Testing multivariate normality in the presence of missing data

Abstract

Missing data are a very common problem in practice. Therefore providing an adequate methodology for statistical inference is of great importance for a wide scientific community. We focus on the problem of testing the hypothesis of multivariate normality in the presence of different missingness mechanisms. In particular, the focus will be on the impact of different imputation algorithms on some of the popular tailor-made normality tests - heir size and powers under commonly used alternatives, and their comparison with the complete-case approach. Finally, potential directions for future research will be discussed.