Biostatistics Seminar Series: Nicole M. Butera

Abstract:Missing data are common in longitudinal cohort studies and can lead to bias, particularly in studies with informative missingness. Common methods for handling missing data include inverse probability weighting, which requires correctly specifying a model for missingness, and imputation, which requires correctly specifying a model for the incomplete variable. Although doubly robust methods exist to provide unbiased regression coefficients in the presence of missing outcome data, these methods do not account for correlation due to clustering inherent in longitudinal or cluster-sampled studies. We developed a doubly robust method to estimate the regression of an outcome on a predictor in the presence of missing multilevel data on the outcome, which consistently estimates the regression coefficients assuming correct specification of either (1) the missingness probability or (2) the outcome model. This method involves specifying separate hierarchical models for missingness and for the outcome, conditional on observed auxiliary variables and cluster-specific random effects, to account for correlation among observations. This proposed estimator has been shown to be doubly robust, and its asymptotic distribution has been derived. Simulation studies will be presented comparing the proposed method to an existing doubly robust method that ignores the clustering, and the method will be applied to data from the China Health and Nutrition Survey, an ongoing multilevel longitudinal cohort study.