Is It MAR or NMAR?

• Activity Number: 584
• Type: Topic Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Michael Sverchkov*+ and Danny Pfeffermann
• Companies: US Bureau of Labor Statistics and Hebrew University of Jerusalem, Israel and University of Southampton, United Kingdom
• Keywords: nonignorable non-response ; sample distribution ; sample-complement distribution ; prediction ; missing information principle

Abstract

Most methods that deal with the estimation of response probabilities assume either explicitly or implicitly that the missing data are 'missing at random' (MAR). However, in many practical situations this assumption is not valid, since the probability to respond often depends on the outcome value. The case where the missing data are not MAR (NMAR) can be treated by postulating a parametric model for the distribution of the outcomes under full response and a model for the response probabilities. The two models define a parametric model for the joint distribution of the outcome and the response indicator, and the parameters of this model can be estimated in principle by maximization of the likelihood corresponding to this distribution. Modeling the distribution of the outcomes under full response can be problematic since no data are available from this distribution. Sverchkov (2008, 2010) proposed a new approach that permits estimating the parameters of the model holding for the response probabilities without modelling the distribution of the outcomes under full response. In the present paper we show how this approach can be used for testing whether the response is MAR or NMAR