Ehri Ryu posted on Monday, June 18, 2007 - 10:09 pm
Do you have any recommendation regarding the choice of the estimator for multilevel SEM (e.g., normal vs. nonnormal cases, balanced vs. unbalanced cases)? It seems that MUMLM is to provide robust standard errors for nonnormal case but it did not work as well as expected. Has there been any attempt to improve the robust estimator?
As far as I understand, MUML was useful before FIML was implemented in SEM softwares. Is there any situation when MUML is preferred to FIML unless the availability of software is limited?