Improving Exploratory Graph Analysis and Network Psychometrics: Community Detection Optimization and Metric Invariance for Cross-Sectional and Intensive Longitudinal Data

This dissertation is a collection of four studies aiming to optimize community detection and measurement model accuracy in the Exploratory Graph Analysis (EGA) framework. The first study addresses the optimization of the Walktrap algorithm, a commonly used community detection algorithm employed by EGA. The second study establishes a method for testing metric invariance in the EGA framework using cross-sectional data. The goal is to now extend this method for intensive longitudinal data. However, network loadings have yet to be investigated in the dynamical case. Therefore, the third study investigates the relationship between true population loadings and recovered network loadings from a Dynamic Exploratory Graph Analysis (DynEGA) model using multivariate time series data. A method for adjusting network loadings in the DynEGA framework is proposed. The fourth study uses these adjusted network loadings to extend the metric invariance permutation testing model into the DynEGA framework. Together, this collection of studies provides methodology for optimizing community detection within the EGA framework as well as powerful techniques for metric invariance in both cross-sectional and intensive longitudinal data.