Title: Conservative approaches to integrative high-dimensional studies of the brain and brain-related measurements
Authors: Brian Avants - University of Virginia and Invicro (United States) [presenting]
Abstract: Wide data collection of neuroimaging and brain-related measurements is increasingly easy in both research and clinical setting. However, this leads to several issues: siloing of analyses, $p$-hacking and other statistical sins that may inadvertently promote irreproducible results. We describe how we use pre-defined integrative statistical models to help reduce multiple modalities to much smaller spaces, identify the reliable signal in datasets with relatively few subjects relative to the number of predictors and how we interpret these results using domain knowledge to enhance systems-level understanding of the brain.