Testing Criterion Validity in Hierarchical Models of Psychopathology: Comparison of Latent-Variable and Factor-Score Approaches

The Hierarchical Taxonomy of Psychopathology is a quantitative diagnostic system that is gaining traction as a framework for studying the correlates of mental-health problems. However, it remains unknown how best to operationalize hierarchically related psychopathology dimensions during criterion validity tests. In a series of simulations, we evaluated the performance of latent-variable (i.e., structural equation modeling [SEM]) and factor-score representations of hierarchical psychopathology constructs in criterion validity analyses. In models based on continuously distributed psychopathology indicators (e.g., symptom composites), SEM and factor-score methods both tended to yield unbiased estimates of criterion validity coefficients. In contrast, for models based on dichotomous indicators (e.g., categorical diagnoses), SEM led to more accurate estimates than factor scores in most cases. We offer recommendations for psychopathology researchers based on these results and provide an R function (https://osf.io/u3j5d/) that investigators can use to apply the approaches studied here in real-world data sets.