B1867
Title: A new approach to Little's MCAR test
Authors: Danijel Aleksic - University of Belgrade (Serbia) [presenting]
Abstract: Little's MCAR test is a very popular method for determining if data is missing completely at random (MCAR). A test for MCAR is constructed in the special case of monotone missing data. The test is based on the estimator of the covariation of the data itself and the response indicator of complete data columns. A connection between Little's and novel statistics is established and those are compared from various points of view.