Issues in the Validation of the General Factor of Psychopathology

There is increasing interest in a General Psychopathology factor that has been invoked to explain the comorbidity among disorders and symptom dimensions across the Externalizing and Internalizing domains, as well as unresolved... [ view full abstract ]

There is increasing interest in a General Psychopathology factor that has been invoked to explain the comorbidity among disorders and symptom dimensions across the Externalizing and Internalizing domains, as well as unresolved issues in the classification and etiology of psychopathology. Despite this enthusiasm, various researchers have questioned the validity and value of the General Psychopathology factor, and the bifactor model that is most often used to characterize it. Using both simulations and real data from 1,568 9-17 year-old twin pairs, we examined several issues in the validation of the General factor of psychopathology. We contrasted the fit and external validity of a bifactor model containing a General and two (i.e., Externalizing, Internalizing) or three (i.e., Externalizing, Fears, Distress) correlated specific factors with a model that only contained the two (or three) correlated factors. We used several criteria to evaluate and compare the validity and utility of these two competing models. In factor analyses, these included various indices of model fit and construct replicability, as well as the sensitivity of the General factor to its constituent first-order symptom dimensions. In external validity analyses, we examined discriminant validity of the factors by characterizing the percent of variance in and patterns of relations with relevant criteria accounted for by the bifactor versus the correlated factors models. In simulations, we examined model fit, statistical power, and parameter bias across the same models. We found evidence for better fit of, and greater discriminant validity in, the bifactor than the correlated factors model and found that the nature of the General factor was largely insensitive to the inclusion of any particular symptom dimension. These validation criteria and our results have implications for the existence and nature of the General psychopathology factor and for evaluating models of the classification of psychopathology in general.