Identifying Psychometric Problems using Exploratory Graph Analysis

This primer introduces Exploratory Graph Analysis (EGA) and associated network psychometric methods to identify common problems that affect the accurate recovery of dimensional structures in psychological assessments. Using a simulated dataset, we show how Bootstrap EGA can identify structural instability often caused by issues such as low loadings, multidimensionality, local dependence, and minor dimensions. Low stability items are then diagnosed for these issues using a combination of network loadings and Unique Variable Analysis. To increase the accessibility of this pipeline, these approaches are consolidated into a single R function, `itemDiagnostics`, in the {EGAnet} package in R to provide an all-in-one function for comprehensive psychometric assessment. This function was then applied to an open-source empirical sample ($N$ = 2,082) that completed items from an inventory on everyday creativity (Creative Behavior Inventory). This analysis revealed an unstable dimension due to achievement-oriented, rather than everyday, creativity that formed a minor dimension. After handling this minor dimension, a robust four-dimensional structure was identified. This primer provides a conceptual background and practical demonstration of how these methods can be interpreted and used so that researchers across psychology can benefit from contemporary psychometric methods in network analysis.