You are here

A question about using OpenMx for twin study

2 posts / 0 new
Last post
zhanglab's picture
Offline
Joined: 06/17/2014 - 10:29
A question about using OpenMx for twin study

Hi,
I am new to OpenMx and trying to use it for twin study (monozygotic vs dizygotic twins). I have some questions and hope someone may help me with the following questions.

I tried the following two scripts:
https://openmx.ssri.psu.edu/svn/tags/stable-1.2/demo/UnivariateTwinAnalysis.R
https://openmx.ssri.psu.edu/svn/tags/stable-1.2/demo/UnivariateTwinAnalysis_MatrixRaw.R

I knew their input data were different, but is there any difference for data processing steps between these two?

I see in order to decide between ACE and AE models, the aforementioned scripts do the log-likelihood ratio test at the end of the script. They got the difference between the log-likelihood of AE model and ACE model. But how could I decide which model I should choose based on this value? Should I do chi-square test using this value and degrees of freedom?

Also, is there any script available online which could also calculate the p-value for the parameters a2, c2 and e2?

Thank you very much!

mhunter's picture
Offline
Joined: 07/31/2009 - 15:26
These scripts appear to me to

These scripts appear to me to be doing the same model.

You could compare the AE model with the ACE model by using the mxCompare() function which does a likelihood ratio (change Chi-Squared / change df) test and shows AIC.

mxCompare(twinACEFit, twinAEFit)
     base comparison ep minus2LL   df      AIC   diffLL diffdf  p
1 twinACE       <NA>  4 115236.3 3996 107244.3       NA     NA NA
2 twinACE    twinACE  3 122406.2 3997 114412.2 7169.898      1  0

To evaluate the significance of parameters in any model, you could include confidence intervals wtih mxCI() objects in the model that refer to any named matrix, algebra or parameter in the model. You could also construct models where parameters were set to zero and use mxCompare() as above to do likelihood ratio tests to see if estimating that parameter is worthwhile.