Understanding fit PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Hillary Schafer posted on Monday, February 11, 2008 - 3:50 pm
Hi,

Can you please explain (or provide a reference for) why there is no chi-square or associated fit statistics (tli, cfi, rmsea) for twolevel sems? Is it strictly related to the estimation technique?

Also, what sample size does your BIC estimate use in version 5 (i.e., level 2 or level 1)?

Thank you.
 Bengt O. Muthen posted on Monday, February 11, 2008 - 7:32 pm
The default estimator for 2-level SEM is ML. ML chi-square testing has not been developed for general two-level models, but is available in Mplus for random intercept models (such as V5 UG ex9.6). With random slope models the typical SEM ML chi-square testing for single-level models (testing wrt second-order moments, i.e. means, variances, and covariances) is no longer straightforward and such moments are no longer the sufficient statistics for the analysis.

In Mplus there is also a weighted least squares estimator for random intercept models which for both single-level and two-level data tests the model against second-order moments. This is available also for categorical outcomes.

The total sample size is used for BIC.
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: