A noninformative Bayesian approach for selecting a good post-stratification (Q1657966)

When there are several ways of stratifying a heterogeneous population using available prior information, one often faces the problem of choosing the best stratification for better estimation. Assuming that before the sample is selected a limited number of stratifications are well demarcated, the authors propose a Bayesian approach which allows the simultaneous utilization of available stratifications. In this approach given the sample, the posterior distribution assigns more weight to good stratifications and less to others. The data set \texttt{nhanes} available in \texttt{R} is used to illustrate how the suggested approach works well.