Analysis of Functional Magnetic Resonance Imaging Data: A Comparison of Extended Unified Structural Equation Modeling (euSEM) and Generalized Structured Components Analysis (GSCA)
Abstract
Structural Equation Modeling (SEM) is a broad family of models which can be utilized to assess the relationship between latent variables. In the context of fMRI analysis, extended unified SEM (euSEM) and Generalized structured components analysis (GSCA) represent two different SEM approaches. This dissertation examined how these models perform under non-normality, low correlations between indicators, and model misspecification via simulations. The four criteria used to assess the performance were bias, Type I error, power, and convergence percentage. The results from the simulations were applied to an empirical dataset. In summary, GSCA was found to be more powerful but also more biased than euSEM. This was true for all levels of normality, correlation, and model specification.
Subject Area
Quantitative psychology|Medical imaging
Recommended Citation
Lipton, Abraham K, "Analysis of Functional Magnetic Resonance Imaging Data: A Comparison of Extended Unified Structural Equation Modeling (euSEM) and Generalized Structured Components Analysis (GSCA)" (2022). ETD Collection for Fordham University. AAI29170023.
https://research.library.fordham.edu/dissertations/AAI29170023
