Questionnaire Dwelling Unit-Level and Person Pair-Level Sampling ...
Questionnaire Dwelling Unit-Level and Person Pair-Level Sampling ...
Questionnaire Dwelling Unit-Level and Person Pair-Level Sampling ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
outwinsor percentages represent the total amount of residual weight when the weights are<br />
trimmed to the critical values (used for ev definition) relative to the total sample weight. For<br />
evaluation purposes, the outwinsor percentage is considered the most important of the three<br />
percentages, as this gave a measure of the impact of winsorization (or trimming) of ev weights<br />
(if we performed this treatment). See Sections 5.1 <strong>and</strong> 5.2 for the domains that were used to<br />
define extreme values.<br />
8.3 Slippage Rates<br />
The slippage rate for a given domain is defined as the relative percentage difference<br />
between the sampling weights <strong>and</strong> the external control totals, both before <strong>and</strong> after ps. The<br />
control totals for QDU <strong>and</strong> person pair ps are derived from the screener dwelling unit (SDU)<br />
weights, which were poststratified to U.S. Bureau of the Census population estimates (Chen et<br />
al., 2008). Table F.1 displays QDU national domain-specific weight sums for both before <strong>and</strong><br />
after ps, as well as the desired totals to be met through ps. Table K.1 shows the same for the pair<br />
sample. These tables also show the relative percentage difference, or the amount of adjustment<br />
necessary (positive or negative) to meet the desired totals. The first relative difference is used<br />
explicitly during the ps modeling procedure to identify potential problems for convergence.<br />
Large differences in domains with relatively small sample sizes are indicative of potential large<br />
adjustment factors, which may cause problems in convergence while satisfying bound<br />
constraints. The reason is that adjustments required for one domain may have an adverse effect<br />
on another domain when a unit belongs to both.<br />
As an example, consider that Table F.1, for the 2006 QDU domain household size of one,<br />
indicates a sample size of 5,641 with a total design-based weight of 30,195,407 <strong>and</strong> a census<br />
total of 30,104,368 with an initial slippage rate of 0.30 percent, which would imply a common<br />
weight adjustment of ≈ 0.996985, if this were the only calibration control. Similarly, looking at<br />
pair data in Table K.1, the pair domain category of pair race other has a sample size of 956, a<br />
design-based weight of 12,778,059, <strong>and</strong> a census total of 13,369,146, showing an initial slippage<br />
of -4.42 percent. The resultant required adjustment would be ≈ 1.046258, if this were the only<br />
control. However, in the generalized exponential model (GEM), all controls are simultaneously<br />
satisfied under a complex algorithm that allows for different adjustment factors for different<br />
units.<br />
8.4 Weight Adjustment Summary Statistics<br />
Tables G.1, G.2, <strong>and</strong> L.1 through L.3 display summary statistics on the product of weight<br />
components before <strong>and</strong> after all stages of adjustment for the QDU <strong>and</strong> person pair, respectively.<br />
The summary statistics include sample size (n), minimum (min), maximum (max), median<br />
(med), 25 th percentile (Q1), 75 th percentile (Q3), <strong>and</strong> the unequal weighting effect (UWE). Note<br />
that in Tables L.2 <strong>and</strong> L.3 the sample size for pair age group, pair race/ethnicity, <strong>and</strong> pair gender<br />
are slightly different. This is because those variables were defined using screening demographic<br />
information in the nonresponse adjustment of respondent pairs, while in the poststratification of<br />
respondent pairs, they were defined from questionnaire demographic information. Because UWE<br />
is directly affected by weight adjustment factors <strong>and</strong> extreme weights, these values—along with<br />
the percentage of extreme weights as noted in Section 8.2—were used as guidelines for<br />
determining model adequacy.<br />
74