ABSTRACT
Missing data are ubiquitous problems in the analysis of survey data. Statistical imputation procedures are techniques for assigning analyzable values to item missing data on survey variables. This chapter talks about some important missing data concepts and then provides an overview of imputation approaches to analysis with missing data, outlining the general framework of method choice, model specification (where applicable), and practical methods for imputation, estimation, and inference for complex sample survey data. It introduces the MI method, including selection of the imputation model, multivariate MI methods and software, and estimation and inference with multiply imputed data. The chapter presents an introduction to Fractional Imputation (FI) theory, methods, and software. The chapter concludes with applications of the MI and FI methods to a missing data problem encountered in an analysis of the 2011–2012 National Health and Nutrition Examination Survey (NHANES) data on diastolic blood pressure (DBP) and hypertension.
