Project
Linking brain function and structure to phenotypes: does more data in fixed sample size lead to higher replicability?

Brain-wide association studies (BWAS) link brain structure and function data (e.g., fMRI) to various outcomes (e.g., phenotypes) but face challenges in reproducibility and replicability (Gell et al., 2023; Marek et al., 2022; Spisak et al., 2023). Current approaches often inefficiently use available data and inadequately account for measurement errors. In this project, we propose developing nuclear norm penalized principal covariate regression (PcovRnnp) for BWAS that 1) uses data more efficiently and 2) accounts for measurement errors. We plan to test the PcovR method with simulated data and new data acquired in the new Gravitation program Growing Up Together in Society (GUTS). We aim to show that the reproducibility and replicability can be improved by using new, more advanced, statistical analysis methods. These advanced methods use more data from each single participant, data that is available but that researchers choose not to take into consideration or do not know how to take into consideration simultaneously.

Supervisors
Prof. Dr. Mark de Rooij
Dr. Wouter Weeda

Financed by
NWO

Period
10 February 2025 – 39 February 20230