WLSMV with unbalanced clusters/catego... PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Jonathon Little posted on Tuesday, August 24, 2010 - 1:53 pm
In reading 'Asparouhov & Muthen (2007) - Computationally efficient estimation of multilevel high-dimensional latent variable models', I observe that WLSM performed well in the two-level case when compared to ML for balanced group sizes.

To what extent can these results be assumed to hold for WLSMV also?

AND

For a two level CFA with unbalanced group sizes (ranging from 1 to 100) and categorical data is there any basis on which to favour one type of estimator (either WLSMV, WLSM) over another?

And is there any reason to worry that either of these two estimators may prove unreliable in the unbalanced senario?
 Linda K. Muthen posted on Tuesday, August 24, 2010 - 2:25 pm
There is no reason to think that both estimators will not perform equally well.
 Jonathon Little posted on Tuesday, August 24, 2010 - 3:18 pm
As always thank you Linda.

It was really the performance of either WLSM or WLSMV in the senario of unbalanced group sizes that I was more concerned with. I have seen nothing in the literature about the use of either of these estimates in the unbalanced senario - are they appropriate for two-level models with clusters of different sizes?
 Linda K. Muthen posted on Tuesday, August 24, 2010 - 3:25 pm
This would need to be studied via Monte Carlo simulation studies. I see no reason weighted least squares estimators would be affected by this any more than any other estimator.
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: