B0556
Title: Effect sizes and replicability in brain-wide association studies
Authors: Simon Vandekar - Vanderbilt University (United States) [presenting]
Kaidi Kang - Vanderbilt University (United States)
Jakob Seidlitz - University of Pennsylvania (United States)
Jonathan Schildcrout - Vanderbilt University (United States)
Ran Tao - Vanderbilt University Medical Center (United States)
A Alexander-Bloch - University of Pennsylvania (United States)
Jiangmei Xiong - Vanderbilt University (United States)
Megan Jones - Vanderbilt University (United States)
Richard Bethlehem - University of Cambridge (United Kingdom)
Abstract: Brain-wide association studies (BWAS) are a fundamental tool in discovering brain-behavior associations. Several recent studies showed that substantial sample sizes are required to improve the replicability of BWAS because standardized effect sizes (ESs) are much smaller than expected. A meta-analysis of a robust effect size index (RESI) is performed using 63 longitudinal and cross-sectional (CS) neuroimaging studies to demonstrate that optimizing study design is an important way to improve standardized ESs in BWAS. The results indicated that the BWAS with low variability in covariate sampling distribution has smaller ES estimates and that longitudinal studies have systematically larger standardized ESs than CS studies. A CS-RESI is proposed to adjust for this systematic difference and quantify the benefit of conducting a longitudinal study and used bootstrapping to show that increasing between-subject variability by implementing different sampling schemes systematically increased standardized ESs. Results provide practical suggestions for future studies regarding improving ESs and replicability. The findings underscore the importance of considering design features in BWAS and emphasize that increasing sample size is not the only approach to improve ESs in BWAS.