Testing independence in incomplete data settings

Abstract

Here we consider the problem of testing independence in the presence of missing data issues and present the overview of recent results. A part of the talk are devoted to the adaptation of tests based on Kendall's τ and its generalizations, for randomly censored data. Their limiting null distributions, in different scenarios, are shown, accompanied by the proposal of bootstrap procedure that might be used for their approximation. In addition, the results of several empirical power studies are summarized and directions for further research are presented.