摘要：In this paper we describe MIDAS: a SAS macro for multiple imputation using distance aided selection of donors which implements an iterative predictive mean matching hot-deck for imputing missing data. This is a flexible multiple imputation approach that can handle data in a variety of formats: continuous, ordinal, and scaled. Because the imputation models are implicit, it is not necessary to specify a parametric distribution for each variable to be imputed. MIDAS also allows the user to address the sensitivity of their inferences to different assumptions concerning the missing data mechanism. An example using MIDAS to impute missing data is presented and MIDAS is compared to existing missing data software.
