![IconGems-Print](images/IconGems-Print.png)
586 – Weighting and Variance Estimation
Survey Designs with Small Stratum Sample Sizes
Wayne A. Fuller
Iowa State University
Emily Berg
Iowa State University
Xiaofei Zhang
Iowa State University
One-per-stratum and Two-per-stratum designs are common survey sampling designs. The one-per-stratum design generally has the smallest variance, but no unbiased estimator of variance is available. We present an intermediate design which is a combination of the one-per-stratum and two-per-stratum designs, and for which an unbiased variance estimator is available. A closely related design is a one-per-stratum design with supplemental observations for variance estimation. We compare the variance and variance of estimated variance for versions of the intermediate design and for the supplement design.