I added data (33 observed variables, 8 latent variables, all are Likert scales) and ran CFA. Sample size is 618 respondents. A result was many fit indices (Chi-squre/DF, CFI, GFI, AGFI, NFI, NNFI, RMSEA, SRMR, and Hoelter's) are acceptable, except P-value. I modified my CFA model many times (about 11 times) by followed modification indices that suggested by a software. However, the P-value is still less than 0.05.

My question: is that acceptable if all fit indices are good, but the P-value is less than 0.05?

CFA model modification progress

  • $\begingroup$ Probably, yes. The reason other fit indices exist is because of problems with chi-square. But I'd be happier saying that if you actually said what the values were. $\endgroup$ – Jeremy Miles Nov 6 '16 at 3:08
  • $\begingroup$ Also, when you ask a question, check and double check the numbers. Presumably you mean 0.05, not 0.5. $\endgroup$ – Jeremy Miles Nov 6 '16 at 3:09
  • $\begingroup$ Thank you very much for the answer. The fit indices were Chi-squre/DF, CFI, GFI, AGFI, NFI, NNFI, RMSEA, SRMR, and Hoelter's. $\endgroup$ – Pakvalit Kurkoon Nov 6 '16 at 13:01
  • $\begingroup$ Ignore chi-square/df, GFI, AGFI and Hoelters, and then tell me the actual values of the others. $\endgroup$ – Jeremy Miles Nov 6 '16 at 19:42
  • $\begingroup$ I uploaded my fit values as a picture file above, but this one, I've tried to push p-value to pass 0.05. $\endgroup$ – Pakvalit Kurkoon Nov 7 '16 at 5:04

If your p-value is 0.05, that means that assuming the null hypothesis is true, then the probability of you getting the above results is 5%. So effectively, a lower p-value is better.

So I wouldn't worry about your p-value, as it looks good. Simultaneously your CFI, GFI and AGFI are increasing, while RMSEA is decreasing, which are all good signs.

You have to find that optimum region where the p-value is still low, while the CFI,GFI and AGFI are reasonably high, and it looks like you have reached it.

| cite | improve this answer | |
  • 1
    $\begingroup$ This is the opposite of true. Lower p-value indicates that the model is unlikely to have generated the data. $\endgroup$ – Jeremy Miles Apr 11 at 2:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.