ABSTRACT

The chapter begins with a discussion of how various procedures in SAS® software treat missing data. Unit nonresponse in surveys, where no data are collected for a unit in the sample, is usually treated through weighting class adjustments or calibration, and the chapter provides references to published macros that perform raking and calibration weighting. Imputation is commonly used for item nonresponse, where a unit is missing data for some but not all of the items in a survey. A simple illustration is given of using the SURVEYIMPUTE procedure to perform a random hot-deck imputation, and the reader is pointed to resources for using the SURVEYIMPUTE and MI procedures to perform fractional and multiple imputation.