Huiping Xu posted on Friday, December 09, 2011 - 9:26 am
I am trying to examine measurement invariance between two groups. My group variable is 0 or 1. I tried fitting the MIMIC model using the multigroup analysis by imposing appropriate constraints. Here is my MIMIC model:
To replicate this using multigroup analysis, I have
model: f by y1-y5; y1 with y2 (1); y1-y5 (2-6); f (7);
model 1: [y3];
I got exactly the same parameter estimates and standard errors. But when comparing the model fit, these two models give very different chi-square (df), and fit indices are different too. The MIMIC model shows very good fit according to chi-square, RMSEA, and SRMR. But the multigroup model returns a very big SRMR that indicates a poor fit. Aren't these two models supposed to give the same results?