Bayesian outcome selection modelling (2203.11439v1)

Abstract: Psychiatric and social epidemiology often involves assessing the effects of environmental exposure on outcomes that are difficult to measure directly. To address this problem, it is common to measure outcomes using a comprehensive battery of different tests thought to be related to a common, underlying construct of interest. In the application that motivates our work, for example, researchers wanted to assess the impact of in utero alcohol exposure on child cognition and neuropsychological development, which were evaluated using a range of different tests. Statistical analysis of the resulting multiple outcomes data can be challenging, not only because of the need to account for the correlation between outcomes measured on the same individual, but because it is often unclear, a priori, which outcomes are impacted by the exposure under study. While researchers will generally have some hypotheses about which outcomes are important, a framework is needed to help identify outcomes that are sensitive to the exposure and to quantify the associated treatment or exposure effects of interest. We propose such a framework using a modification of stochastic search variable selection (SSVS), a popular Bayesian variable selection model and use it to quantify an overall effect of the exposure on the affected outcomes. We investigate the performance of the method via simulation and illustrate its application to data from a study involving the effects of prenatal alcohol exposure on child cognition.