# Fit tests for ACE models

3 posts / 0 new
Offline
Joined: 09/09/2017 - 13:10
Fit tests for ACE models

Good morning,

I am running an ACE model for an ordered categorical outcome using script adapted from Maes. I seem to be able to generate most indicators of model fit, including AIC, BIC, -2LL and p-values associated with likelihood-ratio tests for nested models. However, I cannot seem to generate an associated chi-square value. Is there a simple way to generate chi-square value for these tests? Thanks,

Chance

Offline
Joined: 01/24/2014 - 12:15
"saturated" model...?

If you're using mxCompare(), then the output column named 'diffLL' contains the chi-square test statistics for each pairwise model comparison. The column is poorly named, as it's actually the difference in -2logL.

It sounds like what you might actually want is a chi-square test comparing one of those models with a "saturated" model. OpenMx has mxRefModels() to automate fitting a saturated model suitable for many applications, but maybe not for twin modeling or MxModels involving definition variables (which you'll see if you read the linked man page). Many of the generic reference scripts (such as those from Boulder) manually create a saturated model, so you might want to look for one that does and follow its example.

Offline
Joined: 09/09/2017 - 13:10
Thanks, Rob

Thank you very much for your note, Rob. Makes sense. I should be good to go.