CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Sensitivity analysis for calibrated inverse probability of censoring weighted estimators under non-ignorable dropout Authors:  Li Su - University of Cambridge (United Kingdom)
Shaun Seaman - University of Cambridge (United Kingdom) [presenting]
Sean Yiu - Roche (United Kingdom)
Abstract: Inverse probability of censoring weighting (IPCW) is a popular approach for dealing with dropout in longitudinal studies. The weights can be estimated by specifying a model for the probability of dropout and estimating its parameters using maximum likelihood. More recently, calibrated IPCW estimators have been proposed. These use weights that directly optimize covariate balance in the weighted sample, an approach which has been shown to reduce the mean-squared error of the IPCW estimator. Existing calibrated IPCW estimators are based on the unverifiable assumption of sequential ignorability, and sensitivity analysis strategies to violation of this assumption have been lacking. We shall describe an approach to sensitivity analysis for calibrated IPCW estimators under non-ignorable dropout. We shall illustrate its use on data from an international cohort study of systemic lupus erythematosus.