Optimal design for generalized raking estimators

 

Seminar

Abstract

Two-phase designs measure variables of interest on a subset of cohort where the outcome and covariates are readily available or cheap to collect for the entire cohort. Since the variables of interest are often cost-prohibitive, it is of interest to find an optimal design which will include more informative individuals in the final sample. However, the optimal design will be different for different estimators. In this talk, I will present the optimal design for generalized raking estimators.