Estimating the dimensionality of intelligence like data using Exploratory Graph Analysis

•A new method for estimating intelligence dimensions, Exploratory Graph Analysis, is presented.•Real and simulated data sets are analyzed by Exploratory Graph Analysis and other methods.•Exploratory Graph Analysis is superior to exploratory and confirmatory factor analysis.

This study compared various exploratory and confirmatory factor methods for recovering factors of cognitive test-like data. We first note the problems encountered by several widely used methods, such as parallel analysis, minimum average partial procedure, and confirmatory factor analysis, in estimating the number of dimensions underlying performance on test batteries. We then argue that a new method, Exploratory Graph Analysis (EGA), can more accurately uncover underlying dimensions or factors and demonstrate how this method outperforms the other methods. We use several published data sets to demonstrate the advantages of EGA. We conclude that a combination of EGA and confirmatory factor analysis or structural equation modeling may be the ideal in precisely specifying latent factors and their relations.