I am conducting confirmatory factor analysis using lavaan in R with several both first- and second-order factors in the model (20 latent constructs and 97 items in total; n = 511; no missing values). After running the CFA, the model fit is not sufficient, but some low factor loadings, a couple Heywood cases and several indications from the modification index give some pointers for model improvement. Deleting some items from the model and re-structuring two factors based on an EFA eliminates Heywood cases and results in acceptable model fit (Chi-square=4117, DF=2646, CFI=.94, RMSEA=.038, SRMR=.052), however, the CFA of the adjusted model now returns the following error message: "covariance matrix of latent variables is not positive definite". There aren't any negative values in the covariance matrix though and also not in the correlation matrix.

I'm fairly new to this kind of statistical analysis and have no clue what causes this error and how to fix the issue. I'd appreciate any help. Please find attached the original and adjusted CFA measurement model as well as the covariance and correlation matrices of the adjusted model causing the NPD error.

Thanks a lot for your input!

More Carsten Wendelin Hennig's questions See All
Similar questions and discussions