Full Program »
139. Interpreting Bifactor Structures of Psychopathology: A Comprehensive Evaluation of Alternative Modeling Approaches
Bifactor models of psychopathology are highly visible in clinical science. Confirmatory factor analysis (CFA) is the most common multivariate technique for specifying bifactor models. While CFA offers straightforward hypothesis testing, this advantage is counterweighted by strict pre-specifications of structure onto the data, such as simple structure. Exploratory methods (EFA) allow for inspection of meaningful cross-loadings before estimating more restrictive CFA models and reporting good fit. Limitations of model fit is a timely topic given mounting evidence that bifactor CFAs tend to fit better than alternatives, a hint that they share significant similarities with EFAs. No study has characterized this relationship. The current study aims to explicate bifactor model specification principles through investigating a comprehensive set of parameterizations along the exploratory-confirmatory continuum. Using a nationally representative sample, we evaluated several EFA and CFA bifactor models using various statistical and interpretative criteria. All models fit well, but differences in substantive interpretability and reliability were notable. Several similarities emerged between bifactor and EFA models, including patterns of overfitting and anomalous parameter estimates. Alternative specifications are proposed. Evaluating structures of psychopathology requires multiple approaches, but a critical step is to understand what they can tell us based on alternative ways of thinking about them.