Extensions to the Two-stratum Model for Sampling Rare Subpopulations in Telephone Surveys
Boyle, W. R. (2005, August). Extensions to the Two-stratum Model for Sampling Rare Subpopulations in Telephone Surveys. Presented at Joint Statistical Meetings, Minneapolis, MN.
Population subgroups defined by demographic and other characteristics are often an important focal point of samples in telephone surveys. We consider a class of two-stratum telephone sample designs where part of the frame with higher subgroup concentration (the first stratum) is disproportionately sampled compared to the rest of the frame (the second stratum). The relative intensity of sampling in the first versus the second stratum (r) thus determines the gain in nominal subgroup sample size. Using proportionate sampling as the referent standard, we first compare the effect of r on the nominal and effective change in sample sizes. We then develop the optimum solution for r considering the dampening effect of variable sample weights on effective sample size (due to the varying sampling intensity between strata). Finally, as sample attrition and thus unit costs vary between strata, we also develop a solution for optimum r considering both variable weights and cost. In all findings, we consider the impact of the correlation between the sample weights and key survey measurements.