I've been running SEM models 2 different ways. The first way I run the models, I'm using multiple groups and first running a fully constrained model, and then I run an iunconstrained model where I free up pathes to to compare the coefficients for 2 different age groups (younger and older adolescents).
I then run the same SEM models running each of my age groups separately, e.g., running one model for younger adolescents and running another model for older adolescents.
My problem is that when I run each of these different types of analyses, the findings are not consistent. When I free up the multiple groups, I find that one coefficient will be significant for one age group, but when I run the models separately for each age group, I find that the same coefficient will now be significant for the other age group. I'm very concerned about this issue, because shouldn't the models show consistent findings? Please help!
A multiple group analysis with all parameters free should yield the same results as two single group analyses of the same data. Check to be sure you have the same number of observations and parameters.