Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Eric Reither <eric.reither@usu.edu> |
To | "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |
Subject | st: inflated SEs with FIML in sem |
Date | Mon, 31 Dec 2012 04:46:29 +0000 |
Hello everyone. I have a question about the estimation of SEs with complex survey data in Stata, when using the FIML option in the new sem command. But first, a small bit of background. I'm using FIML as an alternative to MI for missing data in a simple regression model (i.e., 1 dependent variable, no latent variables). Since I'm using complex survey data, I also use the svy: procedures to identify the weight/cluster/strata - and this works fine in sem when I leave out the FIML option. When I do use FIML, the point estimates appear correct - and the sample size increases appropriately - but the standard errors go completely bonkers, which is a highly technical way of saying that they become so inflated that the t-values are reduced to almost zero across the board. Now the question(s): Has anyone else had a similar experience with sem in Stata? Any thoughts about the cause of the problem, or about possible solutions? Thanks so much for considering my problem. Best wishes, and happy new year. - Eric Eric N. Reither, Ph.D. Director, Population Research Laboratory Associate Professor of Sociology Utah State University Office: 435-797-1217 Mobile: 435-757-9313 * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/