Hot Deck Imputation Procedure Applied to Double Sampling Design

Abstract

From an annual sample of U.S corporate tax returns the U.S Internal Revenue Service provides estimates of population and subpopulation totals for several hundred financial items The basic sample design is highly stratified and fairly complex Starting with the 1981 and 1982 samples the design was altered to include double sampling procedure This was motivated by the need for better allocation of resources in an environment of shrinking budgets Items not observed in the subsample are predicted using modified hot deck imputation procedure The present paper describes the design estimation and evalua tion of the effects of the new procedure