Regularized Structural Equation Modeling for Individual-Level Directed Functional Connectivity Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 20, 2019
  • Lane, Stephanie
    • Affiliation: College of Arts and Sciences, Department of Psychology and Neuroscience
  • Within functional magnetic resonance imaging (fMRI) research, one method for evaluating functional brain architecture is directed functional connectivity analysis. Given the potentially exploratory nature of directed functional connectivity modeling, data-driven strategies for identifying individual-level models are necessary. One promising method, the unified SEM, is rooted in the structural equation modeling framework. By representing both the lagged and contemporaneous directed relationships present among regions of interest, it allows for the estimation of individual-level models of connectivity. In this study, I present the regularized unified SEM as an alternative to existing methods, where an individual-level model is selected from a range of possible models with varying degrees of penalization. This method is compared to other existing methods for establishing directed functional connectivity, including an established stepwise model building procedure for the unified SEM as well as the graphical vector autoregressive model. In this evaluation, the regularized unified SEM using the adaptive LASSO outperforms all other methods on simulated time series data, as well as on simulated blood oxygen level dependent (BOLD) data. Performance is optimal in the presence of a long time series, a small number of variables, and a sparse network.
Date of publication
Resource type
Rights statement
  • In Copyright
  • Curran, Patrick
  • Bollen, Kenneth
  • Zeng, Donglin
  • Giovanello, Kelly
  • Gates, Kathleen
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2017

This work has no parents.