Statistical calculations are performed using the survey package in R (https://cran.r-project.org/web/packages/survey/index.html).
With the exception of the unweighted count, estimates displayed in the table are weighted using the analysis weight variable ANALWT_C.
Standard errors and confidence intervals reflect the survey's stratified, clustered design. They are computed using the Taylor series linearization method, assuming a with-replacement design. Strata are defined by the variable VESTR, and clusters are defined by the variable VEREP.
Confidence intervals are computed at the 95% level.
The degrees of freedom for confidence intervals are calculated as the number of clusters minus the number of strata. Unlike the default approach in the R survey package, the strata and cluster counts are taken from the full sample even for domain estimates (that is, estimates based on a subset of the full sample).
Confidence intervals for percentages are constructed on the logit scale. This produces asymmetric intervals that are more accurate near 0% or 100% than symmetric intervals would be.
Chi-squared tests for two-dimensional tables are based on differences between the observed values and the values expected under the assumption of independence. They are calculated using the statistic="Wald" option in the survey package's svychisq function.