511 – Inference and Variance Estimation - 2
A Design Effect Measure for Calibration Weighting in Cluster Samples
Kimberly Henry
IRS
Richard Valliant
University of Michigan and University of Maryland
We propose a model-based extension of weighting design-effect measures for two-stage sampling when calibration weighting is used. Our proposed design effect measure captures the joint effects of a nonepsem sampling design, unequal weights produced using calibration adjustments, and the strength of the association between an analysis variable and the auxiliaries used in calibration. The proposed measure is compared to existing design effect measures in an example involving household-type data.