Login

News

P value of 1 in mxCompare

Posted on
No user picture. Lisa M Joined: 08/19/2013
Forums
OpenMx Structural Equation Modeling
Hello,
I am working on an analysis of a univariate sex limitation model for depression and I have run into a potential problem. When comparing the homogeneity ACE model to the AE model, I get a p-value of 1 in mxCompare. The fit of the models are very similar:


> mxCompare(HomACEModelFit,HomAEModelFit)
base comparison ep minus2LL df AIC diffLL diffdf p
1 HomACE 4 16217.9 21258 -26298.1 NA NA NA
2 HomACE HomAE 3 16217.9 21259 -26300.1 -1.426603e-05 1 1

Does this mean there is something wrong with the models, or is open mx rounding off the value? When I compare them to the full heterogeneity model it works fine and I get p values of 0.20 and 0.12.
Due to confidentiality reasons I am not able to include the data but I have attached the script for the models.
Any input would be greatly appreciated,
Kind regards/
Lisa

Replied on Thu, 10/22/2015 - 15:17
Picture of user. mhunter Joined: 07/31/2009

The difference in log likelihoods is on the order of 1e-5. The p-value associated with this will be essentially one.


> pchisq(1.426603e-05, df=1, lower.tail=FALSE)
[1] 0.9969864

Log in or register to post comments
  Login

News