Timeline for Inverse logistic regression vs. repeated-measures vs. latent class?
Current License: CC BY-SA 2.5
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 1, 2011 at 15:56 | comment | added | alexkchavez | Oh, and the "Path Analysis" section will be most relevant -- and also the SEM sections -- but none of the examples are the exact situation you're looking for. There's another Path Analysis section in Topic 2, which comes close to what you need. Lastly, Mplus has a bit of a learning curve, so be prepared if you go that route. | |
| Mar 1, 2011 at 15:51 | comment | added | alexkchavez | My (also) vague impression of SEM is that it's technically sound but often misapplied by people who by a large margin lack the background necessary to properly formulate/diagnose/test/explain the models they can create using today's software. I'm going to point you to a set of slides with good references and videos (you can skip the videos if you're pressed for time) -- Topic 1 + the ref lists should suffice -- by Muthen & Muthen, who are the authors of Mplus. The demo version of Mplus allows 6 DVs. statmodel.com/course_materials.shtml | |
| Feb 28, 2011 at 6:01 | comment | added | ian242 | I'm very interested in your SEM suggestion. I'll admit that my vague impression of SEM has been that it's like, say, game theory: very general, obstensibly can solve anything, but results in simplistic diagrams and not very satisfying answers. But your description, as simultaneously solving mixed models, makes sense. Can you point me to any papers/examples that seem particularly relevant to my situation? Thanks so much. | |
| Feb 28, 2011 at 5:57 | comment | added | ian242 | Thanks, Alex. Let me address a couple of your questions. In terms of problem's scale, I have ~17,000 events/transport decisions over about 4,000 people. There are around a dozen event level predictors, and a half-dozen at people level. | |
| Feb 24, 2011 at 16:39 | history | answered | alexkchavez | CC BY-SA 2.5 |