B1087
Title: Bayesian analysis of longitudinal dyadic/multiple outcome data with informative missing data
Authors: Jaeil Ahn - Georgetown University (United States) [presenting]
Abstract: Analysis of longitudinal dyadic/multiple outcomes with missing data is challenging due to the complicated correlations within and between dyads/multiple outcomes, as well as non-ignorable missing data. We will introduce a Bayesian mixed-effects hybrid model to analyze longitudinal dyadic data with non-ignorable dropouts/intermittent missingness. To address this, we factorize the joint distribution of the measurement, random effects, and dropout processes into three components. The proposed model accounts for the dyadic interplay using the concept of actor and partner effects as well as dyad-specific random effects. We evaluate the performance of the proposed methods using a simulation study, and apply our method to longitudinal dyadic datasets that arose from a prostate cancer trial. We will introduce a Bayesian mixed-effects selection model to analyze the multivariate quality of life data with non-ignorable missing data. Compared to the first model, we first describe the overall/local effects of predictors on outcomes simultaneously and then incorporate a variable selection feature in the missing data mechanism to evaluate the impact of potentially moderate to high dimensional outcomes on missing data mechanisms.