Nausheen posted on Wednesday, April 06, 2011 - 3:56 pm
First of all, I apologize in advance for writing a question that involves another software program. I will try to focus my questions on Mplus as much as possible.
I'm having trouble understanding whether the multigroups (MG) analyses between AMOS and Mplus are completely reconcilable. I'm trying to replicate a multigroups analysis with okay fit in AMOS (RMSEA = 068; Chi Square = 42.89 df = 18) in Mplus, but the results are a substantially poorer fit in Mplus(RMSEA = .101; Chi Square = 55.761 df = 22) and I can't figure out why. The AMOS model I used is completely unconstrained. The models were the same in both programs before adding the grouping factor.
I've pored over the default settings for both programs, and would like to understand how to manipulate these in Mplus to match what I'm seeing in AMOS, as I would like to conduct more advanced analyses in Mplus.
(Q1) What could possibly account for the 4 extra dfs in Mplus? Here is my Mplus model:
Variable: Names are x1 x2 x3 x4 x6 x5R x7 gp; Grouping = gp (1=SA 2=CA); Missing is x7 (99);
Model: f1 BY x1-x4; f2 BY x5R x6; x1 WITH x2; x3 WITH x4; x3 WITH x7; x4 WITH x7; x7 WITH f1@0; f2 ON f1 x7;
Model SA: f1 BY x2-x4; f2 BY x6;
Nausheen posted on Wednesday, April 06, 2011 - 3:58 pm
More questions: (Q2) Does my syntax above successfully leave only measurement intercepts and means held equal across groups? Or,
(Q3) Are factor intercepts also held equal by default in Mplus?
(Q4) Is there a way to unconstrain measurement intercepts and means without the NOMEANSTRUCTURE setting of the MODEL option (which gives me an error - not sure why)?
(Q5) As far as you know, is multigroups analysis in AMOS and Mplus completely reconcilable?
Sorry for all the text, and thanks in advance for your input!