Chi-Square statistics with multiple I... PreviousNext
Mplus Discussion > Missing Data Modeling >
Message/Author
 Dex  posted on Sunday, November 05, 2017 - 2:08 pm
Hi Drs. Muthen, I was wondering when conducting ordinal factor analysis(using WLSMV), if multiple imputation is used, how does Mpuls compute the Chi-Square test statistics?Is the aggregated Chi-Square simply the average across imputations?
 Tihomir Asparouhov posted on Tuesday, November 07, 2017 - 5:03 pm
That chi-square is not available when multiple imputations are used but you can see some of the complexities involved for the ML estimator when all variables are continuous here:
http://statmodel.com/download/MI7.pdf
You can also see there that the chi-square test is not simply the average.

If you have a particular hypothesis you need to test, Model Test is available with WLSMV and multiple imputations.
 Dex  posted on Tuesday, November 07, 2017 - 5:20 pm
Thank you for your reply. I understand that the Meng and Rubin (1992)'s procedure is only implemented for ML estimator.

When using multiple imputation + WLSMV to fit ordinal factor analysis; Mplus (v 7.4) also print chi-square and RMSEA. I guess my question is under this specific setting, how does Mplus actually compute the chi-square statistics? Is native average approach used here?
 Tihomir Asparouhov posted on Tuesday, November 07, 2017 - 5:57 pm
We just report the average. It is not meant to be treated as an actual test of fit. It should say the "Mean" of the chi-square tests in the output.
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: