A note on post-stratification when analyzing binary survey data subject to nonresponse

Abstract

In this article we presents some results regarding the effect of post-stratification when analyzing binary survey data to nonresponse. Assuming that nonresponse depends on the variable of interest which is missing for nonrespondents. we show that sometimes the relative reduction in the bias due to post-stratification can be estimated from the respondents alone. This provides a simple sensitivity analysis for the potential bias of the post-stratified estimator, under a departure from the nonresponse assumption required for its unbiasedness. We illustrate with an example from the Norwegian Labour Force Survey.