Anne Chan posted on Friday, September 03, 2010 - 8:35 am
Hello! I have a two-level data set, with students variables nested within schools.
I need to run an analysis to check compare students' choices of occupational fields (unordered categorical outcome) in terms of students' individual characteristics (Level 1) and in terms of school characteristics (Level 2).
I used to use HLM in running the above analysese, but I am hoping to switch to MPLUS. What kind of analyses it should be if I do it in MPLUS (2-LEVEL regression for multinomial outcome?) and can you please kindly point me to the relevant example in the User's guide?
Yes, twolevel regression for a multinomial outcome.
I don't think the UG has a specific example for that, so what you do is piece together parts from different examples. Start with ex 3.6 which shows single-level multinomial regression. Then combine with ex 9.1 which is twolevel linear regression. The Within part is the same for multinomial. The Between part refers to the random intercepts - with C categories you have C-1 random intercepts, so with C=3 you say:
%Between% u#1-u#2 on w;
I think that's it. I think you can simplify Between to say u on w and get the same.
Note also how you can expand your modeling using for instance the idea of ex 9.3 where the nominal variable is the ultimate outcome in a path analysis.