141 – Tree-Based Methods for Missing Data and Evaluation of Missingness Mechanisms
Enhanced Tipping-Point Displays
Victoria Liublinska
Harvard University
Donald B. Rubin
Harvard University
Assumptions about the missingness mechanism often cannot be assessed empirically, which calls for the sensitivity analyses. However, few studies with missing values are subjected to such analyses due to the lack of clear guidelines on a systematic exploration of alternative assumptions as well as the difficulty of formulating plausible missing not at random (MNAR) models. We present graphical displays, based on the "tipping-point" analysis first introduced in Yan et al. (2009), that help us visualize the results of a set of sensitivity analyses for missing outcomes in studies that compare two treatments. The resulting "enhanced tipping-point displays" provide compact summaries of conclusions drawn from different alternative assumptions about the missingness mechanism simultaneously. A recent use of these enhanced displays in a medical device clinical trial has helped lead to FDA approval.
