Message/Author 

Abdel posted on Tuesday, April 08, 2008  10:34 am



I did a multigroup CFA on categorical data and I had 2 questions about the output: 1. If I get a chisquare with pvalue=0.000, CFI=.989, TLI=.992, RMSEA=.015 and WRMR=1.530, can I then conclude that it was a good fit? The pvalue and the rest of the statistics seem to contradict each other here.. 2. Where can I see the factor loadings that can be compared to the loadings that an EFA gives for the same data? I hope you can help me out, thanks in advance! 


1. You don't mention your sample size. If it is very large, chisquare can be sensitive as it is a test of exact fit. You can do a sensitivity analysis by freeing the parameters until you obtain an acceptable pvalue. If the original model results have not changed, you can assume that chisquare was sensitive to large sample size. 2. Compare to the factor loadings you obtain from the STDYX standardization. 

Abdel posted on Tuesday, April 08, 2008  5:58 pm



I had a sample size of about 7000, and compared four groups within that sample with the multigroup CFA of about 900, 1000, 2000 and 3000.. 


I would do a sensitivity analysis. 


Dear Dr. Muthén, I performed a CFA analysis with MLR (N=1180) on 26 items and 5 factors. RMSEA (0.048 p=0.896) and SRMR (0.057) indicate good fit, but CFI (0.896) is not acceptable. What should I conclude from it? Is there any other index that I should add? Thank you for your suggestions. 


How does TLI and Chisquare look? 


Dear Dr. Muthén, Chi square=955.9 df=261 TLI=0.881 This CFA model is the result of an E/CFA analysis. Thanks, Robert 


The model does not seem to fit the data well. If this is an EFA in a CFA framework, you should obtain the same fit as the EFA. So perhaps five factors is not optimal. 

Back to top 