The Impact of multiple imputation for DACSEIS

This paper is designed to provide an extensive introduction to the principles of multiple imputation and to give some general recommendations of using multiple imputation techniques in the DACSEIS universes. The definition of an ignorable missingness mechanism is explained, and the concept of the observed-data likelihood is discussed. To introduce the multiple imputation principle a short introduction of Bayesian statistics is provided. A small simulation study is performed comparing different approaches to illuminate the advantages and disadvantages of different imputation techniques. Finally, an overview about recently available multiple imputation software is given and violations of the assumptions made are addressed