The Blocked Imputation Approach for Missing Data

Abstract

This paper introduces a new imputation method,Blocked Imputation (BI),which extends standard multiple imputation (MI) by including the reasons for attrition as part of the imputation procedure.MI and BI are compared by means of a simulation study,where a two-wave design with attrition was considered.The results show that the BI approach outperforms MI in the estimation of the mean and variance of the variable with missing data.Furthermore an empirical example was given to illustrate the applicability of the BI approach. Overall, BI represents a flexible and accurate new method in the missing data methodology.