How to get more fit index PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
Message/Author
 David Lin posted on Thursday, March 20, 2008 - 11:34 pm
I try to conduct a CFA, but it only output chi-square, TLI, CFI, RMSEA, and SRMR. How can I get more fit index like GFI, AGFI, CN, et al.?
 Linda K. Muthen posted on Friday, March 21, 2008 - 6:08 am
Mplus does not give these other fit indices.
 David Lin posted on Friday, March 21, 2008 - 3:16 pm
Does this imply GFI et al. are not good fit indices, and chi-square, TLI, CFI, RMSEA, and SRMR are enough?

Thanks in advance.

David.
 Linda K. Muthen posted on Friday, March 21, 2008 - 5:42 pm
No, this says nothing about other fit measures. There are families of fit measures. We have chosen one from different families rather than several from the same family. Those in the same family would behave the same giving no additional information. See the following paper:

Hu, L. & Bentler, P.M. (1998). Fit indices in covariance structure analysis: Sensitivity to underparameterized model misspecification. Psychological Methods, 3, 424-453.
 Stephan posted on Thursday, August 28, 2008 - 7:51 pm
Dear Linda,
After reading Hu & Bentler I know that Mplus covers most if the incremental fit indices. When it comes to absolute fit indices RMSEA/SRMR are provided but the GFI/AGFI are not. Large GFI may not mean an impressive result and it's problems with complex models are obvious. But these criteria are easy for readers because it is analogous to the R-squared in regression analysis. Could you please provide some hints to explain why Mplus doesn't show AGFI in the output, why RMSEA etc. is sufficient to cover absolute fit indices? Thanks for your help.- Stephan
 Linda K. Muthen posted on Friday, August 29, 2008 - 7:19 am
AGFI was not excluded for any particular reason. We feel the set of fit measures provided are sufficient.
 George Deitz posted on Wednesday, March 11, 2009 - 12:04 pm
Please note another request for AGFI.

While I agree that what is provided by MPlus should be sufficient, many reviewers (perhaps trained using other SEM software) favor fit indices that are not offered currently by MPlus.

While I'm sure there are "sexier" development projects on board than simply adding fit indices, my guess is that you would be making life just a little bit easier for many loyal users by adding a few more.

Just my two-cents.
 Linda K. Muthen posted on Wednesday, March 11, 2009 - 5:24 pm
Our philosophy is to provide one fit statistic from several different families of fit statistics rather than many fit statistics from one family. We chose fit statistics based on which performed well for a family. I think often people look at several fit statistics from the same family and conclude that model fit is good without realizing that the fit statistics are really one and the same. If you know of a family we have missed or a better representative for a family, let us know.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: