CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Bayesian sensitivity analysis for a missing data model Authors:  Bart Eggen - Delft University of Technology (Netherlands) [presenting]
Aad van der Vaart - TU Delft (Netherlands)
Stephanie van der Pas - Amsterdam University Medical Centres (Netherlands)
Abstract: In many fields, sensitivity analysis is very important to assess the robustness of study conclusions to key assumptions. We consider the missing outcomes model and perform sensitivity analysis under the assumption that missing outcomes are missing completely at random. We provide theoretical guarantees for a Bayesian approach to estimating the mean outcome, conditional on the sensitivity parameter. We show two Bernstein-von Mises theorems for different parametrisations of the model. The results are obtained using Dirichlet process priors on the distribution of the outcome and on the distribution of the outcome conditional on being observed. We also provide a simulation study, showing the performance of the methods in finite sample scenarios.