Hello all. I hope someone can help with this, it's been driving me a little crazy this afternoon! (But maybe nobody else on this Forum is using AMOS - I hope someone is!!)

I'm using AMOS software to run Confirmatory Factor Analysis as the first step in Structural Equation Modeling. I've created my default model and a constrained model for comparison, and that's fine, but when I go on to do post-hoc model changes I'm hitting problems. I've named a new model and made the first change suggested by the Model Fit Indices, but when I run the analysis AMOS is changing the details on the default model to match the new model, so I can't compare the default model to the new one. I assume there must be a way of showing both sets of data but I can't work out how - has anyone got any ideas?

Thanks!

erm, I've only ever compared groups on the same model. Have you got the Byrne book? I have it on pdf (care of Walminski) if you'd like a copy.


Have to say that I've now decided AMOS is just rubbish, and use mplus instead - SO much better and far better online support/info for it on semnet and the mplus forums.

Thanks, I have got the Byrne book but it seems light on actual instructions to follow to get the software to do what I want. I'm coming to the conclusion that the whole of SEM is a bit rubbish, to be honest - the whole idea of post-hoc adjustments to a confirmatory model seems well dodgy to me! But maybe that's just because it's not working with my data :$

yeah, I think the same about SEM, its a bit like data fishing really!

Having said that, it is great for path analysis and when you want to use latent variables in path analysis.

Can you not just run one model, run the other and then do the chi squared difference test?

I suppose I could. Thanks, that's a solution if it comes to it.

And yes, data fishing describes it! I'm trying to be rigorous - and it's the path analysis with latent variables that I'm really interested in.