
per stratum. Exhibit 16 shows the Exhibit 9 sample design adjusted In a
similar manner, but to a sample size of about 2,000, rather than a sample
size of 5,000.
2-18
-------
As can be seen in the last column of the top portion of Exhibit 16,
for a sample of 2,130 HUs the national estimate of 0.5 percent of the homes
having radon concentration above 10 pCi/L, is subject to an expected
relative standard error of 0.49, which meets the precision requirement that
the relative standard error be no more than 0.50. Similarly, for regional
estimates of the percent of homes with radon levels over 4 pCi/L, the
constraints that were used for the recommended design, and listed in
section 2.3.3, are satisfied. These constraints were: a maximum relative
standard error of 0.5 for estimates of 5.0 percent or larger, a maximum
relative standard error of 1.0 for estimates of 2.0 percent or more but
less than 5.0 percent, and a maximum relative standard error of 2.0 for
estimated percentages less than 2.0.
A sample of 5,000 obviously costs considerably more to implement than
a sample of 2,130. We will examine here some of the benefits obtained from
using the recommended design rather than the minimal design.
The most apparent advantage a sample of 5,000, rather than a sample
of 2,130, is that we will pick up about 2 1/2 times as many HUs with high
radon levels. For example, we expect about 36 sample homes with radon
levels over 10 pCI/L using the recommended design compared to only about 15
using the Minimal design, providing us with a greater ability to
characterise high radon level homes.
Even though the minimal sample design, shown in Exhibit 16, satisfies
the minimum precision constraints that were set, estimates using this
design will not be as precise as those based on the 5,000 size sample
design. The relative standard errors contained in Exhibit 17 can be
compared to those 1n Exhibit 13 to see what differences in precision are
2-19
-------
likely to occur as a result of the difference in sample size. Comparing
the values in the fifth columns of the two exhibits reveals that the
expected relative standard errors of estimated percentages of homes with
radon concentration over 4 pCi/L are 40 to 60 percent higher for estimates
based on a sample size of 2,130 than they are for estimates based on a
sample size of 5,000. The same relationship holds when comparing the
precision of estimates of the proportion of homes with concentrations over
10 pCi/L, the values found in the sixth columns of the two exhibits.
Comparing the expected precision of the estimates described in
Exhibit 18 with those shown in Exhibit 15 shows again that a sample of
2,130 HUs can be expected to produce estimates having relative standard
errors in the neighborhood of 40 to 60 percent higher than those obtained
from a sample of 5,000.
In general, we can conclude that the confidence intervals for domain
estimates would be about one and one half times as large using the smaller,
minimal sample design than they would be if the recommended design were
used. Increase in precision is one of the major advantages derived from an
increase in sample size. This type of advantage will also occur when
carrying out more complicated statistical analyses of the survey data. The
larger sample will of course support more detailed data analysis than the
2,130 sample will.
In addition to comparing the precision of estimates made from each of
the two designs, we can also investigate their respective ability to detect
differences that exist between different subgroups in the population.
Power curves, which will enable us to make these comparisons, are presented
in Figures 2, 3, and 4.
2-20
-------
To help the reader better understand the concept of power, let us
consider three equal size population subgroups that have different
percentages of homes with radon concentrations in excess of 4 pCi/L. Let
us suppose that the percentage of homes above 4 pC1/L is 7 percent for
group A, 8 percent for group B, and 93 percent for group C. What size
samples must we have to be able to show that the percentage of homes with
radon levels over 4 pCi/L is smaller in group A than it is in group B?
What size samples must we have to be able to show that the percentage of
such high radon homes is smaller in group A than it is in group C?
Intuitively, we would guess that we would need much larger samples to
demonstrate the group A and group B relationship than we would need to
demonstrate the group A and group C relationship. Our ability to detect a
difference between two population subgroups is in fact related to: a) the
magnitude of the difference between the two population subgroups, b) the
variability of the characteristic within each of the two population
subgroups, and c) the size and design of the samples from the two
population subgroups. The power tables described in this section
demonstrate the effects of two different sample sizes, 5,000 and 2,130, on
our ability to detect differences that exist in population subgroups of
different sizes. The type of question that the power tables answer is the
following: If two population subgroups differ on the percentage of homes
with radon levels over 4, with the percentage being larger for group 2 than
it is for group 1, what is the likelihood that our survey results will show
that group 2 has a higher percentage of high radon homes than group 1?
The curves in Figure 2 show the power of the surveys to detect a
difference in the residential radon concentrations of two population
2-21
-------
subgroups that do in fact differ from one another. More specifically, the
curves show the probability of our detecting a difference in the predicted
direction if 7 percent of the homes in population subgroup j have radon
levels over 4 pCi/L and only 1 percent of the homes in population subgroup
i have radon levels over 4. To test for such a difference, we would test
the hypothesis that groups i and j have the same proportion of homes with
radon levels over 4 pCi/L, which is called the null hypothesis, against the
alternative hypothesis that a larger proportion of homes in group j than in
group i have radon concentrations over this level. If we use an alpha of
.05, the probability of our detecting a difference between p,- and p.,- would
be that shown by the four curves in Figure 2. If each of two such
subgroups 1s a geographical subpart of the nation, and each contains about
0.05 of the nation's housing units, the sample of 2,130 has only a 40
percent chance of detecting the difference between the groups (the far left
point on the lowest curve). On the other hand, the sample of 5,000 has
about a 64 percent chance. As we follow the two lower curves to the right
to determine the power associated with larger and larger subgroups, we can
see that If each geographic subgroup contains about 40 percent of the
nation's housing units, the chance of detecting the difference becomes
almost a certainty for both of the sample sizes.
The top two curves on the Figure 2 graph show the ability of the two
different sample designs to detect differences between subgroups defined on
other than geographic lines. These are power curves for somewhat of a
"best case" scenario, where instead of comparing two different geographic
subgroups, we ownpare two subgroups, each of which is widely dispersed
across all of the sample segments. The "best case" scenario assumes that
2-22
-------
there is a cancelling out of certain correlation terms, resulting in error
variances equivalent to those of a simple random sample. For small
subgroups of approximately 0.05 of the population, the sample of 2,130 has
only about a 73 percent chance of detecting a difference between the two
subgroups, whereas the sample of 5,000 has about a 96 percent chance. This
can be seen in the far left points of the two top curves in Figure 2. For
larger subgroups, each containing about .20 of the population, the power is
about the same for the two designs. Figure 2 shows, therefore, that the
recommended sample design has a much greater power than the alternative
design to detect a difference between "low" and "medium" radon levels in
small population subgroups.
Figure 3 shows comparable power curves for detecting differences
between medium level and high level groups, the former defined as 7 percent
of homes with radon concentrations over 4 pCi/L and the latter as 13
percent of homes over that level. Notice that the curves are less steep
than those shown in Figure 2 and that the differences between the 5,000
sample size design and the 2,130 sample size curves are more pronounced
than they were in Figure 2. Even when comparing two large geographic
groups, each containing about half of the nation's population, the larger
sample size has considerably more power than the smaller sample size,
demonstrating the consistent greater ability of the larger sample to detect
differences between medium and high level radon groups.
The powers to detect a difference between a "low" radon level group
having only 1 percent of the homes over 4 pCi/L and a "high" group having
13 percent of the homes over 4 pC1/L are shown in Figure 4. Again, the
larger sample size consistently shows a greater power of detecting
2-23
-------
differences between even large geographical groups. For nongeographically
defined subgroups (the two top curves), the recommended design is clearly
more powerful than the alternative design for small subgroups, but for
subgroups, each of which is about one quarter or more of the population,
the two designs are about equally powerful.
There are, of course, many other ways in which the recommended sample
design and the minimal sample design can be compared. The comparisons
described in this section simply highlight the greater precision of
estimates provided by the recommended design and the greater ability of the
recommended design to detect existing differences between population
subgroups. The additional cost associated with implementing the larger
design does indeed "buy" some important advantages.
2.6 Procedures for Selecting the Sample
In earlier sections we briefly discussed the manner in which the
three stage area probability sample of HUs would be selected for the
ational Residential Radon Surveys. These procedures will described in
greater detail in this section.
2.6.1 Selecting PSUs
A probability sample of PSUs will be selected within each of
the 22 strata shown in Exhibit 8. We will select the exact number of
PSUs called for in the recommended design described in Exhibit 12.
Using the latest (1987) Market Statistics, Inc. data that provide
estimates of the number of housing units for each county in the nation,
we will assign a current housing unit count to each county or county
equivalent. Those with fewer than 1,000 HUs will be linked with
adjoining counties or county equivalents to form PSUs having a minimum
of 1,000 HUs.
2-24
-------
We will append the following 1980 Census information to each PSU
record:
• SMSA Status: Whether or not the PSU is part of an SMSA.
• Urbanization: Percent of HUs located in areas classified as
urban.
• Heating equipment: Percent of HUs with warm air furnace or heat
pump.
• Heating fuel: Percent of HUs with utility gas.
Using these variables, the PSUs within a stratum will be ordered in a
serpentine fashion, so that PSUs with similar characteristics will be
next to one another on the list. We will then use Chromy's sequential
selection procedure, which was referenced earlier in this chapter, to
select the specified number of PSUs from each stratum, with probability
proportional to the estimated number of HUs in the PSU. Using a
sequential selection procedure on an ordered list produces an effect
similar to the effect that could be derived from a formal stratification
procedure. Such procedures help assure that all aspects of the
population are properly represented in the sample and may provide some
increase In the precision of the sample estimates.
2.6.2 Selecting SSUs
The SSUs within each sample PSU will be made up of Census
defined blocks or enumeration districts (EDs). Units containing fewer
than 25 occupied HUs In the 1980 Census will be combined with nearby
units to for* SSUs having a minimum size of 25 HUs. We will append the
following 1980 Census Information to each SSU record:
• Urbanization: The place size code for the largest place in the
PSU will be Identified. An SSL) that is all or
partly in a place of this size will be classified
as high urban. Other SSUs will be classified as
low urban. (A place is a heavily populated area,
typically, but not always, incorporated as a city
or town. Place size codes of 01, indicating a
population under 200, to 16, indicating a
population of 1,000,000 or more, were assigned to
the 23,000 places by the Census Bureau.)
2-25
-------
• House value: An index of house value will be computed by taking
the weighted average of 1 percent of the house
value for owner occupied HUs and the rent value for
rental HUs.
• Ownership: Percentage of occupied HUs that are owner occupied.
Within each sample PSU, the SSUs will be ordered in a serpentine fashion
using the variables listed above, and Chromy's sequential selection
procedure will be used to select exactly 8 SSUs with probability
proportional to the number of occupied HUs in the SSU.
Each sample SSU will be identified on a map, and a sketch of the SSU
will be prepared. Maps, sketches, listing forms and a copy of RTI's
Field Listing Manual will be sent to experienced field representatives,
who will visit each sample SSU in the PSU. Each sample SSU that contains
150 or fewer HUs will be listed in its entirety. Using explicit
instructions, the field representative will proceed around the SSU in a
prespecified manner listing the addresses of all HUs located within the
SSU boundaries. In order to avoid listing an excessively large number of
HUs, SSUs containing more than 150 HUs may be partitioned into subparts,
one or more of which will be randomly selected for listing.
Field representatives will maintain close contact with RTI's sampling
personnel, so that any discrepancies between the expected number of HUs
and the actual number can be resolved. This close interaction will help
avoid problems of improper field identification of sample SSUs and will
help maintain field listing costs. Field materials will be returned to
RTI and carefully edited, on a flow basis. This will permit us to
identify problems at an early stage so that corrective action can be
taken.
2-26
-------
2.6.3 Selecting HUs
A random systematic sample of a specified number of listed
addresses will be selected from each sample SSU. The sample addresses
will be sent to the field for inclusion in the National Residential Radon
Survey. When a field interviewer visits a sample address, she will
include in the sample all HUs at the sample address and all HUs between
the sample address and the "next listed" address on the SSU listing form.
Using this "half open interval" procedure assures that HUs that were
missed in the listing process will be given their proper chance of being
included in the survey.
Records showing the probability of selection of each sample address
will be maintained so that sampling weights that reflect the sample
inclusion probabilities can be assigned. Such sampling weights are
needed for the generation of unbiased population estimates from the
sample data. All data analysis will be carried out using properly
weighted data, that reflect sampling inclusion probabilities, adjustments
for nonresponse, and the full complexity of the sample design.
2.6.4 Controlling Sample Size
In order to obtain a sample of 5,000 participating HUs, a
much larger sample must be selected. Some sample addresses will yield
more than one survey-eligible HU while other sample addresses will yield
only HUs that are ineligible for the survey, for reasons such as vacancy
or residents having definite plans to move. In addition, some sample
households will refuse to participate, others will participate but fall
to return their detectors, and others will return detectors that prove to
be unreadable. We must estimate the amount of expected loss for each of
2-27
-------
these types of nonresponse in order to determine the number of sample
addresses to be fielded. If we underestimate the expected loss, our
sample will end up smaller than the 5,000 we had hoped for. This
likelihood could be reduced by using somewhat higher estimates of
expected loss, but this approach could bring about higher field costs and
a sample considerably larger than the 5,000 targeted.
We do plan to implement the survey using methods that will provide us
with additional control over the size of the sample. Two SSUs in each
PSU will be randomly identified as "early report" SSUs. The field
interviewers will be instructed to work the sample addresses in these
SSUs first. Evaluation of the early returns will provide us with
additional information on vacancy rates, expectation of moves, response
rates, etc. Using this information, we can make a more informed decision
on the number of sample addresses that should be included in the survey.
We will have selected from the SSUs that were not designated to be
"early report" SSUs a somewhat larger number of addresses than we think
we will need. These sample addresses will be randomly partitioned into a
regular sample and two reserve samples. Each sample address will be
identified and color coded to indicate its classification. Field
interviewers will implement only the regular sample addresses until
instructed to do otherwise. After we have had a chance to evaluate the
results obtained from the early report SSUs, we will decide whether one
or both of the reserve sample of addresses is needed. The field
interviewers will then be provided with Instructions pertaining to the
implementation of the reserve samples. Use of this early report-reserve
sample methodology will provide us with some additional control over the
sample size for a very minimal cost.
2-28
-------
Figure 1. Summary of the Exhibits and Figures Presented
The Exhibits that follow are based on a) three different sets of
assumptions about the distribution of residential radon concentration and
b) the following two types of precision constraints on the survey
estimates:
The estimated percent of homes
concentration above 10 pci/1, for the
have a relative standard erroro7nomore
estimate is in the neighborhood of 0.5 percent.
with residential radon
nation as a whole, should
than0.5 when the
The estimated percent of homes with
concentration above 4 pci/1, for each of
should have a relative standard error of not more than 0.5 when
the estimate is in the neighborhood of 7.0 percent.
residential radon
the 10 EPA Regions,
Exhibits 1-3 are based on the assumption that the distribution
residential radon concentration is the same for all 10 EPA Regions.
of
Exhibit 1
Exhibit 2
Exhibit 3
The minimum-cost sample allocation when only
national level precision constraint is considered.
the
The minimum-cost sample allocation when both the
national level and the regional level precision
constraints are considered.
The Exhibit 2 sample allocation
size of 5,000 housing units.
expanded to a sample
Exhibits 4-7 are based on the assumption that the distribution of
residential radon concentration varies across the 10 EPA Regions.
Exhibit 4
Exhibit 5
Exhibit 6
The minimum-cost sample allocation when only the
national level precision constraint is considered.
The minimum-cost sample allocation when both the
national level and the regional level precision
constraints are considered.
The Exhibit 5 sample allocation
size of 5,000 housing units.
expanded to a sample
2-29
-------
Figure 1 (continued)
Exhibit 7 The expected relative standard errors of estimates based
on the Exhibit 6 sample allocation, if in fact the
distribution of residential radon were actually constant
across the 10 EPA Regions.
Exhibits 8-12 are based on the assumption that the distribution of
residential radon concentration can vary from one geographical area to the
next within each of the 10 EPA Regions.
Exhibit 8
Exhibit 9
Exhibit 10
Exhibit 11
Exhibit 12
The minimum-cost sample allocation when only
national level precision constraint is considered.
the
The minimum-cost sample allocation when both the
national level and the regional level precision
constraints are considered.
The Exhibit 9 sample allocation
size of 5,000 housing units.
expanded to a sample
The expected relative standard errors of estimates based
on the Exhibit 10 sample allocation, 1f in fact the
distribution of residential radon were actually constant
across the different substrata.
The recommended design. This is the Exhibit 10 sample
allocation adjusted as follows:
• The number of PSUs in a substratum is rounded to a
whole number and set at a minimum of 2 PSUs.
• The number of SSUs and HUs were also rounded to
whole numbers and for interviewer work-load reasons
were set at 8 SSUs per PSU and an expected 5
participating HUs per SSU.
Exhibits 13-15 present expected relative standard errors of estimates based
on the recommended sample design described in Exhibit 12.
Exhibit 13 Expected relative standard errors for domain estimates,
when the distribution of residential radon concentration
varies across substrata, as shown in Exhibit 12.
2-30
-------
Figure 1 (continued)
Exhibit 14 Expected relative standard errors for domain estimates,
if in fact the distribution of residential radon were
actually constant across substrata.
Exhibit 15 Examples of relative standard errors for domain
estimates, when the distribution of residential radon
concentration varies across substrata, as shown in
Exhibit 12.
Exhibit 16-18 present expected relative standard errors of estimates based
on the minimal design in Exhibit 9.
Exhibit 16
Exhibit 17
Exhibit 18
The minimal practical design. This
sample allocation adjusted as follows:
is the Exhibit 9
• The number of PSUs in a substratum is rounded to a
whole number and were set at a minimum of 2 PSUs.
• The number of SSUs and HUs were also rounded to
whole numbers and were set at 6 SSUs per PSU and an
expected 5 participating HUs per SSU.
Expected of relative standard error for domain estimates
for the minimal design, when the distribution of
residential radon concentration varies across substrata,
as shown in Exhibit 16.
Examples relative standard error of domain estimates for
the minimal design, when the distribution of residential
radon concentration varies across substrata, as shown in
Exhibit 16.
Figures 2-3 illustrate the power of detecting differences between subgroups
for the recommended design and the minimal practical design.
2-31
-------
Exhibit 1. The Minimum-Cost Sample Allocation When Only the National Level
Precision Constraint is Considered. (Calculations Assume that the
Distribution of Residential Radon Concentration Is the Same For All 10 EPA
Regions.)
Assumed cost components
CO =' $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RHO1 =0.05
Assumed intracluster corr coef. RHO2 = o.io
Domain
U. S. Total
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
Prop>lo
0.00500
Desired
Maximum
RSE
0.500
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
9.000
9.000
9.000
9.000
9.000
9.000
9.000
9.000
9.000
9.000
Achieved
RSE
0.500
0.563
0V399
0.400
0.303
0.298
0.384
0.577
0.728
0.350
0.668
Optimal sample allocation by EPA Region
Region
1
2
3
4
5
6
7
8
9
10
No. of
Cty's
4.7
9.3
9.3
16.2
16.7
10.1
4.5
2.8
12.1
3.3
NO. Of
SEGs
21.0
41.7
41.5
72.4
74.7
45.1
19.9
12.5
54.3
14.9
No. of
HUS
98.9
196.9
195.7
342.1
352.8
212.8
94.1
59.2
256.4
70.2
Pop %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Total
88.9
398.0
1,879.2
100.0
100.0
2-32
-------
Exhibit 2. The Minimum-cost Sample Allocation When Both the National Level
and the Regional Level Precision Constraints are Considered. (Calculations
Assume that the Distribution of Residential Radon Concentration Is the Same
for All 10 EPA Regions.)
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RHO1
Assumed intracluster corr coef. RHO2
Domain
U. S. Total
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
Prop>10
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.05
0.10
Desired
Maximum
RSE
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
Achieved
RSE
0.500
0.500
0.413
0.414
0.313
0.308
0.397
0.500
0.500
0.362
0.500
Optimal sample allocation by EPA Region
Region
1
2
3
4
5
6
7
8
9
10
Total
No. of
Cty's
5.9
8.7
8.7
15.1
15.6
9.4
5.9
5.9
11.4
5.9
No. of
SEGs
26.6
39.0
38.8
67.8
69.9
42.2
26.6
26.6
50.8
26.6
No. of
HUs
125.5
184.3
183.2
320.1
330.2
199.1
125.5
125.5
240.0
125.5
Pop %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp
of HUs
6.4
9.4
9.4
16.3
16.9
10.2
6.4
6.4
12.3
6.4
92.6
414.8
1,958.8
100.0
100.0
2-33
-------
Exhibit 3. The Exhibit 2 Sample Allocation Expanded to a Sample Size of
5,000 Housing Units. (Calculations Assume that the Distribution of
Residential Radon Concentration Is the Same for All 10 EPA Regions.)
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RHO1
Assumed intracluster corr coef. RHO2
Domain
U. S. Total
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
Prop>10
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0:00500
0.05
0.10
Desired
Maximum
RSE
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
0.500
Achieved
RSE
0.372
0.372
0.307
0.308
0.233
0.229
0.295
0.372
0.372
0.269
0.372
Optimal sample allocation by EPA Region
Region
1
2
3
4
5
6
7
8
9
10
No. of
Cty's
8.0
11.8
11.7
20.4
21.1
12.7
8.0
8.0
15.3
8.0
No. Of
SEGs
64.0
94.1
93.5
163.4
168.6
101.7
64.0
64.0
122.5
64.0
No. of
HUs
320.2
470.5
467.6
817.2
842.8
508.3
320.2
320.2
612.6
320.2
Pop %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp
of HUs
6.4
9.4
9.4
16.3
16.9
10.2
6.4
6.4
12.3
6.4
Total
125.0
1,000.0
5,000.0
100.0
100.0
2-34
-------
Exhibit 4. The Minimum-Cost Sample Allocation When Only the National Level
Precision Constraint is Considered. (Calculations Assume that the
Distribution of Residential Radon Concentration Varies across the 10 EPA
Regions.)
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RHO1
Assumed intracluster corr coef. RHO2
Domain
U. S. Total
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
0.070
0.130
0.130
0.010
0.130
0.010
0.070
0.130
0.010
0.070
Prop>10
0.00499
0.00200
0.01000
0.01000
0.00100
0.01000
0.00100
0.00200
0.01000
0.00100
0.00200
0.05
0.10
Desired
Maximum
RSE
0.500
9.000
9.000
9.000
9.000
9.000
9.000
9.000
9.000
9.000
9.000
Achieved
RSE
0.500
0.751
0.253
0.254
1.310
0.189
1.661
0.770
0.462
1.513
0.891
Optimal sample allocation by EPA Region
Region
1
2
3
4
5
6
7
8
9
10
No. of
Cty's
2.6
11.7
11.6
6.4
20.9
4.0
2.5
3.5
4.8
1.9
No. of
SEGS
11.8
52.3
52.0
28.8
93.6
17.9
11.2
15.7
21.6
8.4
No. of
HUs
55.7
246.8
245.3
136.2
442.1
84.7
53.0
74.2
102.1
39.5
Pop %
Of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp
of HUs
3.8
16.7
16.6
9.2
29.9
5.7
3.6
5.0
6.9
2.7
Total
70.0
313.4
1,479.5
100.0
100.0
2-35
-------
Exhibit 5. The Minimum-Cost Sample Allocation When Both the National Level
and the Regional Level Precision Constraints Are Considered. (Calculations
Assume that the Distribution of Residential Radon Concentration Varies
across the 10 EPA Regions.)
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RHO1
Assumed intracluster corr coef. RHO2
Domain
U. S. Total
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
0.070
0.130
0.130
0.010
0.130
0.010
0.070
0.130
0.010
0.070
Prop>10
0.00499
0.00200
0.01000
0.01000
0.00100
0.01000
0.00100
0.00200
0.01000
0.00100
0.00200
0.05
0.10
Desired
Maximum
RSE
0.500
0.500
0.500
0.500
2.000
0.500
2.000
0.500
0.500
2.000
0.500
Achieved
RSE
0.500
0.500
0.261
0.262
1.353
0.195
1.715
0.500
0.477
1.563
0.500
Optimal sample allocation by EPA Region
Region
1
2
3
4
5
6
7
8
9
10
No. of
Cty's
5.9
10.9
10.9
6.0
19.6
3.8
5.9
3.3
4.5
5.9
No. of
SEGs
26.6
49.0
48.7
27.0
87.8
16.8
26.6
14.7
20.3
26.6
No. of
HUs
125.5
231.4
230.0
127.7
414.5
79.4
125.5
69.5
95.7
125.5
Pop %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp
of HUs
7.7
14.2
14.2
7.9
25.5
4.9
7.7
4.3
5.9
7.7
Total
76.8
344.1
1,624.7
100.0
100.0
2-36
-------
Exhibit 6. The Exhibit 5 Sample Allocation Expanded to a Sample Size of
5,000 Housing Units. (Calculations Assume that the Distribution of
Residential Radon Concentration Varies across the 10 EPA Regions.)
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RHO1
Assumed intracluster corr coef. RHO2
Domain
U. S. Total
Region 1
Region 2
Region 3
Region 4
Region .5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
0.070
0.130
0.130
0.010
0.130
0.010
0.070
0.130
0.010
0.070
Prop>10
0.00499
0.00200
0.01000
0.01000
o.ooioo
0.01000
0.00100
0.00200
0.01000
0.00100
0.00200
0.05
0.10
Desired
Maximum
RSE
0.500
0.500
0.500
0.500
2.000
0.500
2.000
0.500
0.500
2.000
Achieved
RSE
0.339
0.339
0.177
0.177
0.916
0.132
1.161
0.339
0.323
1.058
0.500
0.339
Optimal sample allocation by EPA Region
Region
1
2
3
4
5
6
7
8
9
10
Total
No. of
Cty's
9.7
17.8
17.7
9.8
31.9
6.1
9.7
5.4
7.4
9.7
No. of
SEGs
77.2
142.4
141.6
78.6
255.1
48.9
77.2
42.8
58.9
77.2
No. of
HUs
386.1
712.2
707.8
393.0
1,275.7
244.4
386.1
214.0
294.6
386.1
Pop %
Of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp
of HUs
7.7
14.2
14.2
7.9
25.5
4.9
7.7
4.3
5.9
7.7
125.0
1,000.0
5,000.0
100.0
100.0
2-37
-------
Exhibit 7. The Expected Relative Standard Errors of Estimates Based
on the Exhibit 6 Sample Allocation, If in fact the Distribution of
Residential Radon Were Actually Constant across the 10 EPA Regions.
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RH01
Assumed intracluster corr coef. RHO2
Domain
U. S. Total
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
Prop>lO
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.05
0.10
Desired
Maximum
RSE
0.500
0.500
0.500
0.500
2.000
0.500
2.000
0.500
0.500
2.000
0.500
Achieved
RSE
0.431
0.339
0.249
0.250
0.336
0.186
0.425
0.339
0.455
0.388
0.339
Optimal sample allocation by EPA Region
Region
1
2
3
4
5
6
7
8
9
10
No. of
Cty's
9.7
17.8
17.7
9.8
31.9
6.1
9.7
5.4
7.4
9.7
No. of
SEGs
77.2
142.4
141.6
78.6
255.1
48.9
77.2
42.8
58.9
77.2
No. of
HUs
386.1
712.2
707.8
393.0
1,275.7
244.4
386.1
214.0
294.6
386.1
Pop %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp
of HUs
7.7
14.2
14.2
7.9
25.5
4.9
7.7
4.3
5.9
7.7
Total
125.0
1,000.0
5,000.0
100.0
100.0
2-38
-------
Exiribit 8 Classification of States into Substrata within Strata Defined
Along the Lines of the EPA Regions
EPA Region Substratum
State
Region 1:
Region 2:
Region 3:
Region 4:
High:
Medium:
High:
Medium:
Low:
High:
Low:
High:
ME, NH, VT.
MA, CT, RI.
northern NJ.
NY.
southern NJ.
PA, western MD, WV, western VA.
DE, central and eastern VA.
western NC, western SC, northern GA, northern AL
Region 5:
Region 6:
Region 7:
Medium:
Low:
High:
Low:
High:
Medium:
Low:
High:
Medium:
Region 8: High:
eastern TN.
KY, western and central TN.
central and eastern NC, eastern SC, southern GA,
southern AL, MS, FL.
MN, WI, IL, IN, OH.
MI.
OK, western and central TX, northern AR.
LA, southern AR, southeastern TX.
NE, IA.
KS, MO.
MT, WY, UT, CO, ND, SD.
Region 9:
Region 10:
High:
Low:
High:
Low:
NV.
CA, AZ, HI.
AK, ID.
WA, OR.
The above classifications were made using the following data and procedures:
• Using data from the first two years of State Radon Surveys.
• Using the results of ATD testing by Landauer and Terradex,
presented at the Radon Symposium in Denver in October, 1988.
• Grouping states with similar geographic location and terrain
into the same stratum as contiguous areas.
• The percentage of homes with a basement was not taken into
consideration. This information is available for the State
Radon Survey samples, but was not provided for the Landauer-
Terradex homes. The vast majority of homes in the northern part
of the nation do have basements, but the proportion of homes
with basements is quite small in some southern states.
2-39
-------
Exhibit 9. The Minimum-Cost Sample Allocation When Both the National Level
and the Regional Level Precision Constraints are Considered. (Calculations
Assume that the Distribution of Residential Radon Concentration Can Vary
from One Geographical Area to the Next within Each of the 10 EPA Regions.)
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RH01 = 0.05
Assumed intracluster corr coef. RHO2 = 0.10
Domain
U. S. Total
Prop>4
0.070
Prop>10
0.00500
Desired
Maximum
RSE
0.500
Achieved
RSE
0.500
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
0.085
0.081
0.106
0.042
0.113
0.044
0.095
0.138
0.014
0.031
0.00471
0.00459
0.00834
0.00298
0.00887
0.00244
0.00592
0.01080
0.00134
0.00260
0.500
0.500
0.500
1.000
0.500
1.000
0.500
0.500
2.000
1.000
0.500
0.407
0.286
0.440
0.204
0.588
0.500
0.441
1.105
1.000
Optimal sample allocation by EPA Region
EPA
Region
1
2
3
4
5
6
7
8
9
10
No. of
Cty's
4.7
7.4
9.8
9.5
18.5
5.7
4.2
3.6
5.1
2.3
NO. Of
SEGs
21.2
33.0
43.9
42.7
82.7
25.7
18.8
16.1
23.0
10.2
No. Of
HUS
99.9
155.8
207.1
201.7
390.4
121.4
88.6
75.9
108.4
48.3
Pop %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp %
of HUs
6.7
10.4
13.8
13.5
.1
,1
26.
8.
5.9
5.1
7.2
3.2
Total
70.8
317.2
1,497.6
100.0
100.0
2-40
-------
Exhibit 9 (Continued)
Sample allocation by stratum
EPA
Region
1
1
2
2
2
3
3
4
4
4
5
5
6
6
6
7
7
Radon
class
HIGH
MED
HIGH
MED
LOW
HIGH
LOW
HIGH
MED
LOW
HIGH
LOW
HIGH
MED
LOW
HIGH
MED
No. of
Cty's
1.5
3.2
2.7
4.4
0.3
8.9
0.9
3.5
1.9
4.2
17.2
1.3
0.7
3.1
1.9
2.2
2.0
No. of
SEGs
6.8
14.3
11.9
19.9
1.2
39.8
4.1
15.5
8.4
18.8
76.9
5.8
3.1
14.0
8.6
9.8
8.9
No. of
HUs
32.2
67.7
56.1
94.0
5.7
187.9
19.3
73.4
39.6
88.8
363.2
27.2
14.6
66.1
40.8
46.4
42.2
Pop %
of HUs
1.2
4.1
2.3
7.4
0.8
7.8
2.6
3.0
3.1
12.1
15.1
3.7
0.6
5.2
5.5
1.9
3.1
Samp %
of HUs
2.2
4.5
3.7
6.3
0.4
12.5
1.3
4.9
2.6
5.9
24.3
1.8
1.0
4.4
2.7
3.1
2.8
Samp
Rate
1.86
1.10
1.61
0.85
0.49
1.61
0.49
1.61
0.85
0.49
1.61
0.49
1.61
0.85
0.49
1.65
0.90
Rel . Samp
Rate
3.79
2.24
3.27
1.73
1.00
3.2-7
1.00
3.27
1.73
1.00
3.27
1.00
3.27
1.73
1.00
3.36
1.83
HIGH
3.6
16.1
75.9
3.1
5.1 1.61
3.27
9
9
10
10
HIGH
LOW
HIGH
LOW
0.5
4.6
0.9
1.4
2.4
20.6
4.0
6.2
11.3
97.0
19.0
29.2
0.5
13.2
0.6
3.1
0.8
6.5
1.3
2.0
1.61
0.49
2.09
0.62
3.27
1.00
4.24
1.27
Total
70.8
317.2 1,497.6 100.0
100.0
2-41
-------
Exhibit 10. The Exhibit 9 Sample Allocation Expanded to a Sample Size of
5,000 Housing Units. (Calculations Assume that the Distribution of
Residential Radon Concentration Can Vary from One Geographical Area to the
Next within Each of the 10 EPA Regions.)
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RH01 = 0.05
Assumed intracluster corr coef. RHO2 = 0.10
Domain
U. S. Total
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
Prop>10
0.00500
0.085
0.081
0.106
0.042
0.113
0.044
0.095
0.138
0.014
0.031
0.00471
0.00459
0.00834
0.00298
0.00887
0.00244
0.00592
0.01080
0.00134
0.00260
Desired
Maximum
RSE
0.500
0.500
0.500
0.500
1.000
0.500
1.000
0.500
0.500
2.000
1.000
Achieved
RSE
0.325
0.325
0.265
0.186
0.286
0.133
0.382
0.325
0.287
0.718
0.650
Optimal sample allocation by EPA Region
EPA
Region
1
2
3
4
5
6
7
8
9
10
No. of
Cty's
8.3
13.0
17.3
16.8
32.6
10.1
7.4
6.3
9.0
4.0
No. of
SEGs
No. of
HUs
66.7
104.0
138.3
134.7
260.7
81.1
59.2
50.7
72.4
32.2
333.7
520.2
691.6
673.5
1,303.5
405.4
295.8
253.3
361.9
161.1
Pop %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp %
of HUs
6.7
10.4
13.8
13.5
26.1
8.1
5.9
5.1
7.2
3.2
Total
125.0
1,000.0
5,000.0
100.0
100.0
2-42
-------
Exhibit 10 (Continued)
Sample allocation by stratum
EPA
Region
1
1
2
2
2
3
3
4
4
4
5
5
6
6
6
7
7
Radon
class
HIGH
MED
HIGH
MED
LOW
HIGH
LOW
HIGH
MED
LOW
HIGH
LOW
HIGH
MED
LOW
HIGH
MED
No. of
Cty's
2.7
5.7
4.7
7.8
0.5
15.7
1.6
6.1
3.3
7.4
30.3
2.3
1.2
5.5
3.4
3.9
3.5
No. of
SEGs
21.5
45.2
37.5
62.8
3.8
125.4
12.9
49.0
26.4
59.3
242.5
18.2
9.8
44.1
27.2
31.0
28.2
No. of
HUs
107.6
226.1
187.4
313.8
19.1
627.2
64.4
245.0
132.1
296.4
1,212.6
90.9
48.8
220.5
136.1
154.9
140.9
Pop %
of HUs
1.2
4.1
2.3
7.4
0.8
7.8
2.6
3.0
3.1
12.1
15.1
3.7
0.6
5.2
5.5
1.9
3.1
Samp %
of HUs
2.2
4.5
3.7
6.3
0.4
12.5
1.3
4.9
2.6
5.9
24.3
1.8
1.0
4.4
2.7
3.1
2.8
Samp
Rate
1.86
1.10
1.61
0.85
0.49
1.61
0.49
1.61
0.85
0.49
1.61
0.49
1.61
0.85
0.49
1.65
0.90
Rel. Samp
Rate
3.79
2.24
3.27
1.73
1.00
3.27
1.00
3.27
1.73
1.00
3.27
1.00
3.27
1.73
1.00
3.36
1.83
HIGH
6.3
50.7
253.3
3.1
5.1
1.61
3.27
9
9
10
10
HIGH
LOW
HIGH
LOW
0.9
8.1
1.6
2.4
7.6
64.8
12.7
19.5
37.9
324.0
63.5
97.5
0.5
13.2
0.6
3.1
0.8
6.5
1.3
2.0
1.61
0.49
2.09
0.62
3.27
1.00
4.24
1.27
Total
125.0 1,000.0 5,000.0 100.0
100.0
2-43
-------
Exhibit 11. The Expected Relative Standard Errors of Estimates Based on the
Exhibit 10 Sample Allocation, If in fact the Distribution of Residential
Radon Were Actually Constant across the Different Substrata.
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RHO1 = 0.05
Assumed intracluster corr coef. RHO2 = 0.10
Domain
U. S. Total
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
Prop>10
0.00500
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.070
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
0.00500
Desired
Maximum
RSE
0.500
0.500
0.500
0.500
1.000
0.500
1.000
0.500
0.500
2.000
1.000
Achieved
RSE
0.415
0.373
0.307
0.288
0.283
0.206
0.350
0.404
0.418
0.359
0.580
Optimal sample allocation by EPA Region
EPA
Region
1
2
3
4
5
6
7
8
9
10
No. of
Cty's
8.3
13.0
17.3
16.8
32.6
10.1
7.4
6.3
9.0
4.0
No. of
SEGs
No. of
HUs
66.7
104.0
138.3
134.7
260.7
81.1
59.2
50.7
72.4
32.2
333.7
520.2
691.6
673.5
1,303.5
405.4
295.8
253.3
361.9
161.1
Pop %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp %
of HUs
6.7
10.4
13.8
13.5
26.1
8.1
5.9
5.1
7.2
3.2
Total
125.0
1,000.0
5,000.0
100.0
100.0
2-44
-------
Exhibit 11 (Continued)
Sample allocation by stratum
EPA
Region
1
1
2
2
2
3
3
4
4
4
5
5
6
6
6
7
7
Radon
class
HIGH
MED
HIGH
MED
LOW
HIGH
LOW
HIGH
MED
LOW
HIGH
LOW
HIGH
MED
LOW
HIGH
MED
No. of
Cty's
2.7
5.7
4.7
7.8
0.5
15.7
1.6
6.1
3.3
7.4
30.3
2.3
1.2
5.5
3.4
3.9
3.5
No. of
SEGs
21.5
45.2
37.5
62.8
3.8
125.4
12.9
49.0
26.4
59.3
242.5
18.2
9.8
44.1
27.2
31.0
28.2
No. of
HUs
107.6
226.1
187.4
313.8
19.1
627.2
64.4
245.0
132.1
296.4
1,212.6
90.9
48.8
220.5
136.1
154.9
140.9
Pop %
of HUs
1.2
4.1
2.3
7.4
0.8
7.8
2.6
3.0
3.1
12.1
15.1
3.7
0.6
5.2
5.5
1.9
3.1
Samp %
of HUs
2.2
4.5
3.7
6.3
0.4
12.5
1.3
4.9
2.6
5.9
24.3
1.8
1.0
4.4
2.7
3.1
2.8
Samp
Rate
1.86
1.10
1.61
0.85
0.49
1.61
0.49
1.61
0.85
0.49
1.61
0.49
1.61
0.85
0.49
1.65
0.90
Rel . Samp
Rate
3.79
2.24
3.27
1.73
1.00
3.27
1.00
3.27
1.73
1.00
3.27
1.00
3.27
1.73
1.00
3.36
1.83
HIGH
6.3
50.7
253.3
3.1
5.1
1.61
3.27
9
9
10
10
HIGH
LOW
HIGH
LOW
0.9
8.1
1.6
2.4
7.6
64.8
12.7
19.5
37.9
324.0
63.5
97.5
0.5
13.2
0.6
3.1
0.8
6.5
1.3
2.0
1.61
0.49
2.09
0.62
3.27
1.00
4.24
1.27
Total
125.0 1,000.0 5,000.0 100.0
100.0
2-45
-------
Exhibit 12. The Recommended Design.
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = $432.78
C3 = $174.71
Assumed intracluster corr coef. RH01 = 0.05
Assumed intracluster corr coef. RHO2 = 0.10
Domain
U.S. Total
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
Prop>10
0.00500
Desired
Maximum
RSE
0.500
0.085
0.081
0.106
0.042
0.113
0.044
0.095
0.138
0.014
0.031
0.00471
0.00459
0.00834
0.00298
0.00887
0.00244
0.00592
0.01080
0.00134
0.00260
0.500
0.500
0.500
1.000
0.500
1.000
0.500
0.500
2.000
1.000
Achieved
RSE
0.329
0.337
0.273
0.188
0.281
0.134
0.395
0.340
0.294
0.700
0.664
Optimal Sample Allocation by EPA Region
EPA
Region
1
2
3
4
5
6
7
8
9
10
Total
No. of
Cty's
8.0
14.0
17.0
17.0
32.0
10.0
7.0
6.0
10.0
4.0
No. of
SEGs
No. of
HUs
125.0
64.0
112.0
136.0
136.0
256.0
80.0
56.0
48.0
80.0
32.0
320.0
560.0
680.0
680.0
1,280.0
400.0
280.0
240.0
400.0
160.0
1,000.0
5,000.0
2-46
Pop %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
100.0
Samp %
of HUs
6.4
11.2
13.6
13.6
25.6
8.0
5.6
4.8
8.0
3.2
100.0
-------
Exhibit 12 (Continued)
Optimal Sample Allocation by Stratum
EPA
Region
1
1
2
2
2
3
3
4
4
4
5
5
6
6
6
7
7
Radon
class
HIGH
MED
HIGH
MED
LOW
HIGH
LOW
HIGH
MED
LOW
HIGH
LOW
HIGH
MED
LOW
HIGH
MED
No. of
Cty's
3.0
5.0
5.0
7.0
2.0
15.0
2.0
6.0
4.0
7.0
30.0
2.0
2.0
5.0
3.0
4.0
3.0
No. Of
SEGs
24.0
40.0
40.0
56.0
16.0
120.0
16.0
48.0
32.0
56.0
240.0
16.0
16.0
40.0
24.0
32.0
24.0
No. of
HUs
120.0
200.0
200.0
280.0
80.0
600.0
80.0
240.0
160.0
280.0
1,200.0
80.0
80.0
200.0
120.0
160.0
120.0
Pop %
of HUs
1.2
4.1
2.3
7.4
0.8
7.8
2.6
3.0
3.1
12.1
15.1
3.7
0.6
5.2
5.5
1.9
3.1
Samp %
of HUs
2.4
4.0
4.0
5.6
1.6
12.0
1.6
4.8
3.2
5.6
24.0
1.6
1.6
4.0
2.4
3.2
2.4
Samp
Rate
2.079
0.973
1.717
0.759
2.060
1.539
0.611
1.576
1.031
0.465
1.592
0.433
2.636
0.772
0.434
1.709
0.765
Rel . Samp
Rate
4.801
2.247
3.965
1.754
4.758
3.553
1.412
3.639
2.380
1.073
3.676
1.000
6.086
1.782
1.002
3.947
1.767
HIGH
Total
6.0
48.0
240.0
3.1
4.8
125.0 1,000.0 5,000.0 100.0
100.0
No of Expected HUs with radon level > 4 = 491.6
No of Expected HUs with radon level > 10 = 36.3
1.524
3.519
9
9
10
10
HIGH
LOW
HIGH
LOW
2.0
8.0
2.0
2.0
16.0
64.0
16.0
16.0
80.0
320.0
80.0
80.0
0.5
13.2
0.6
3.1
1.6
6.4
1.6
1.6
3.397
0.486
2.628
0.512
7.844
1.122
6.069
1.182
2-47
-------
Exhibit 13 Expected Precision of Domain Estimates of the Percent of Homes
with Radon Concentration Above 4pCi/L and Above lOpCi/L for the
Recommended Design if the Residential Radon Levels Vary Across
Strata, as Shown in Exhibit 12
Row
No.
1
2
Percent
Domai n
(1)
100
50
Percent
of PSUs
(2)
100
100
Percent
Segments
(3)
100
100
Percent HU's
within Segment
(4)
100
50
Expected Precision
(RSE) for Percent
>4pCi/L >10pCi/L
(5) (6)
0.09 0.33
0.10 0.37
50
100
0.10
0.40
4
5
25
100
100
25
25
100
0.11
0.13
0.44
0.49
6
7
10
100
50
10
20
100
0.16
0.18
0.61
0.70
8
9
5
100
25
5
20
100
0.21
0.25
0.83
0.96
For example, if we estimate that 0.5 percent of all homes in the nation
have radon concentrations above 10 pCi/L, the "Percent Domain" is 100
because the estimate pertains to 100 percent of the survey eligible homes
in the nation. Each of the valuesforcolumns2, 3, and 4 is also 100
because our estimate would use the entire sample, all sample HUs in all
sample SSUs in all sample PSUs. We would therefore read the relative
standard error for our estimate from row 1 column 6, and find that it is
0.33. Not that this is the same relative standard error that was provided
in Exhibit 12 under the "Achieved RSE" for the "U.S. Total" domain.
Let us consider now estimates describing a subgroup in the population.
Suppose that subgroup A consists of half of the homes in the nation, and
that these homes are distributed uniformly over the entire nation. We
would use row 2 to determine the relative standard error of our estimate
because we would expect our subgroup to be represented in all of the sample
PSUs, all of the sample SSUs, and half of the sample HUs in each sample
SSU. If we consider subgroup B that also consists of half of the homes in
the nation, but homes in only half of the SSUs, we would use row 3 to
determine the relative standard error of the estimate.
2-48
-------
Exhibit 14 Expected Precision of Domain Estimates of the Percent of Homes
with Radon Concentration Above 4pCi/L and Above lOpCi/L for the
Recommended Design if the Radon Distribution is Uniform
Across Strata
Row
No.
Percent
Domain
(1)
Percent
of PSUs
(2)
Percent
Segments
(3)
Percent HU's
within Segment
(4)
Expected Precision
(RSE) for Percent
>4pCi/L >10pCi/L
(51(61
2
3
100
50
100
100
100
100
50
100
50
100
0.11
0.12
0.13
0.42
0.47
0.50
4
5
25
100
100
25
25
100
0.15
0.16
0.56
0.63
6
7
10
100
50
10
20
100
0.20
0.23
0.79
0.91
8
9
5
100
25
5
20
100
0.27
0.32
1.06
1.24
For example, if we estimate that 0.5 percent of all homes in the nation
have radon concentrations above 10 pC1/L, the "Percent Domain" is 100
because the estimate pertains to 100 percent of the survey eligible homes
in the nation. Each of the valuesforcolumns2, 3, and 4 is also 100
because our estimate would use the entire sample, all sample HUs in all
sample SSUs in all sample PSUs. We would therefore read the relative
standard error for our estimate from row 1 column 6, and find that it 1s
0.42. Note that this is larger than the relative standard error that was
shown in row 1 column 6 of Exhibit 13 because Exhibit 14 assumes that the
radon levels do not vary from one geographic area to the next, as we
assumed in developing the optimal sample design. Exhibit 14 provides a
"worst case" scenario, showing the relative standard errors for estimates
if the assumptions we made about radon levels differing from one geographic
area to the next were not correct and that instead residential radon levels
were the same all over the nation.
The procedures for using rows 2-9 of
as was described in Exhibit 13.
Exhibit 14 follow the same logic
2-49
-------
Exhibit 15 Some Additional Domain Estimates Where a Maximum Relative
Standard Error of 0.50 is Desired for the Percent of Homes
with Radon Levels above 4 pCi/L (Recommended Design)
Domain
Description
Likely Distribution'
Percent of
total
population
Percent
of PSUs
Percent
of SSUs
Percent of
HUs within
a segment
Expected
Precision
for percent
> 4 pCi/L
Single family
homes 65
Residences in
multi-unit
structures
Homes with 1+
smokers
Homes with 0
smokers
Homes with 1+
children
under age 12 15
Homes without
children
under age 12 85
Homes with 1+
children
under age 12
and 1+
smokers 5
Average radon
level for
homes with
basements 40
Average radon
level for
homes without
basements 60
100
100
100
100
80
80
50
100
25
60
80
30
85
20
70
100
80
75
0.09
35
25
75
100
100
100
60
100
100
60
25
25
0.10
0.11
0.09
0.15
0.09
0.21
0.10
0.10
3 Footnote appears on following page.
2-50
-------
Exhibit 15 (Continued)
The "Likely Distribution" information provided in Exhibit 15 1s based on the
most recent data we could find plus some educated guessing. For example, the
1988 Statistical Abstract provides an estimate that 1n 1983 67 percent of the
housing units in the nation were in one unit structures. Because we were
providing only very rough "estimates," we rounded the number to 65 percent.
It seemed reasonable to expect that there will be single family homes in all
or virtually all PSUs, so we listed 100 as the percent of PSUs that would
contain single family homes. Because neighborhoods tend to be somewhat
homogeneous on a characteristic such as single family homes, we "guessed"
that about 80 percent of the SSUs would have one or more such homes and that
on the average about 80 percent of the homes in these SSUs would be in single
family structures. This type of reasoning provided some rough guidelines for
us to use to determine the approximate relative standard errors that we would
expect for estimates of the percentage of homes with radon levels greater
than 4pCi/L. The precision figures listed 1n the far right column of Exhibit
15 were obtained a) assuming that the "likely distribution" information was
correct, and b) using the values in Exhibit 13 and interpolating.
2-51
-------
Exhibit 16. Minimal Practical Design. (Based on Allocation in Exhibit 9)
Assumed cost components
CO = $810,928.00
Cl = $4,567.57
C2 = * $497.78
C3 = $174.71
Assumed intracluster corr coef. RH01 = 0.05
Assumed intracluster corr coef. RHO2 = 0.10
Domain
U.S. Total
Region 1
Region 2
Region 3
Region 4
Region 5
Region 6
Region 7
Region 8
Region 9
Region 10
Prop>4
0.070
Prop>10
0.00500
Desired
Maximum
RSE
0.500
0.085
0.075
0.108
0.044
0.112
0.048
0.095
0.137
0.014
0.031
0.00468
0.00417
0.00850
0.00322
0.00877
0.00256
0.00587
0.01068
0.00133
0.00258
0.500
0.500
0.500
i.ooo
0.500
1.000
0.500
0.500
2.000
1.000
Achieved
RSE
0.490
0.460
0.409
0.289
0.403
0.199
0.636
,469
,385
.198
0.
0.
1.
0.710
Optimal Sample Allocation by EPA Region
EPA
Region
1
2
3
4
5
6
7
8
9
10
No. of
cty's
5.0
8.0
9.0
9.0
17.0
6.0
4.0
4.0
5.0
4.0
No. of
SEGs
30.0
48.0
54.0
54.0
102.0
36.0
24.0
24.0
30.0
24.0
No. of
HUs
150.0
240.0
270.0
2'70.0
510.0
180.0
120.0
120.0
150.0
120.0
Pop %
of HUs
5.3
10.5
10.4
18.2
18.8
11.3
5.0
3.1
13.6
3.7
Samp %
of HUs
7.0
11.3
12.7
12.7
23.9
8.5
5.6
5.6
7.0
5.6
Total
71.0
426.0
2,130.0
100.0
100.0
2-52
-------
Exhibit 17 Expected Precision of Domain Estimates of the Percent of Homes
with Radon Concentrations above 4 pCi/L and above 10 pCi/L for
the Minimal Design If the Residential Radon Levels Vary across
Strata, as Shown in Exhibit 16
Row Percent Percent Percent Percent HU's
No. Domain Of PSU's Segments within segments
(1) (2) (3) (4)
Expected Precision
(RSE) for Percent
>4pCi/L
(5)
>10pCi/L
(6)
2
3
4
5
6
7
8
9
100
50
25
10
5
100
100
100
100
100
100
100
50
100
25
50
10
25
5
100
50
100
25
100
20
100
20
100
0.13
0.14
0.15
0.18
0.20
0.25
0.29
0.34
0.40
0.49
0.56
0.59
0.68
0.76
0.97
1.12
1.32
1.54
2-53
-------
Exhibit 18 Some Additional Domain Estimates Where a Maximum Relative
Standard Error of 0.50 Is Desired for the Percent of Homes
with Radon Levels above 4 pCi/L (Minimal Design)
Domain
Description
Likely Distribution
Percent of
total
population
Percent
of PSUs
Percent
of SSUs
Percent of
HUs within
a segment
Expected
Precision
for percent
> 4 pCi/L
Single family
homes 65
Residences in
multi-unit
structures
Homes with 1+
smokers
Homes with 0
smokers
Homes with 1+
children
under age 12 15
Homes without
children
under age 12 85
Homes with 1+
children
under age 12
and 1+
smokers 5
Average radon
level for
homes with
basements 40
Average radon
level for
homes without
basements 60
100
100
100
100
80
80
50
100
25
60
80
30
85
20
70
0.13
35
25
75
100
100
100
60
100
100
60
25
25
0.14
0.18
0.14
0.27
0.13
0.34
0.16
100
80
75
0.15
2-54
-------
•r-,-
5 1iF IF
q 5r 5:
UJ Q> O
_J tJ TJ
o o
O K>
§8°™
§ - II II
S CM C C
II II - -
c c fO to
r^ us
C «C
0) O
T» -O
m
•
o
CM
O
Q.
fr?
TT
'5.
8
O)
JQ
JD
Q.
O
c
.2
o
Q.
o
T3
5
CO
s *,
2 c
0) •**
w
o «
"" ^
D II
a
t:
jca i:
« - •§* ~
« « * !:
§ 5
O.
it
bi)
o
o
II
c
• *
m
C
w>
"53
v
o
a.
S
ca
CO
o o
— o
o
en
o
CO
o
(O
o
m
o o
CM »-
a_ o
2-55
-------
o o
O ro
r
o II II
m CM c c
.
5 IT IT IT IF
O C 5= 3: C
^^ Ml Ml fl% rtj
— ' T3 -0 tJ t>
I
in
•
o
ro
¥_
'a.
fr?
r>v
II
s
o>
•o
3
C0
s
CO
CO
e? n*
^ 10 ^
^ q ^
*> •- 2 eo
£ Q- A -H
t) • 31 W
03 r ffl c
a
— o
0)
5 2 S. -w
SB §* « ^
§ o w 7
fc a> it
Q.
I
»
o
Q.
8
60 9
••* 41
o £ SB
w §- S
o> ^
Q «2 *®
"o a, C
. Q) O
« §
£ "•
O
o
c
• •
00
Q
.
g
co
CO
o o
"- O
o
a>
o
CO
o
o
CM
Q_ O
2-56
-------
m
o o
O K)
II II
m CM c c
5 IT IT ¥ IT
O<*- -4— t- **~
H- <*- >«- H_
UJ 4> Q
I i
K)
liT
2
eo
S «
S e
zj
h> CS U
^ *J nj
® C3 C
O. o> y;
,
g
CO
CO
o
o
o
o>
o
to
o
(O
o
m
o o
CM •—
Q. O
2-57
-------
3.0 SURVEY IMPLEMENTATION
This chapter is divided into four sections that address those
components necessary for successful implementation of the National
Residential Radon Study. Section 3.1 discusses those data collection
instruments already developed, as well as those to be developed. The OMB
submission and pretest plans are discussed in this section. Section 3.2
describes the training materials that will be utilized in the study. Those
materials specific to the study and those that are universal to any survey
are presented In this section.
Section 3.3 details the selection, training, and responsibilities of
the field staff, at both the supervisor and interviewer levels. Among the
topics covered are the counting and listing training of the supervisors,
the responsibilities of the supervisors in addition to counting and
listing, the training of Interviewers, and the subsequent responsibilities
of these sa»e Interviewers. Finally, Section 3.4 provides information on
the steps we will take to assure that the highest quality data are
collected.
3.1 Data Col lection Instruments
3.1.1 Existing Instruments
A household questionnaire has been developed for the National
Radon Study. This Instrument incorporates comments to a draft questionnaire
that was subnltted to the Science Advisory Board and also incorporates the
result of an Iterative review process that has included both EPA and RTI
project staff. The questionnaire Includes items that address the substructure
.
3-1
-------
type, heating, cooling and ventilation systems, and construction techniques
and materials. Household composition and occupancy patterns of the residences
have been incorporated in this revised instrument. Other items address the
number and ages of the household occupants and the approximate percentage of
time each occupant spends in each level of the house.
Since the primary risk factor of lung cancer is cigarette smoking, a
smoking history section has been added to the questionnaire to assess the
smoking patterns in the sample households. A draft of the questionnaire will
be submitted for inclusion in the OMB package. The OMB version of the
questionnaire will be pretested on less than sixty households to: 1) assess
the comprehensibility of the questions to the respondent, 2) determine if
adequate data are obtained and 3) identify other problems with the instrument
and the data collection protocol.
In addition to pretesting the questionnaire, we will also pretest the
control/screening form and the consent form for these same three areas of
concern. The control/screening form has been developed for use by the field
interviewer to determine if a housing unit is eligible for inclusion in the
study. Eligibility criteria are: 1) the housing unit must be a primary
residence, and 2) the occupants for the household must have no firm plans to
move within the next 12 months. The placement of ATDs will be tested as a
part of the overall data collection protocol.
Along with the other components of the survey, the lead letter will be
scrutinized during the pretest. Since we are conducting a pretest and not a
pilot study, it will not be possible to utilize the lead letter in the same
manner as it will be for the main survey. However, we plan to focus attention
on the language of the letter and its subsequent impact on respondent
3-2
-------
participation. To do this, the pretest staff members will present the letter
to one-half of the pretest population prior to the onset of screening or
household data collection. The remaining half will see the letter only after
deployment of the dummy detectors. This will allow us to administer the
entire protocol to a respondent who is blind to the presurvey publicity and
other information contained in the lead letter.
Should the pretest experience indicate changes are necessary in either
the questionnaire, the control/screening form, the lead letter, or the data
collection protocol, these changes will be made. A copy of the questionnaire
appears in Appendix A; a copy of the control/screening form appears in
Appendix B. The lead letter and consent form will be development in
conjunction with EPA and our Human Subjects Committee at a later date.
3.1.2 Forms to be Developed
Four other forms will be developed for use in the national study.
These include a lead letter for the housing unit, a consent/incentive receipt
form for the occupant of the housing unit in which detectors are to be placed,
a respondent notification-of-results letter, and a landlord notification-of-
results letter. The lead letter will be mailed before the onset of data
collection activities. This letter will describe the study and inform the
household residents that an interviewer will visit. An EPA brochure on radon
gas or an excerpt from "A Citizen's Guide to Radon" may be included with the
lead letter. In the case of rental property, landlords will not be sent lead
letters or contacted prior to dector placement.
We will develop a consent form for the occupant of the housing unit that
will explain the purpose of the study and Its confidential nature. One copy
will be retained by the occupant and another copy will be kept with the
3-3
-------
completed data forms. The consent form will include an incentive receipt
section for the respondent to sign to verify that the incentive was received
from the interviewer.
The respondent notification-of-results letter will be designed to inform
respondents of the results of the radon testing. The landlord notification-
of-results letter will be designed to inform landlords radon testing has been
conducted on one of their properties and that results are available upon
request.
Additionally, various administrative forms will be developed for use
during data collection. These forms will facilitate the administration and
monitoring of data collection activities. The primary administration form
will be the combination Control/Screening Form used for recording the outcomes
of each attempted contact with the household (see Appendix B).
3.2 Training Materials
While the field supervisors are listing segments and recruiting Inter-
viewers, RTI in-house project staff will develop materials for supervisor and
interviewer training and data collection. This will include the preparation
of field training and procedural manuals, and the development of the training
agenda.
We will prepare a Field Supervisor's Manual, which will cover project
specific supervisory procedures. Supervisors will use this manual as a
supplement to the interviewer manual. The manual will include information on
such topics as:
• assisting with interviewer training,
• preparing interviewer assignments and work schedules,
• maintaining close communications with each interviewer through
regular telephone and at least one in-person observation visit,
3-4
-------
• reviewing the status of each active case during each supervisory
contact, authorizing appropriate additional action for potential
noninterviews, and assisting the interviewer with difficult cases as
necessary,
• contacting households that refuse to be interviewed and attempting
to obtain their cooperation,
• monitoring interviewer production, efficiency, and costs,
• validating a sample of each interviewer's field work,
• editing completed data collection forms, and
• reporting field work progress to the RTI central office on a weekly
basis.
Additionally, we will prepare a Field Interviewer's Manual that will
contain detailed descriptions of all field procedures that interviewers will
be required to follow during the data collection period. The manual will
serve both as a training manual and as a reference manual during field work.
The topics covered will include:
• study sponsorship, background, specific objectives, and importance
of the study,
• other EPA programs, such as the EPA/State Radon Surveys,
• importance of collecting accurate data,
• schedule of work,
• informed consent and confidentially procedures and their importance,
• procedures for screening households and identifying eligibles,
• contact procedures, explaining the study, overcoming objections,
scheduling appointments, and using other basic interviewing
techniques,
• conducting the interview, including detailed instructions for
questionnaire administration,
• question-by-question specifications containing a detailed descrip-
tion of each question, including the purpose, definitions of terms,
and special probing instructions and other instructions as required,
3-5
-------
• procedures for preserving confidentiality of data and respondent
anonymity in the return of completed questionnaires to RTI,
• quality control procedures, including in-house scan edits, valida-
tions, and the need for immediate referral of problems that might
affect the quality of data collection to the supervisor for
resolution,
• refusal conversion strategies and approaches for handling problem
situations that may arise during contact with households, and
• completing project administrative forms and reporting progress to
the field supervisor.
Three existing RTI manuals will be used by the field staff during
training and data collection: the RTI Field Interviewer's General Manual, the
RTI Field Supervisor's General Manual, and the RTI Field Counting and Listing
Manual. These three manuals contain basic information applicable to field
work for all of our surveys, and their use will eliminate the need to provide
extensive coverage of general subject in the project field manuals. These
three manuals are attached for reference.
A series of handout cards will be developed to aid the interviewer in
capturing accurate data regarding various aspects of the questionnaire. These
cards will graphically illustrate house types, construction materials,
foundation types, HVAC systems, and other items as determined by the project
staff.
3.3 Field Staff Selection, Training, and Responsibilities
Field work for the National Radon Study will require a large, highly
qualified field staff. Selection of this staff and the provision of thorough
training will be critical to the success of the study. The approaches we
plan to use in achieving this goal are described in this section.
3-6
-------
3.3.1 Field Supervisors
The quality of supervisory direction and support received by the
field interviewers will impact greatly on the quality of data collected and
the overall success of the study. Supervisors will be located as close as is
practical to the interviewers with whom they will work.
3.3.1.1 Selecting and Assigning Supervisors
Field supervisors for this study will be selected from
our pool of approximately 75 individuals who have worked as field supervisors
on other RTI projects. The criteria for selection will be quality past
performance and proximity to the sampled geographic areas. Since these
individuals have worked for RTI in the past, we will not need to interview
them, only obtain their commitment by phone.
Each field supervisor will supervise approximately 15 interviewers, so we
will require 12 to 15 field supervisors and approximately 200 to 250 field
interviewers for a sample size of 7,400.
3.3.1.2 Supervisor Training
Field supervisors will attend a two-day briefing session
at RTI. The briefing will cover:
• study purpose and procedures,
• field interviewer recruiting and selection criteria,
• questionnaire specifications,
• interviewer training,
• supervisory and reporting responsibilities,
• extensive review of counting and listing procedures, and
• counting and listing exercises in the local community.
3-7
-------
The RTI Field Supervisor's Manual discussed In an earlier section
explains these topics in detail. Supervisors will attend and participate in
the field interviewer training. Prior to the commencement of interviewer
training, each field supervisor will be given a special briefing at their
training site. This briefing for supervisors will be conducted by the central
office'trainers and will cover all supervisory procedures in detail.
3.3.1.3 Supervisor Responsibilities
Field supervisors will recruit field interviewers in
their area and count and list housing units. Supervisors will screen the
interviewer prospects by phone. During their counting and listing circuit
through the interviewers' area, supervisors will interview the strongest
candidates in person and hire two field interviewers for every primary
sampling unit, giving optimal staff coverage for the projected sample yield
and allowing for interviewer attrition and project scheduling constraints.
Supervisors will use standard RTI procedures for counting and listing
housing units within their assigned SSUs. Sampling materials will include:
• a census map showing the SSU location,
• a hand-drawn map showing the SSU boundaries and the number of
housing units expected, and
• a SSU listing form for recording the address or a description of
each housing unit.
Initially the supervisors will verify the physical boundaries of the SSU
as shown on the SSU map and do a quick count of housing units in the SSU.
Boundaries and counts that vary significantly from what is expected will
require the supervisor to call the RTI sampling staff for instruction. In
some cases, supervisors may be instructed to list only part of the SSU, to
subsegment.
3-8
-------
The.supervisors will apply prescribed rules for traveling through the
SSU, following roads in a specified order to ensure that all possible housing
locations are investigated. Standard rules will specify handling of multiple-
unit dwellings, such as apartment or condominiums and atypical structures,
such as boarding houses.
Supervisors will mail completed listing materials to RTI sampling staff.
If errors are detected, the supervisors will be contacted and asked to resolve
same.
3.3.2 Field Interviewers
3.3.2.1 Recruiting and Hiring Interviewers
RTI will identify potential interviewers listed in our
National Interviewer File, a computerized file of over 3,000 individuals who
have worked, or applied to work, as field interviewers on other RTI projects.
The file includes information about each individual's capabilities, exper-
ience, language fluencies, location and availability. The field supervisors
will receive a list of individuals from the file who appear to have the
necessary qualifications and live in the appropriate geographic area for this
survey. Other interviewer candidates may be identified by contacting other
survey organizations, by advertising in local newspapers, or by recruiting
through state employment agencies.
Interviewer applicants will be screened by field supervisors over the
phone and strong candidates interviewed in person during the supervisors
counting and listing circuit in the interviewers area.
3.3.2.2 Interviewer Training
Approximately 200 to 250 interviewers will be trained for
the survey. Two 2.5 day training sessions will be held, each involving three
3-9
-------
RTI in-house trainers, the field supervisors for that region of the country,
and the interviewers assigned to work that region. Field supervisors will
assist trainers during each training session.
A pre-training study package, which will include study materials and a
questionnaire, will be mailed to each field interviewer several days before
the training session, with specific self-study instructions and authorization
for four hours of pre-training study time. The RTI in-house training staff
will meet prior to the onset of training to to review the training guide,
discuss training procedures and clarify training instructions and related
materials.
The training methods used will include lectures, question-and-answer
review sessions, demonstrations, small group discussions, and structured
drills. Field supervisors will work with their interviewers during training
exercises. A formal training guide will maximize participation and require
trainees to react to potential field situations and to deal with field forms
and procedures. Topics to be covered during the 2.5 day session will include:
• mailing lead letters to sample households,
• identifying a sample housing unit,
• contacting a household member,
• determining eligibility and completing the screener,
• introducing the study and explaining its importance and confidential
nature,
• obtaining participation and signed consent from all occupant
respondents,
• administering the questionnaire, question-by-question,
• answering questions raised by the respondent in clarifying terms,
• placing the radon detectors and recording their location prior to
the interviewer leaving the residence,
3-10
-------
• paying the incentive and getting a signed receipt,
• editing completed questionnaires,
• reporting to field supervisors, and
• completing time and expense reports.
Interviewer trainees will be encouraged to ask questions about any of
these topics. They will practice and review all procedures, in a "mock
interview" setting. Emphasis will be placed on the objectives of the study,
housing characteristics and the importance of correct detector placement.
We anticipate that we will lose some interviewers during training.
However, by hiring two field interviewers for every county, we compensate for
this attrition.
At the conclusion of training, the interviewers will be given their
assignments, materials, and picture ID badges. They will begin their
interviewing assignments Immediately after training.
3.3.2.3 Interviewer Responsibilities
The Interviewer will be responsible for contacting the
household, determining eligibility, administering the questionnaire, aiding in
the placing of, or placing, radon detectors, and paying the incentive. The
interviewer will also be responsible for identifying any missed listing units
in the segment.
The Interviewer will mall a lead letter, a copy of the Citizen's Guide,
and an Information sheet to each sampled housing unit a few days before
beginning work In a segment. The letter will explain the survey and its
importance and will alert the household members that an RTI interviewer will
be contacting then In the next week. Our experience dictates that a lead
letter on government letterhead (i.e., EPA) increases response rates.
3-11
-------
Initial visits to sample households will be made in the late afternoon or
early evening on a weekday or on a Saturday to maximize the chance of finding
either the householder or the spouse at home. If neither the householder nor
spouse are at home, the interviewer will call a second time between 5:30 and
9:00 pm on a weekday or Saturday. Additional visits will be made at varying
times and days in order to find someone at home.
Vacant housing units, non-residential units, and temporary or vacation
homes will be reported to the field supervisor after the interviewer has
verified the status of these units. Upon approval of the supervisor, these
units will be classified as ineligible for the survey and this information
will be routed to RTI to update the control system.
If the householder is at home, the interviewer will identify the study
and ask if the lead letter was received by the household. If not, the
interviewer will give a copy to the respondent and either read the letter out
aloud or allow time for it to be read before proceeding. Then the interviewer
will determine if the households is eligible for the survey and if so, will
proceed with questionnaire administration. If the household is ineligible,
the interviewer will thank the respondent and proceed to the next housing
unit.
Prior to administering the questionnaire, the interviewer will perform
the following tasks:
• describe the survey,
• explain the kinds of questions that will be asked,
• emphasize the importance of the information,
• assure the respondent that the study is completely confidential, and
• read the consent form and obtain verbal agreement to continue with
the questionnaire.
3-12
-------
The Interviewer will have the respondent sign the informed consent form
before proceeding with detector placement.
Other adult household members will be invited to join in the interview
process, as necessary, to increase the accuracy and completeness of the infor-
mation obtained. If the respondent does not know the answer to a question
about the house, the interviewer may accompany the respondent while he/she
investigates (e.g., by going outside, they can determine if the entire house
is over a crawl space or if it is partially on a concrete slab.)
The radon detectors will be placed in the home by an occupant according
to the instructions of the interviewer. As the detectors are placed, the
interviewer will record the location and ID number of each detector and give
the respondent information about the detectors and precautions for their use.
The interviewer will give the participating respondent $5.00 in cash
after the detectors are placed. The respondent will sign the incentive
receipt portion of the consent form to verify that the money was received.
3.4 Field Quality Assurance
Close supervision of field data collection activities in a study of this
kind is essential to achieve high quality data and the maximum response rate
possible within the allotted time and budget. The staffing plan we will use
for the National Residential Radon Study will provide this supervision. The
issues and procedures associated with this project's overall quality assurance
plan are addressed in Section 5.
3.4.1 Supervision
The field supervisors will oversee the work of the field inter-
viewers assigned to them. Each supervisor will recruit, assist in.training,
and monitor the data collection activities of their respective interviewers.
3-13
-------
Each interviewer will contact his/her supervisor at least once a week, usually
by telephone. During these contacts, the supervisor will review each active
case with the interviewer and resolve any problems. The supervisor will
summarize the information received from the interviewers and send an updated
field report to the RTI in-house regional supervisor every week.
The supervisor will also visit each interviewer at least once during the
data collection period to observe their work, help with problem cases, answer
questions, and offer suggestions to improve participation rates. Also during
this visit, the field supervisor will make in-person follow-up contacts with
refusals and perform validation checks on the interviewer's work.
Each field supervisor will report to an in-house regional supervisor at
RTI's central office. The regional supervisor will review the field super-
visor's weekly reports and help resolve problems. The regional supervisor
will monitor field costs and will advise the field supervisor if there are
budgetary problems. The data collection manager will have overall responsi-
bility for all data collection activities and questions which cannot be
resolved by the field and regional supervisors will be referred the manager.
3.4.2 Validation
A rando* sample of at least 10 percent of each interviewer's
completed Interviews will be validated by telephone. Supervisors, who will be
maintaining the completed household screening forms containing name and
address of the respondents, will be responsible for this activity. Non-
interviews will not be Included In the validation sample since follow-up
efforts will be Made on all such cases routinely.
3-14
-------
4.0 DATA REDUCTION PROCEDURES
The data reduction procedures include all data processing steps,
including data receipt, edit/coding, error resolution, data entry, machine
edit, data file preparation, and document storage. An automated event
monitoring system will monitor the receipt of all survey forms, as will
each data processing step. Also included are the linking of data from the
survey forms with the sample design information such as stratum, PSU, SSU,
and HU. This linkage is necessary to calculate sampling weights for each
sample HU and to compute appropriate nonresponse adjustments to the
sampling weights for the statistical analysis.
4.1 Survey Monitoring System
Effective monitoring of the data receipt and data processing task
requires an efficient and flexible system. We will use our automated
survey monitoring system for the National Residential Radon Survey, a
software tool developed by RTI to assist in monitoring data collection and
data processing operations. The system operates under control of FICS
(Fully Integrated Control System), which is a database management system
developed at RTI for managing research data on studies where tracking is
involved and data access is required. FICS allows interactive and batch
processing. It is fully menu driven and supported by a flexible language
that allows users to develop their own routines.
The monitoring activity is controlled by an event table defined by the
user. The events are defined as actions within the task being monitored.
The events mirror a sequence of activities that follow a logical pattern
dictated by the task. Each event is defined within the system as a
4-1
-------
predecessor or outcome of another event. Each processing event is dated so
that duration times may be followed and the user can be alerted when
activities fall behind schedule. Reports may be produced directly from
FICS or the user can design his own reports. The user can tabulate and
print any items from the database. In addition to reports, FICS can make
additional files available to other software systems for management
activities.
4.2 Questionnaire Processing
The data processing task include data receipt, data preparation, data
entry, and hardcopy storage. Each of these activities is discussed in this
section.
4.2.1 Data Receipt
Data receipt involves receipt of the mail, check-in and
identification, scan-edit, and batching. Part of the check-in procedure is
keying events to the automated monitoring system.
After being checked in, a scan-edit checks the forms for completeness
and usability. If the form fails to meet preset specifications for
processing, it is turned over to the field operations manager for
resolution.
All data forms that pass scan-edit are batched for easier handling by
the receipt clerks. These batches are numbered and logged into the data
receipt log book and the monitoring system. The form is entered into the
monitoring system by the data receipt clerk. If the package does not match
a study ID, it is turned over to the data collections operations staff for
resolution. Each form is identified by number and assigned to a batch and
entered into the log. We expect batches to have 25 questionnaires or 75 to
100 screening forms.
4-2
-------
4.2.2 Data Preparation
The batches will be assigned to an edit clerk. The edit
specifications and coding requirements will be described in a Edit and
Coding Manual that will be written for the task. A training session will
instruct the clerks on these procedures. Once work has started, the
supervisor will periodically check a sample of each clerks work to insure
quality work as a quality control measure.
We will follow-up with the respondent on questionnaires that fail edit.
The edit clerk will call the respondent in an attempt to resolve any
problems we find. We expect to recontact 10 percent of the respondents
that complete the questionnaire.
4.2.3 Data Entry
All data received 1n hard copy form will be keyed by RTI 1n our
data entry unit. Our full-time data entry operators work up to two shifts
a day as required by study schedules. They use a data entry software
system to check formats and ranges; 1t can also be programmed to make logic
checks. Skip patterns are checked through detailed machine edits.
A detailed machine-readable codebook describes every element of the
instrument. The codebook contains a short item name and a longer
description for use In analysis. The format type and the acceptable range
of each Item Is part of the codebook record and is used by the data entry
program to check the data as 1t 1s being entered. These codebooks become
part of the permanent documentation for the data entry program and are used
to create documentation for the edit files.
Each keyer Is trained by the programmer and each supervisor is
instructed on how to monitor the work. We will rekey two 10 percent
4-3
-------
samples of the data as a quality control measure. The rekeyed data is
compared to the original to ensure that the data is being keyed within the
prescribed specifications. Two different keyers will do the rekeying. If
poor quality data is found, those batches will be rekeyed and the keyer
will be retrained.
4.2.4 Questionnaire Storage
After data entry, the survey forms will be stored in a secure,
vault-like storage area until the project is completed; they will then be
disposed of according to SC&A or EPA instructions and confidentiality
requirements.
4.3 Data File Preparation
Once the data is in machine-readable format, it will undergo some level
of machine edit. At a minimum, the data will be checked for skip-pattern
errors. More extensive edits may be programmed by specifying the logic
checks of certain key items.
The final data file will be documented with a complete codebook,
formatted as described in the section on Data Entry above. In addition the
frequency of each variable will be calculated and become part of the
codebook. The data file and documentation will be submitted to SC&A as
part of the final report.
4.4 Constructing the Statistical Analysis File
Once all of the data from the survey instruments have been keyed the
raw data file will be merged with the design information file. Result
codes from the control/screening form will be used to ascertain which HUs
were eligible for the study. Among the eligible HUs, each unit will be
classified as a respondent or nonrespondent. The ellgibles will be
4-4
-------
partitioned into weighting classes that are similar by known important
characteristics associated with radon levels and other parameters of
interest. The sampling weights will then be ratio adjusted so that the sum
of the respondents' weights equals the sum of the eligible's weights within
each weighting class.
The nonresponse adjusted weights will be used to inflate parameter
estimates to population levels in the statistical analysis. All estimates
of population parameters and their variances will properly reflect the
stratification and stages of selection employed in the sample design.
4-5
-------
5.0 QUALITY ASSURANCE
A draft Quality Assurance Project Plan (QAPP) has been prepared for the
NRRS according to current EPA guidelines (QAMS 005/80). The specific and
detailed steps that will be implemented to assure that the NRRS will
generate high quality data and satisfy the objectives of the survey are
given in the QAPP. This chapter briefly overviews the issues addressed in
the QAPP, the interested reader is referred to the QAPP for specific
details.
Quality assurance procedures will be applied to essentially all
components of the survey including
• Sampling procedures
• Sample custody
• Measurement calibration
• Analytical procedures
• Data receipt and processing procedures.
The QAPP also outlines steps that will be taken to assure that the five
above items will be performed correctly. These include
• Internal quality control checks
• Audits of the analytical laboratory
• Specific procedures for assessing the precision and bias of the
measurements and the estimates of radon levels for the nation and
important subsets of the nation.
• Corrective action.
The QAPP is organized into 15 Sections and begins in with a general
description of the project. Section 2 specifies the project organizational
5-1
-------
structure and describes the responsibilities of the each organizational
unit. The QA Objectives are stated in terms of precision, bias,
representativeness, completeness, and comparability in Section 3. The
planned sampling procedures are described next including radon measurements
within each home and the process used to select homes in which detectors
will be placed.
In Section 5, the sample custody procedures that will track detectors
from the point of purchase, during in-house and field handling, and return
to the laboratory for analysis. Section 5 also describes the procedures
that will track all other survey forms including the questionnaire and
control/screening form.
Calibration procedure and their frequency are specified in Section 6.
The analytical procedures including the preprocessing of detectors and
their counting are covered in Section 7. Chapter 8 describes the receipt
and processing of all data forms, their manual editing, and automated
processing.
Section 9 outlines the internal quality control checks that will be
required of the laboratory, sample design, survey operations, and data
processing. Audits to assure that the procedures specified 1n Section 9
are followed are planned. Section 10 describes the form and frequency of
the audits. Section 11 specifies the preventive maintenance planned for
the laboratory equipment used to analyze the detectors. Section 12
specifies the procedures that will be used to assess the precision, bias,
and completeness of the measurement data. The steps to be taken if
corrective actions are necessary are described in Section 13.
The content the required QA reports to management are described 1n
Section 14. The references are given Section 15.
5-2
-------
APPENDIX A.
GLOSSARY OF TERMS USED IN CHAPTER 2
-------
Alpha
CO
Cl
C2
C3
Deff.
deft.
Domain
Housing Unit
HU
The alpha level is set by the researcher and is
the probability of rejecting the null hypothesis
when, in fact, it is true. This type of error is
called a type 1 error. (A type 2 error, as you
might expect, is the probability of failing to
reject the null hypothesis when, in fact, it is
false.
The fixed survey cost, that is, costs that are
not affected by the size or distribution of the
sample.
The cost associated with the number of PSUs in
the sample, on a cost per PSU basis.
The cost associated with the number of SSUs in
the sample,, on a cost per SSU basis.
The cost associated with the number of HUs in the
sample, on a cost per HU basis.
Deff is an abbreviation for the term "design
effects" and is a measure of the inefficiency of
a sample design. The Deff for an estimate is
computed as the ratio of two variances, the error
variance of the estimate based on the sample
design used to the error variance of a comparable
estimate if a simple random sample of the same
size had been used.
Deft is merely the square root of Deff.
A subportion of the population for which
statistical estimates are desired. A domain can
be defined in a variety of ways. Examples of
domains of Interest for the National Residential
Radon Survey are geographical regions within the
United States and houses with certain
construction characteristics, such as having a
basement.
A house, apartment, mobile home or trailer, group
of rooms, or single room occupied as a separate
living quarter or, if vacant, intended for
occupancy as a separate living quarter. Separate
living quarters are those in which the occupants
live and eat separately from any other persons in
the building and which have direct access from
the outside of the building or through a common
hall. The occupants of a housing unit may be a
single family, one person living alone, two or
more families living together, or any other group
-------
Place
Place size code
Primary Sampling Unit
PSU
County
of related persons or up to 9 unrelated persons
who share living arrangements.
Each sample SSU is visited by a field
representative. Typically the sample SSU
contains no more than 150 housing units (HUs), in
which case the addresses of all HUs are listed on
a special listing sheet, using a separate line
for each address, and following explicit
instructions contained in a special RTI Field
Listing Manual. When an SSU contains more than
150 HUs, it is usually "subsegmented," and one or
is randomly selected for
more of the
listing.
subparts
A concentration of population which may or may
not have legally prescribed limits, powers, or
functions. Most of the places identified in the
1980 census are incorporated as cities, towns,
villages, or boroughs. In addition, census
designated places (called "unincorporated places"
in earlier censuses) are delineated for 1980
census tabulations. There are about 23,000
places recorded in the 1980 census, of which
about 20,000 are incorporated places and 3,000
are not incorporated. Places may cross county
boundaries but do not cross state boundaries.
00 Not in a place
01 Under 200
02 200-499
03 500-999
04 1,000-1,499
05 1,500-2,499
06 2,000-2,499
07 2,500-4,999
08 5,000-9,999
09 10,000-19,999
10 20,000-24,999
11 25,000-49,999
12 50,000-99,999
13 100,000-249,999
14 250,000-499,999
15 500,000-999,999
16 1,000,000 or more
The population to be sampled is partitioned into
primary sampling units (PSUs) each of which is a
county or county equivalent (or on occasion a
combination of two counties or county
equivalents.) A probability sample of PSUs is
selected. The PSU is sometimes referred to as a
county.
-------
RH01
The intracluster
PSUs.
correlation among SSUs within
RH02
Relative Standard Error
RSE
Secondary Sampling Unit
SSU
Segment
The intracluster correlation
SSUs.
among HUs within
The ratio of the standard error of an estimate to
the parameter being estimated.
A sample PSU is partitioned into smaller areas
such as census defined blocks, block groups or
enumeration districts (EDs). These are secondary
sampling units (SSUs). A probability sample of
SSUs is selected within each PSU. The SSU is
typically referred to as a segment in
instructional materials prepared for field
interviewers.
-------
APPENDIX B
NATIONAL RESIDENTIAL RADON SURVEY
QUESTIONNAIRE
-------
OMB Number:
Expires:
March 20, 1989
QUESTIONNAIRE
THE NATIONAL RESIDENTIAL RADON SURVEY
Sponsored by
U.S. Environmental Protection Agency
The purpose of this study is to determine the extent of radon concentrations
in residential structures throughout the United States. Radon is a radio-
active gas that occurs naturally in soil and rocks and in building materials.
Your household was randomly selected for this important study.
Interviewer Name:
Date of Interview:
PLACE I'D LABEL HERE
Time Interview Began: am/pm
Time Interview Ended: am/pm
-------
1. IF READILY OBSERVABLE, DO NOT ASK.
Which of the following best describes this residence?
[CIRCLE ONE]
A multi-unit building 01 —> GO TO Q3
A mobile home 02 —> GO TO Q6
A single unit, detached dwelling 03 —> GO TO Q2
2. Which of the following house-types best describes this house.
[CIRCLE ONE]
Ranch style or 1 story 01
Split level 02
Split foyer 03 —> Q6
2 story 04
3 or more stories 05 .
3. How many housing units are in this building?
/ / HOUSING UNITS
4. How many floors are in this building, including any below ground?
/ FLOORS
DK 94
RE 97
5. In this building, on which floor is the lowest level of your home located?
/ LOWEST LEVEL OF HOME IN BUILDING
DK 94
RE 97
March 20, 1989
-------
About how old is this (home/building)?
LESS THAN 1 YEAR OLD 01
1-5 YEARS OLD 02
6-10 YEARS OLD 03
11-20 YEARS OLD 04
21-40 YEARS OLD 05
OVER 40 YEARS OLD 06
DK 94
RE 97
7. Do you have a full or partial basement, a cellar, or a level of the house that
has one or more walls partially or completely below ground level? (PROBE: A
level is the area within a home that is all of the same height and is not
separated by any stairs.)
YES 01 —-> CONTINUE
NO 02
DK 94 —-> Q12
RE., 97 .
8. For the purpose of this study we are calling the floor or level that has one or
more walls partially or completely below ground a basement. Can you enter the
area we are calling a basement from inside your home?
YES 01 —> CONTINUE
NO 02
DK 94 —> Q10
RE 97 .
9. Is there a door that can be closed between what we are calling a basement and
the next higher level?
YES 01
NO 02
DK 94
RE 97
March 20, 1989
-------
10. Which of the following describes the construction of most of the outside
basement walls? Are they mostly made of...
[CIRCLE ALL THAT APPLY]
concrete block or cinder block 01
poured concrete 02
stone and mortar 03
wood 04
brick/brick veneer 05
t
earth/dirt or 06
something else 07 —> [SPECIFY] |_|_
DK 94
RE 97
11. Is any part of your basement floor exposed earth?
YES 01
NO 02
DK 94
RE 97
12. Is any part of this home excluding a basement on a concrete slab?
[IF MOBILE HOME: AXLE OR WHEELS RESTING ON CONCRETE PAD SHOULD BE CODED AS A
"NO".]
YES 01
NO 02
DK 94
RE 97
13. Is any part of your home over a crawl space? (PROBE: A crawl space is space
between the ground and the floor structure that cannot be occupied. This is
space other than a basement or cellar.)
YES 01 —> CONTINUE
NO 02
DK 94
RE 97
-> Q20
March 20, 1989
-------
14. Does any part of the crawl space have exposed or visible earth, sand or rock?
YES 01 —> CONTINUE
NO 02
DK 94 —-> Q15
RE 97
14a. What, if anything, covers all or part of the surface of your crawl space?
!_!__!
15. How much of the crawl space is enclosed by foundation walls- all, part or none?
ALL 01
PART 02 —> CONTINUE
NONE 03
DK 94 —> Q20
RE 97 .
16. Which of the following describes the construction of most of the outside
foundation or crawl space walls? Are they mostly made of....
[CIRCLE ALL THAT APPLY]
concrete block or cinder block 01
poured concrete 02
stone and mortar 03
wood 04
brick/brick veneer 05
earth/dirt or 06
something else 07—> [SPECIFY]
DK 94
RE 97
March 20, 1989
-------
17. Can you enter any part of the enclosed crawl space without going outside the
L. A ..»,«. 1
house?
YES 01
NO 02
DK 94
RE 97
18. Are there air vents in the foundation walls of the crawl space?
YES ..01 —> CONTINUE
NO 02
DK 94 —> Q20
RE 97 J
19. What percentage of the time during the year are the vents open? ENTER WHOLE
NUMBERS.
/ / PERCENT
DK 94
RE 97
20. Is any part of your home over open air, that is, on blocks, or pillars?
(PROMPT: These may also be know as columns, pylons, or piers)
YES 01
NO 02
DK 94
RE 97
March 20, 1989
-------
21. What percent of your home rests over a,
a. basement
b. concrete slab | | | | %
c. crawl space | | | | %
d. open air | | | | %
e. something else | | | | % —-> [SPECIFY]
f. DK | | | | %
g. RE I_I_I_J *
(INTERVIEWER: THESE PERCENTAGES MUST ADD UP TO
100%. IF NOT, RECALCULATE OR PROBE AS NECESSARY.)
SCRIPT 1 — READ TO RESPONDENT:
Radon levels tend to be different in different levels of homes. We need to determine
how many levels or floors your home has. A level is the area within a home that is
all at the same height and is not separated by any stairs. We are not counting as
separate levels such things as sunken living rooms or sunken foyers.
22. How many levels or floors are in your home? (PROBE: This includes basement or
cellar.) [CIRCLE ANSWER BELOW AND RECORD ANSWER IN BOX AT THE INTERVIEWER
CHECKPOINT ON PAGE 19.]
[CIRCLE ONE]
One level 01
Two levels 02
Three levels 03
Four levels 04
Five levels 05
Six levels 06
DK 94
RE 97
March 20, 1989
-------
23. Which of these levels has one or more walls partially or completely below
ground?
[CIRCLE ALL THAT APPLY]
Level one 01
Level two 02
Level three 03
Level four 04
Level five 05
Level six 06
None 00
DK 94
RE 97
24. We are interested in the levels of your home that are used as living quarters.
By living quarters we mean the rooms in which you sleep, eat, watch television,
or do other activities of daily life. Starting with the lowest level of your
home as level one, which of these levels do you use as living quarters? (PROBE:
This includes a basement or cellar if it is used as living quarters.)
[CIRCLE ALL THAT APPLY]
Level one 01
Level two 02
Level three 03
Level four 04
Level five 05
Level six 06
None 00
DK 94
RE 97
March 20, 1989
-------
25. Think about the building material used for the floor of the lowest level of your
home, or what we think of as your primary radon barrier. Which of the following
building materials makes up your lowest level floor?
[CIRCLE ALL THAT APPLY]
poured concrete 01
wood 02
earth, or 03
something else 04—> [SPECIFY] l_|_|
DK 94
RE ...97
26. Do you have a garage or underground parking structure?
YES 01 —> CONTINUE
NO 02
DK 94
RE 97
—> Q29
27. Is this garage or underground parking structure attached to your home?
YES 01
NO 02
DK 94
RE 97
28. Does this garage or underground parking structure rest on a concrete or asphalt
surface that is attached to or bordering the foundation of your home? (PROBE:
Bordering means actually touching the foundation.)
YES 01
NO 02
DK 94
RE 97
8 March 20, 1989
-------
29. Are there any other concrete or asphalt surfaces attached to or bordering the
foundation of your home? (PROBE: Bordering means actually touching the
foundation and includes such things as a carport or a patio.)
NO
DK
RE....
02
94
97
/ UUN 1
— > Q31
30. Which of the following structures are attached to or border the foundation of
your home? [CIRCLE ONE RESPONSE FOR EACH ITEM, a-i]
Yes
No
DK
RE
a.
h
c.
d.
e.
f
Q.
h
carport
driveway
sun room
porch (slab on
patio
workshop
sidewalk
anything else?.
01
01
01
grade, only).... 01
01
, 01
01
, 01
I
[IF YES, SPECIFY]
02.
02.
02.
....02.
02.
02.
02.
02.
94...
94...
94...
Q4
• •••*••*• J~ * • •
94...
94...
94...
94...
97
97
97
97
97
97
97
97
1 1 1
31. The next few questions are about the heating, air conditioning and ventilation
systems in your home. Does this home have a main or primary heating system?
YES.
NO.
DK.
RE.
.01
.02
.94
.97
CONTINUE
—> Q37
March 20, 1989
-------
32a. Do you use the following appliances for main or primary heat during the heating
season?
FOR EACH YES CIRCLED, ASK:
32b. Do you use this (NAME)...
1. unvented kerosene
space heater
Y N DK RE
01 02 94 97
2. kerosene space heater 01 02 94 97
vented to the outside
3. unvented gas or propane 01 02 94 97
space heater
4. gas or propane heater 01 02 94 97
vented to the outside
5. woodstove
6. fireplace
01 02 94 97
01 02 94 97
ALWAYS/
ALMOST
ALWAYS
01
01
01
01
01
01
ABOUT
HALF
THE OCCASSION-
TIME ALLY DK
02
02
02
02
02
02
03
03
03
03
03
03
94
94
94
94
94
94
RE
97
97
97
97
97
97
33. Which one of the following fuels do you use for your main or primary heating
system?
[CIRCLE ONE]
oil 01
electricity 02
coal 03
other 04 —> [SPECIFY] |_|
none used 05
DK 94
RE 97
10
March 20, 1989
-------
34. Which one of the following best describes the type of distribution system you
use for your main or primary heating system?
forced air ..01 —> CONTINUE
hot water (i.e. radiator or baseboard) 02
natural convection (i.e. fireplace, woodstove
or floor furnace: without a blower) 03 —> Q37
or, something else 04
i
[SPECIFY] l__l_l
DK .94
RE 97
35. Is this system in a basement or crawl space?
YES 01
NO 02
DK 94
RE 97
36. .Do any ducts carrying the warm air for this system run under the house?
YES 01
NO 02
DK 94
RE 97
37. Do you use any supplemental system to heat your home?
YES 01 —> CONTINUE
NO 02 —> Q43
11
March 20, 1989
-------
38a. Do you use the following appliances for supplemental heat during the heating
season?
FOR EACH YES CIRCLED, ASK:
38b. Do you use this (NAME)...
1. unvented kerosene
space heater
Y N DK RE
01 02 94 97
2. kerosene space heater 01 02 94 97
vented to the outside
3. unvented gas or propane 01 02 94 97
space heater
4. gas or propane heater 01 02 94 97
vented to the outside
5. woodstove
6. fireplace
01 02 94 97
01 02 94 97
ALWAYS/
ALMOST
ALWAYS
01
01
01
01
01
01
ABOUT
HALF
THE
TIME
02
02
02
02
02
02
OCCASION-
ALLY
03
03
03
03
03
03
DK
94
94
94
94
94
94
RE
97
97
97
97
97
97
39. Which of these other fuels are used for supplemental heat?
[CIRCLE ALL THAT APPLY]
oil ..01
electricity 02
coal 03
other 04 —> [SPECIFY]
none used 05
DK 94
RE 97
12
March 20, 1989
-------
40. Do you use any of the following distribution systems for supplemental heat?
Y N DK RE
a. forced air 01 02 94 97
b. hot water 01 02 94 97
c. natural convection
(fireplace, woodstove,
floor furnace,
without blower) 01 02 94 97
d. something else 01 02 94 97
I
[SPECIFY] | | |
[IF YES TO Q40A CONTINUE; IF NO, DK OR RE TO Q40A —> Q43]
41. Is this system in a basement or crawl space?
YES 01
NO 02
DK 94
RE 97
42. Do any ducts carrying the warm air for this system run under the house?
YES 01
NO 02
DK 94
RE 97
43. How many fireplaces do you have?
DK 94
RE 97
13 March 20, 1989
-------
44a. Do you use any of the following humdification devices during the heating season?
FOR EACH YES CIRCLED, ASK:
445. Do you use this (NAME)
MORE LESS
THAN THAN
ONCE ONCE
DAILY PER/WK PER/WK DK RE
a. ultrasonic humidifier
(PROBE: This humidifier
reduces water to a fine
spray using ultrasound;
ultrasound is inaudible
to the human ear.)
b. cool mist humidifier
(PROBE: This humidifier
uses the blade of a rotor to
spray small droplets of
water into the air.)
c. steam mist humidifier
(PROBE: This humidifier
uses a heating coil to
create steam.)
d. humidifier, DK type
Y
01
01
01
01
N DK RE
02 94 97
02 94 97
02 94 97
02 94 97
01 02
01 02
01 02
01 02
03 94 97
03
03
03
94 97
94 97
94 97
45. Is there an air-to-air heat exchanger or heat-recovery ventilator in your home?
(PROBE: This is a system that blows stale air out of the house, brings in fresh
air from outside, and transfers heat from the stale air to the fresh air.)
YES.
NO..
DK..
RE..
.01
.02
.94
.97
46. On the average, what temperature do you usually keep your entire home in the
winter? [USE MARGIN/MARGINAL NOTES TO CALCULATE AVERAGE TEMPERATURE.]
DK.
RE.
DEGREES
.94
.97
14
March 20, 1989
-------
47. Do you have any of the following gas or propane fueled appliances in this home?
[CIRCLE ONE RESPONSE FOR EACH ITEM.]
YES NO DK RE
a.
b.
c.
d.
e.
f.
q.
y
h.
water heater. . . .
clothes dryer.. .
stove
oven
refrigerator
air conditioner.
heat pump
other
01
01
01
01
01
01
01
•••••Ux«**»««
....02.
....02.
02.
02.
.....02.
....02.
02.
....02.
94
• •*••••••• J^ •
94
•••••*•••• J" •
94.
94.
94.
Qd
• • • • • •••• • J ~ •
94.
94.
97
97
97
97
«••••••••*.?/
• • • • * * • • • • _7 /
97
• •••*•••*•_//
[SPECIFY] l_|_l
48. Does this home have a central air conditioning system?
YES 01
NO 02
DK 94
RE........97
49. Do you use any of these other air conditioning systems? [IF YES TO A OR B, ASK
NUMBER USED AND ENTER AT Q50 FOR THE APPROPRIATE ITEM(S). IF NO TO A AND B,
SKIP TO Q51.]
049
DK RE NO YES NUMBER
USED
a. window or wall mounted 94 97 02 01 —> | | |
unit(s)
b. swamp or evaporative coolers..94 97 02 01 —> | | |
51. On the average, what temperature do you usually keep your entire home in the
summer? [USE MARGIN/MARGINAL NOTES TO CALCULATE AVERAGE TEMPERATURE.]
I I I DEGREES
DK 94
RE ...97
15 March 20, 1989
-------
52a. Some stove fans or bathroom fans simply blow air through a filter, or blow
exhaust to an unvented attic. They do not blow air or exhausts out of the
house. We are only interested in exhaust fans which are vented to a vented
attic or to the outside.
Do you use these built in exhaust fans vented to the outside?
FOR EACH YES CIRCLED, ASK:
52b. Do you use this (NAME)...
1.
2.
3.
YES
cook stove
exhaust fans 01
YES
bathroom
exhaust fans 01
YES
clothes dryer 01
ON
02
NO
02
NO
02
DK
94
DK
94
DK
94
RE
97
RE
97
RE
97
WHENEVER
COOKSTOVE
USED
01
WHENEVER
BATHROOM
USED
01
DAILY
01
OCCASIONALLY
WHEN
COOKSTOVE
USED
02
OCCASIONALLY
WHEN
BATHROOM
USED
02
MORE THAN
ONCE PER WEEK
02
SELDOM
OR
ALMOST
NEVER
03
SELDOM
OR
ALMOST
NEVER
03
SELDOM
ALMOST
NEVER
03
DK
94
DK
94
OR
DK
94
RE
97
RE
97
RE
97
53. Does the home have a whole house exhaust fan which blows air out of the home?
YES 01
NO.
DK.
RE.
02
94
97
—> 54a
16
March 21, 1989
-------
53a. How frequently do you use the whole house fan in the cooling season:
Use every day...* 01
Not daily but more than once a week 02
Use regularly but less than once a week 03
Seldom or never use during cooling season 04
DK 94
RE 97
53b. How frequently do you use the whole house fan during other seasons?
Use every day 01
Not daily but more than once a week 02
Use less than once a week ,.. 03
Seldom or never use during other seasons 04
DK 94
RE 97
54a. Does this home have...
FOR EACH YES CIRCLED, ASK:
54b. What percent of your home has (TERM)?
Y N DK RE
1. double pane windows
2. storm windows
3. insulation in the walls
4. insulation in the ceiling
5. storm doors
6. weather stripping
01
01
01
01
01
01
02
02
02
02
02
02
94
94
94
94
94
94
97
97
97
97
97
97
PERCENT
DK RE
% 94 97
% 94 97
94 97
94 97
94 97
94 97
17
March 20, 1989
-------
55. All things considered regarding the insulation and weatherization of your home,
would you say that your home is tight, leaky, or are you uncertain about the
tightness of your home?
TIGHT 01
LEAKY 02
UNCERTAIN 03
RE 97
56. Which months of the year is your home usually closed up for heating season, that
is, your windows and doors are usually kept closed?
[CIRCLE ALL THAT APPLY]
Jan. Feb. Mar. Apr. May June July Aug. Sep. Oct. Nov. Dec.
01 02 03 04 05 06 07 08 09 10 11 12
57. Which months of the year is your home usually closed up for cooling season, that
is, your windows and doors are usually kept closed?
[CIRCLE ALL THAT APPLY]
Jan. Feb. Mar. Apr. May June July Aug. Sep. Oct. Nov. Dec.
01 02 03 04 05 06 07 08 09 10 11 12
58. During the time when your home is usually closed for the heating and cooling
seasons, you may take steps to ventilate your home. This may involve such
things as opening windows, or when cleaning, cooking, using insecticides, etc.
Which of the following best describes your ventilation practices during heating
and cooling seasons?
Do you....
ventilate on a daily basis 01
not daily, but more than once a week 02
regularly, but less than once a week 03
only during specific activities 04
do not ventilate at all 05
DK 94
RE 97
18 March 20, 1989
-------
• *
I—
z.
LU
O
O
a.
10
LU
ce.
o
h-
o
^C
LU
a:
\
i
m
H-
a.
i— t
o:
C-)
to
c
^ 01
«3 E
3 ••-
CO 4-1
3
0 T-
JZ Ol
4->
O)
— TJ
CL c
o o>
oi a.
a. co
JZ Ol
4-> JZ
3 O)
o s-
C/l
c Jx:
O CO
•r- 03
C/) C
O) O)
3 JZ
CT4->
o) XJ
E c:
O 03
CO
Ol
c/> a.
'O O
o a.
4-1
0)
TJ co
01 01
C 4->
1— 1 4->
< 3
•- O
C JD
O 03
TJ
O3 CO
i- C
o
O •!-
. CO
J- 3
3 CT
co
O XJ
Q. C
X 3
ft) O
. fc.
i— Ol
03 -^
3 0
TJ 03
•^ (~*
>
•r- 0)
XJ E
C O
•r— CO
*4— ^«£
O CO
CO
0) i—
4-* ^~
O3 *r—
E Z
•r-
4-* t— t
CO
0) 4->
CO
4-) J—
O> -r—
O)U-
0
i- S- co
O) O> 3
XJ JZ O
O O)
> Ol
C 'r- JZ,
)— ( ^— 4-1
r— i
LU
CD
a.
1—
x:
LU
^^
O
to
a:
LU
to
z
Q
cr
o
CJ
LU
a:
*— '
t>-
TJ
^™
O
ft)
tn
3
O
JZ
•r—
5
C
ftl
•r-
'•^
ftl
r—
a.
0
oi
Q.
>,
c
03
E
5
.
cn
in
c=
•r—
cn
O)
JD
CO
Ol
o>
03
C
s-
3
O
s_
O)
H-
XJ
'o
JZ
a>
CO
o
VI
4->
c
ft)
o
X
ft)
"o.
£
a.
JC
H- «-•
O 4^
ftl
•ft O*
i§
>
«-< ft)
u «-«
^ o
4^
ftl
jc c
«-> *
o
I10
J*
^» >W
5 ^
** TJ
- C
3 «
O
^> 4M
in
JC ft)
*J TJ
•^ ^"
jl O
Ol ft)
C JC
J-.JZ
03 **>
.
O
VD
ur/the1r] last birthday.
o
>>
c
o
ft)
o*
43
"v?
-
UJ
1
^^
u
3
O
^•^
•MH«
TJ
ftl
ftl
C
"1
X
ftl
z
,
^•4
VO
r female?
0
ftl
I^B
1
4MB*,
UJ
5
2L
IA
*"^
^»
VO
«c
00
dD
2
>
oo
O
i
u.
.
CM
VO
O
JZ
c
XJ
ft)
UJ
z
z
JC
a
o
D
>
*C
£.
i/)
,^
4vl
i
•o
c
fQ
|/)
(-
03
ft)
-^
C
03
O
,
CO
VO
p-
10
ft)
4^
4->
ftl
(_
O3
Cf
u
ft>
^x
i
^-^
LU
s:
^^
CO
O)
0
O
3
o
o
o
.
^^
VO
I—I
VO
Or
•z.
I-M
c
to
UJ
E
a:
^f
LU
ce.
o
u.
^,
VD
cr
Q
•^
^^
VO
cr
in
VO
o-
t^s*
to
•a:
01
o>
u
Q.
4J
03
01
3
TJ
O)
O
CO
LU"
2:
•z.
CO
O)
O
TJ
3
O
o
CO
4->
4->
Ol
01
O
CO
U
03
O.
^»
03
4-1
03
Q.
S_
O
JK:
u
03
CL
C
E
^
o
JZ
*
Ol
Ol
03
Ol
o
t
in
VO
03
TJ
jx:
Ol
Ol
3
O
CM
•r™
CL
Ol
C
si
3
TJ
(J
Ol
01
o
01
4-1
c
o
CO
Uj"
o
T3
3
0
o
XJ
CO
O)
O)
S-
03
Ol
"o
C
03
E
^
O
^
O)
O)
03
s-
Ol
03
C
o
.
f«^
VO
THROUGH 71 FOR R, THEN PROCEED WITH QUESTIONS 68 THROUGH 71 FOR EACH
R OF THE HOUSEHOLD, ONE PERSON AT A TIME]
CO UJ
VO CO,
to LU
HH _J
1— ^C
to Z
UJ O
Z3 I— '
o-t—
i— i
i«i O
CO Q
a:
I—
to
•z.
1— 1
a:
UJ
§
;>
i 1 1
z.
i— i
u?
z.
CO
01
o
TJ
3
O
O
-o
CO
3
0
03
O
03
TJ
a>
CD
3
O
1
CM
03
U
Q.
03
0)
C
•r-
J_
3
0
03
TJ
•r—
J-
u_
4= a>
en 01
3 03
0 S-
S- 0)
jz >
03 -C
TJ 4J
c .
o c
s: o
,_ ^
03 ftl
3 CO
CO ZS
3 O
03
ftl
3 4J
O
03 ••-
jxc TJ
C C
•i- ft)
J= CL
^
OO
VO .
O)
JZ
4-1
C
O
TJ
C
Ol
a.
co
UJ*
S
Z.
co
ai
o
•a
3
O
o
TJ
CO
S-
3
O
C
03
O
a>
o
Ol
JZ
4J
C
•r—
TJ
C
Ol
D.
CO
LU C-
Z V
«=C E
Z O
3
0 t-
>i 3
*— ' O
CO
3 O
O
Ol
>^ >
03 ai
XJ r—
01. — •
ai •
S u
^ ft>
j- ^.
ai XJ
J3 C
E' O
3' O
C O>
•>_^ co
JZ' CO
4-1 t-
^
(71
VO
LU
SI
Z.
CO
O)
O
TJ
3
O
o
TJ
CO
t-
3
O
J=
C
03
O
TJ
TJ
C
0
w
0)
3
o
1
CM
03
U
a.
03
en
c
s-
0
03
TJ
Z3 ^*
to ai
Ol
TJ 03
C i-
03 O)
£*
TJ ftl
5- JC
3 4-1
4->
03 C
to o
_ ^
03 O)
3 CO
CO 3
3 O
JZ
03
Ol
4-> JC
3 4->
O
J3 C
03 -r-
-* TJ
CZ C
T- 01
JZ Q.
1— CO
B
o
r-.
cn
CO
en
LU O
3E CM
o
CO S-
Ol 13
0 2!
XJ
3
O
o
XJ
(/)
s-
Z5
o
c
i
o
ro
XJ
TJ
CD
CD
03
O
a.
03 cn
i— i
1= C^.
O O)
J= J-
3
a> o
<+-
c o
XJ Ol
c >
ft) O)
Q-r—
CO
UJ U
s: *J
z -^
3 C
O O
>> o
tu
CO
co
l_ -~^
3 4->
O CO
-C I-
^M^l^
J- — •
a> .
jQ 0)
6 -C
=3 +->
C
O
a>
JC TJ
4-1 C
ft)
<*- a.
O co
• '
t— 4
-------
M
3 M
O •
IB >
o
c~
o
S M
« « P «
o> o o •>
4 89
H TJ O
« B — m
> « T
•< X M a
5 — •*
a
M
3
O 1 U «
tP M g a
o
> -H s
rt O M
g s ^
03 I Q
•e o
O H
»J
V O
C M
M h
MO
• «
te
a
8.5
1 s
§
j|
O
u>
o
iSSiJ.. .
t, ~ In O X
o o o o o
-------
72. Are you currently renting this home or is it owned or being bought by you?
OWN 01
RENT 02—> Landlord:
DK 94 Address:
RE 97
73. Finally we would like to ask about any other radon testing. Has your home been
tested for radon in the past?
YES 01 —> CONTINUE
NO 02 '
DK 94 —> SCRIPT 2
RE 97 .
74. How many measurements have been taken?
/
DK 94
RE 97
75. Did (any of) the tests indicate that the radon levels in your home were high or
elevated?
YES 01
NO 02
DK 94
RE 97
76. Have you had anything done to reduce the levels of radon in your home?
YES 01 —> CONTINUE
NO 02 '
DK 94 —-> SCRIPT 2
RE 97 .
77. What was done?
21
March 20, 1989
-------
SCRIPT 2
READ TO RESPONDENT
That's all the questions I have. There are just two more tasks I have to do — place
the detectors and get your name, mailing address, and telephone number.
To obtain accurate measurements, the detectors should remain in place for approxi-
mately 12 months. At that time, we will send you instructions on how to package and
return them. We will also send you postage-paid envelopes for their return.
PLACE (OTHER) DETECTOR AND ENTER INFORMATION BELOW.
DETECTOR IDI
#1
LEVEL
V H
#1.
DUPLICATE
DETECTOR ID
#2
1 2 #2.
#3
1 2 #3.
#4
01
02
03
Level One
Level Two
Level Three
CIRCLE ONE NUMBER BELOW TO INDICATE DETECTOR PLACEMENT.
PLACED BY RESPONDENT WITH INTERVIEWER DIRECTION...01
PLACED BY INTERVIEWER 02
PLACED BY RESPONDENT ALONE 03
REFUSED PLACEMENT 04
PHYSICALLY/MENTALLY INCAPACITATED 05
OTHER (EXPLAIN IN FULL BELOW) 06
DETECTORS
Originals Duplicates
01 02 03 04
01 01 01 01
02 02 02 02
03 03 03 03
04 04 04 04
05 05 05 05
06 06 06 06
01 02 03 04
01 01 01 01
02 02 02 02
03 03 03 03
04 04 04 04
05 05 05 05
06 06 06 06
INTERVIEWER:
WHEN YOU HAVE COMPLETED THE QUESTIONNAIRE, COMPLETE PART E OF THE CONTROL FORM.
22 March 20, 1989
-------
APPENDIX C
NATIONAL RESIDENTIAL RADON SURVEY
CONTROL/SCREENING FORM
-------
March 2, 1989
CONTROL/SCREENING FORM - NATIONAL RESIDENTIAL RADON SURVEY
OMB Number:
Expires:
PART A. HOUSEHOLD IDENTIFICATION
IF ADDED HU, CHECK HERE
AND ENTER LINK LINE NO.
ATTACH AND COMPLETE ADDED HU LABEL
IDENTIFICATION LABEL
PART B. RECORD OF CALLS
Day of Week
Date
Time
am
pm
am
pm
am
pm
am
pm
am
pm
am
pm
Results
Code*
FI No.
*CODES: 10 = No Answer 31 = Vacant
11 = Appointment 32 = Not a housing unit
12 = Refused 33 = HU Demolished or moved
13 = Callback 34 = Not Year Round Residence
20 = Complete, detectors left 35 = Moving within 12 months
21 = Complete, detectors not left 36 = Other (Specify Below)
30 = Terminated
PART C. INTRODUCTION AND SCREENING QUESTIONS
Hello. My name is (YOUR NAME). I'm an interviewer from Research Triangle Institute which is
located in North Carolina. We are conducting an important research project called the National
Residential Radon Survey, which is sponsored by the U.S. Environmental Protection Agency.
You may have seen reports about radon on.TV or in the newspaper. Many of these reports were
based on EPA sponsored State Radon Surveys. This survey is to determine household radon
levels, nationwide. Your household is among several from this area that have been randomly
selected through scientific sampling procedures. Your home will represent thousands of homes
in this part of the United States. A letter and other materials were sent to each household.
Did you receive these materials? IF NOT, HAND LETTER AND INFORMATION MATERIALS TO RESPONDENT
AND EITHER READ OR ALLOW TIME FOR READING. SHOW SAMPLE DETECTOR IF NECESSARY.
-------
First, I need to ask a few questions to find out if this household is eligible for the
survey.
1. Does anyone in this household live here at least nine or more months out of the year?
IF RECENT MOVE IN - Does anyone from this household plan to live here nine or more
months out of the year?
YES 01 -> CONTINUE.
NO 02 •» STOP, NOT ELIGIBLE. ENTER CODE 34 IN PART B.
2. Does your household have firm plans to move from this home within the next 12 months?
YES, FIRM PLANS TO MOVE . 01 * STOP, NOT ELIGIBLE. ENTER CODE 35 IN PART B.
NO FIRM PLANS TO MOVE . . 02 •» CONTINUE.
Your home is eligible for the study.
To answer the questions in the questionnaire, I need to talk with the head of the household or
another person who is responsible for the upkeep of the house. Are you that person?
IF "YES", CONTINUE WITH THE QUESTIONNAIRE.
IF "NO", ASK TO SPEAK TO A KNOWLEDGEABLE INFORMANT. IF NOT AVAILABLE, DETERMINE BEST
TIME TO RETURN AND END. DOCUMENT RESULT IN PART B.
WHEN KNOWLEDGEABLE INFORMANT AVAILABLE/CONTACTED, RE-READ THE INTRODUCTION AND CONTINUE
WITH THE QUESTIONNAIRE, PART D.
PART D. QUESTIONNAIRE ADMINISTRATION
PART E. RESPONDENT IDENTIFICATION
So that we can contact you about the return of the detectors, please give me your name,
correct mailing address, and telephone number.
NAME:
ADDRESS:
TELEPHONE: ( )
In case we can't get in touch with you, we need to have the name and telephone number of some-
one who does not live with you but who will always know how to get in touch with you.
NAME: Phone # ( )
Thank you for being a part of this important study. To show our appreciation for answering
our questions and placing the 12 months detectors, we want you to have this $5.00. So that I
may be reimbursed, please sign this receipt portion of the consent form. RTI will send you
the rest of the incentive as soon as the detectors are received after the 12 month exposure
period.
GIVE THE INCENTIVE AND INSTRUCTION SHEET TO RESPONDENT.
HAVE RESPONDENT SIGN AND YOU COMPLETE THE INCENTIVE RECEIPT PORTION OF THE CONSENT FORM.
I am going to leave this instruction sheet with you. This sheet has more information about
the detectors and it also has a phone number to call if you have questions now or at anytime
during the time the detectors are in place.
Thanks again for your help.
-------
-------
-------
Recycled/Recyclable • Printed on 100% Postconsumer, Process Chlorine Free Recycled Paper
that has been manufactured with Wind Power
------- |