Mixed effects models for adolescent brain and cognitive development
- Smith, Diana Mun Yee
- Advisor(s): Jernigan, Terry L
Abstract
Adolescence is a critical period during which the brain undergoes structural and functional changes. The Adolescent Brain Cognitive Development℠ Study (ABCD Study®) presents unique opportunities and challenges for modeling longitudinal trajectories of these changes over time. In this dissertation I explore a novel statistical technique, Fast Efficient Mixed-Effects Algorithm (FEMA), for applying mixed-effects models to longitudinal datasets with participants nested within families. First, I present a comparison of the random effects estimation within FEMA to OpenMx for estimation of heritability and other components of phenotypic variance. Next, I present results of a similar investigation using FEMA to partition the variance of cortical phenotypes into components including family effects, genetic effects, and subject-level effects. Finally, I document a procedure for incorporating spline functions into the linear mixed-effects model framework and present results from two investigations using this method: the first presents age associations with brain morphometry at the regional, vertexwise, and voxelwise level; and the second presents associations between pubertal development and the intracellular compartment of the diffusion MRI signal modeled on restriction spectrum imaging. By first validating this statistical method against a gold standard, using it to ask a novel scientific question, and then extending the computational framework to incorporate curvilinear trajectories, this work provides a nuanced understanding of adolescent maturational changes in the brain, and also provides methodological resources for future work using complex modeling techniques in this and similar datasets.