Date of Award

Summer 2025

Document Type

Open Access Dissertation

Department

Epidemiology and Biostatistics

First Advisor

Alexander C. McLain

Abstract

Lesion-symptom mapping (LSM) studies offer insight into the brain areas involved in various aspects of cognition. This is commonly done via behavioral testing in patients with a naturally occurring brain injury or lesions (e.g., strokes or brain tumors). This results in high-dimensional observational data where lesion status (present/absent) is non-uniformly distributed, with some voxels having lesions in very few (or no) subjects. In this situation, mass univariate hypothesis tests have severe power heterogeneity where many tests are known a priori to have little to no power. Additionally, high-dimensional observational data can be grouped according to brain anatomical structure.

In this dissertation, three multiple testing approaches using side-information for LSM studies are proposed to increase testing power under the control of FDR. Chapter two employs the weighted adaptive Benjamini-Hochberg (WABH) procedure, which utilizes p-value weighting based on side-information (e.g., information on power heterogeneity). The weights are created using the distribution of lesion status and spatial information to estimate different non-null prior probabilities for each hypothesis test through some common approaches. We provide a monotone minimum weight criterion, which scales the overall spread of the weights to ensure monotonicity while minimizing the impact of weighting. In Chapters 3 and 4, we introduce HIMTEST and iHIMTEST, which both incorporate hierarchical group information into local false discovery rate (lfdr) procedures with a group-adjusted two-class model. These procedures are helpful when there is a natural way to divide all tests into different groups. HIMTEST utilizes group-level side information, which can provide insights into heterogeneity in the power or non-null probability of different tests. iHIMTEST aims to improve HIMTEST by utilizing individual side information to enhance the accuracy of the lfdr estimation when there are significant differences in conditional non-null probabilities within each group. All methods are demonstrated on simulated data and an aphasia study investigating the regions of the brain that are associated with the severity of language impairment among stroke survivors. The results demonstrate that the proposed methods have robust error control and can increase power.

Rights

© 2025, Siyu Zheng

Download

Included in

Biostatistics Commons

Share

COinS