|
Activity Number:
|
268
- SBSS Student Paper Competition II
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 9, 2022 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #322179
|
|
|
Title:
|
Bayesian Image Mediation Analysis
|
|
Author(s):
|
Yuliang Xu* and Jian Kang
|
|
Companies:
|
University of Michigan and University of Michigan
|
|
Keywords:
|
Spatially varying mediation effects;
Soft thresholded Gaussian process;
Posterior consistency
|
|
Abstract:
|
Mediation analysis aims to separate the indirect effect through mediators from the direct effect of the exposure on the outcome. It is challenging to perform mediation analysis with high dimensional, correlated, and sparse mediators on the outcome. We develop mediation models with the soft-thresholded Gaussian processes priors for Bayesian image mediation analysis (BIMA). The proposed model aims to make inferences on sparse and piecewise smooth spatially-varying natural indirect effects mediated through brain images. We develop an efficient posterior computation algorithm for BIMA. We also perform rigorous theoretical analysis of BIMA by establishing posterior consistency for mediation effects and selection consistency to identify brain regions contributing to mediation effects. Simulation studies demonstrate that our proposed method has better accuracy and computational efficiency over the existing methods with high-dimensional correlated mediators. We apply the proposed method to the Adolescent Brain Cognitive Development (ABCD) study and estimate the effects of the parental education level on the children's cognitive ability mediated through the working memory brain activities.
|
Authors who are presenting talks have a * after their name.
