Reliability and Validity of Bifactor Models of Dimensional Psychopathology in Youth from three Continents

Abstract

Bifactor models are a promising strategy to parse general from specific aspects of psychopathology in youth. Currently, there are multiple configurations of bifactor models originating from different theoretical and empirical perspectives. Our aim is to identify and test the reliability, validity, measurement invariance, and the correlation of different bifactor models of psychopathology using the Child Behavior Checklist (CBCL). We used data from the Reproducible Brain Charts (RBC) initiative (N=7,011, ages 5 to 22 years, 40.2% females). Factor models were tested using the baseline data. To address our aim, we a) mapped the published bifactor models using the CBCL; b) tested their global model fit; c) calculated model-based reliability indices. d) tested associations with symptoms' impact in everyday life; e) tested measurement invariance across many characteristics and f) analyzed the observed factor correlation across the models. We found 11 bifactor models ranging from 39 to 116 items. Their global model fit was broadly similar. Factor determinacy and H index were acceptable for the p-factors, internalizing, externalizing and somatic specific factors in most models. However, only p- and attention factors were predictors of symptoms' impact in all models. Models were broadly invariant across different characteristics. P-factors were highly correlated across models (r = 0.88 to 0.99). Homotypic specific factors were also highly correlated. Regardless of item selection and strategy to compose CBCL bifactor models, results suggest that they all assess very similar constructs. Our results provide support for the robustness of the bifactor of psychopathology and distinct study characteristics.

The development of a Brief Working Alliance Inventory for clients and therapists using multilevel factor analysis and item response theory in the United States and China.

“…This specification allowed for the examination of the extent to which a general latent WA factor could explain the total variances among all the 12 items, while also accounting for the “nuisance variances” due to the three specific grouping factors along which the items were originally written. The WAI measure would be considered as having a strong enough, predominant general factor to “justify a unidimensional measurement model” (Rodriguez et al, 2016, p. 137) with the specific grouping factors accounted for, if the bifactor model showed the best overall model fit, and the model-derived indices reached proposed thresholds: for the general WA factor, its omega hierarchical (ω H ) > 0.8, explained common variance > 0.7, and percent of uncontaminated correlations > 0.7 (Hoffmann et al, 2021; Rodriguez et al, 2016).…”

Creating a brief form of the Supervisory Working Alliance Inventory in the Chinese cultural context using multilevel factor analyses.

“…Specifically, the bifactor model was specified on both the within- and between-trainee levels as (a) a general factor (i.e., general working alliance) loaded on all six items representing shared variance among all of them, (b) two specific factors loaded on their corresponding items (i.e., Items 1, 3, and 7 on the Rapport factor, and Items 6, 16, 17 on the Client Focus factor) to explain the shared variance over and above the general factor among the subset of items, and (c) the general factor and two specific factors were set to be orthogonal to each other. Following Rodriguez et al (2016) and Hoffmann et al (2021), we used a set of various indices to determine the dimensionality on each level, including hierarchical omega (ωH), explained common variance (ECV), and percent uncontaminated correlations (PUCs). When ωH is >0.8 and ECV and PUC are >0.7, the SWA construct can be interpreted as unidimensional.…”