In evaluating my Confirmatory Factor Analysis (CFA) model, I focused on CFI (Comparative Fit Index) and RMSEA (Root Mean Square Error of Approximation) as the two main fit ...
In evaluating my Confirmatory Factor Analysis (CFA) model, I focused on CFI (Comparative Fit Index) and RMSEA (Root Mean Square Error of Approximation) as the two main fit indices.
For the CFI, I used a threshold of ≥ .95 to indicate good model fit. This cutoff is widely recommended in the literature, especially by Hu and Bentler (1999), who suggested that values above .95 represent a well-fitting model.
For the RMSEA, I followed the commonly accepted standards:
- < .05 = close fit
- < .08 = acceptable fit
- < .06 = good fit (stricter criterion)
I personally used RMSEA < .08 as acceptable and < .06 as ideal, because many scholars acknowledge that strict thresholds may vary depending on sample size and model complexity.
In my model, the results were:
- CFI = .94
- RMSEA = .067
Based on these values, the model almost met the ideal standard. The CFI is slightly below .95, but still very close, while the RMSEA falls within the acceptable range below .08.
Therefore, I would consider this model acceptable rather than perfect.
I believe a model that “almost” fits can still be accepted, especially when the theoretical framework is strong and the factor loadings are meaningful. However, I would also review modification indices and cross-loadings before making final conclusions.
I respectfully think that relying only on rigid cutoffs may be problematic because recent discussions suggest that fit indices should be interpreted in context rather than as absolute rules