Degrees of Freedom in "Baseline model... PreviousNext
Mplus Discussion > Confirmatory Factor Analysis >
 Dmitriy Poznyak posted on Tuesday, November 23, 2010 - 1:40 pm

I am trying to estimate a fairly simple and identical CFA model on a set of different countries. This is being done separately for every country.

The observed variables are measured and coded exactly the same in all cases (that is for all countries), but for whatever reason my degrees of freedom in the "Chi-square Test of Model Fit for the Baseline model" output section differ for some countries (ranging from 3 to, most frequently 5, if this is relevant). My degrees of freedom for the final model are always the same though and = 1, as would be expected. The fit of the separate models is adequate and quite nice - really no surprises here.

I am curious about this discrepancy in the "Baseline model" output and where it comes from. Does it necessarily point at any problem with the models?

Many thanks for your feedback,
 Linda K. Muthen posted on Tuesday, November 23, 2010 - 4:59 pm
Which estimator are you using? Which version of the program are you using?
 Dmitriy Poznyak posted on Tuesday, November 23, 2010 - 10:58 pm
I didn't specify an estimator, so I would imagine it's a default one. I need t check for the version, but I am sure it's version 5. Will check later at work
 Dmitriy Poznyak posted on Wednesday, November 24, 2010 - 1:57 am
I just upgraded to version 6 and the problem seems to be solved. Now the DFs in the baseline model are equal to the number of free parameters = 6. All other model fit estimates remain the same.
Not sure what was the reason of the problem in the previous version though. Can it possibly result from the fact that the CFA model I am estimating includes 3 categorical (2- 3- and 4- point scale) and 1 continious (11-point scale) variable?

Thanks again for your help,
 Linda K. Muthen posted on Wednesday, November 24, 2010 - 6:25 am
This is because of the following change in Version 6:

New method for second-order chi-square adjustment for WLSMV, ULSMV, and MLMV resulting in the usual degrees of freedom

Prior to Version 6 for those estimators, chi-square and degrees of freedom were adjusted to obtain a correct p-value so only the p-value was interpretable.
 Dmitriy Poznyak posted on Wednesday, November 24, 2010 - 9:14 am
This makes perfect sense now. Thank you so much again for helping me out

Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message