EFA | CFA | ESEM | |
---|---|---|---|
Theory/Data | Data driven approach, (exploratory) | Theory-driven, (confirmatory) | Fundamentally, confirmatory technique that integrates exploratory elements |
Item-Factor Loadings | Cross-loadings are allowed and not fixed. loadings are freely estimated | Cross-loadings are not allowed Unique specification of items onto respective latent factors applies | Non-target cross-loadings are constrained to be as close to zero as possible, but are still allowed |
Factor Structure & Parsimony | Complex factorial structures may emerge. Issues with parsimony may be present | Parsimonious models. Simple/clear factorial structures (sometimes criticised as overly simplistic) | Complex factorial structures, especially, in large datasets. However, more control applies compared to EFA |
Interpretability Risks | Extracted factors may not always be meaningful | Despite the increased insight into scale-scoring, adequate item loadings and high levels of reliability it provides, positively biased factor correlations and lower goodness of fit may be present | With non-zero cross-loadings, the bias and inflated statistics are reduced |