Date created: | Last Updated:

Category: Project

Description: Recovering the number of factors in multivariate data is at the crux of psychological measurement. Factor analysis has a long tradition in the field but it’s been challenged recently by exploratory graph analysis (EGA), an approach based on network psychometrics. EGA first estimates a regularized partial correlation network using the graphical least absolute shrinkage and selection operator (GLASSO), and then applies the Walktrap community detection algorithm, which identifies communities in the network. Simulation studies have demonstrated that EGA has comparable or better accuracy for recovering the number of factors than contemporary state-of-the-art factor analytic methods (e.g., parallel analysis). Despite EGA’s effectiveness, there has yet to be an investigation into whether other network estimation methods or community detection algorithms could achieve equivalent or better perfomance. Furthermore, unidimensional structures are foundational to psychological measurement yet they have been sparsely studied in simulations using different algorithms. In the present study, we performed a Monte Carlo simulation using correlation matrix, GLASSO, and two variants of a non-regularized partial correlation network estimation method with several community detection algorithms. We examined the performance of these method-algorithm combinations in both continuous and polytomous data across a variety of conditions (including unidimensional structures). Our goal was to examine whether the network estimation and community detection components of EGA are optimal for recovering the number of factors in psychological data. The results indicate that the Fast-greedy, Louvain, and Walktrap algorithms paired with the GLASSO method were consistently among the most accurate and least biased across conditions.