tag:blogger.com,1999:blog-4594832939334410220.post2761906572202405042..comments2024-02-12T06:23:51.153-06:00Comments on Deeply Trivial: Statistics Sunday: Dealing with Missing DataUnknownnoreply@blogger.comBlogger2125tag:blogger.com,1999:blog-4594832939334410220.post-20385931137595957692017-08-29T11:12:50.500-05:002017-08-29T11:12:50.500-05:00Thanks, Jay, for sharing the work by Craig Enders!...Thanks, Jay, for sharing the work by Craig Enders! I'll definitely look into it. And look for a post (or perhaps a handful of posts) about estimation techniques in the future. My goal with this blog is to make these topics approachable to non-statisticians, so the struggle is in translating these topics into plain language. I hope to get to estimation techniques soon!Sarahttps://www.blogger.com/profile/13213593768515404983noreply@blogger.comtag:blogger.com,1999:blog-4594832939334410220.post-89115191466081905812017-08-29T08:15:53.937-05:002017-08-29T08:15:53.937-05:00Nice post. SEM and IRT models work under MAR for t...Nice post. SEM and IRT models work under MAR for the same reason: They're both from a general class of models (generalized SEMs) that allow for full information maximum likelihood estimation. Missingness in exogenous variables is, however, much more problematic in an SEM, and they do not cope with MNAR situations. I'd highly recommend taking a look at the recent work by Craig Enders (http://www.appliedmissingdata.com), who's synthesized a lot of the missing data literature and provided some new programs to do techniques like multiple imputation. Jay Verkuilenhttps://www.blogger.com/profile/07461798676830653869noreply@blogger.com