Bifactor exploratory structural equation modeling: A meta-analytic review of model fit

Abstract

Multivariate behavioral research often focuses on latent constructs—such as motivation, self-concept, or wellbeing—that cannot be directly observed. Typically, these latent constructs are measured with items in standardized instruments. To test the factorial structure and multidimensionality of latent constructs in educational and psychological research, Morin et al. (2016a) proposed bifactor exploratory structural equation modeling (B-ESEM). This meta-analytic review (158 studies, k = 308, N = 778,624) aimed to estimate the extent to which B-ESEM model fit differs from other model representations, including confirmatory factor analysis (CFA), exploratory structural equation modeling (ESEM), hierarchical CFA, hierarchical ESEM, and bifactor-CFA. The study domains included learning and instruction, motivation and emotion, self and identity, depression and wellbeing, and interpersonal relations. The meta-analyzed fit indices were the χ2/df ratio, the comparative fit index (CFI), the Tucker-Lewis index (TLI), the root mean square error of approximation (RMSEA), and the standardized root mean squared residual (SRMR). The findings of this meta-analytic review indicate that the B-ESEM model fit is superior to the fit of reference models. Furthermore, the results suggest that model fit is sensitive to sample size, item number, and the number of specific and general factors in a model.

Validation of the Italian version of the Neuroception of Psychological Safety Scale (NPSS)

“…Interestingly, this pattern is reflected in our ESEM models from two to three factors. In particular, CFI and TLI changed from 0.865 to 0.867 to 0.945 and 0.934, respectively; while χ2/df ratio and RMSEA changed from 2.68 to 0.055 to 1.84 and 0.035, respectively [ 79 ]. Regarding newly generated items, saturation per item is recommended to be > 0.5, whereas for existing items, saturation for each item is recommended to be ≥ 0.6 [ 80 ].…”

The Gothenburg Trismus Questionnaire in China: Cross‐cultural adaptation and measurement invariance

“…Therefore, the CFA results for the female group are considered to be within an acceptable range 47 . This may be attributed to a multitude of factors, such as gender‐specific traits in women, sociocultural influences, and model heterogeneity 48,49 …”

“…47 This may be attributed to a multitude of factors, such as genderspecific traits in women, sociocultural influences, and model heterogeneity. 48,49 The present study's innovative contribution is the examination of measurement invariance across different clinical status and genders, which was not mentioned in the previous GTQ. This study found that the configural invariance of the GTQ scale was established in different clinical status (H&N cancer and TMD), which indicated that the scale can effectively capture traits and structures related to trismus regardless of an individual's clinical status.…”