OpenMx Scripts for Gene-Environment correlation (rGE)
Posted on
Qiuzhi Xie
Joined: 12/27/2018
Forums
Hi,
I currently need to examine the correlation between additive gene (A) and moderators (such as SES and home language exposure). However, I don't have the OpenMx scripts to conduct rGE (gene-moderator correlation). Can anyone here provide me with an example script of rGE or tell me where I can find such scripts? I would be grateful very much!
Could you say more about the
Could you say more about the kind of analysis you want to do?
Log in or register to post comments
In reply to Could you say more about the by AdminRobK
Many thanks!
Many thanks!
In a model containing A, C, E, and M (moderator, identified environment), I want to see the correlation between A and M (rGE), similar to the one in the picture attached. Could anyone here provide me with example OpenMx scripts to run this model?
Log in or register to post comments
Start with bivariate model
Hi,
If I understand you right, you want to estimate a bivariate AE-model where the unique paths of P are moderated by M.
You could begin with a bivariate model (see [here](https://hermine-maes.squarespace.com/#/two/)) and then add some objects with M defined as a definition variable, the moderated paths coefficients and the moderated variance components (see [here](https://vipbg.vcu.edu/media/course/HGEN619_2015/twinModAceCon.R): I think the objects `defAge1`to `covPI` are what you are looking for.). To get used to the syntax I recommend the materials of the [Quantitative Genetics course](https://vipbg.vcu.edu/academics/courses/course-materials/hgen619-2015/) by Hermine Maes and you might be interested in the materials of the Boulder Workshops. Another possibility is to estimate a "normal" bivariate moderation model and constrain the moderation of the cross-paths to zero. I think with umx the implementation is quite straightforward. You might want to check out a [wrapper function](https://github.com/mirkoruks/twinflex) for OpenMx where you can estimate the desired model directly. Note, however, that it's still work in progress.
All in all, I would begin with a bivariate moderation model unless you have strong apriori reasons not to moderate the cross-paths.
The rGE you can calculate as a secondary statistic. The formulas are strightforward and I'm sure, here in the forum or in the course materials you will find them.
I hope this helps!
Log in or register to post comments
In reply to Start with bivariate model by Benny
Many sincere thanks, and your
Many sincere thanks, and your information provided to me is indeed useful.
We want to have the Bivariate moderation model as the attached picture shows. I find the OpenMx scripts on the website (https://vipbg.vcu.edu/media/course/HGEN619_2015/twin2ModBivAceCon.R), but how can we see whether the genetic influence on moderator (am) is significant or not?
Log in or register to post comments
In reply to Many sincere thanks, and your by Qiuzhi Xie
hypothesis test
That depends: are you going to test that hypothesis in a model that includes moderation effects or not? My guess is "not". In that case, do the following. First, use `omxSetParameters()` to make a new MxModel object that has 'am' and 'ac' fixed to zero. Then, run the new MxModel. Finally, use `mxCompare()` to compare that model, and the model you ran that had 'am' and 'ac' freely estimated.
Log in or register to post comments
In reply to hypothesis test by AdminRobK
Many thanks for your reply!
Many thanks for your reply!
After reading posts in this forum, now I understand that the Bivariate moderation ACE model is similar to the Bivariate Cholesky model. We simply see the moderator as the first phenotype. Therefore, I can set confidence interval in the model to see whether the moderative effect is significant or not (just like the Cholesky model).
If I am not interested to see the moderative effect on E, I can simply build the Bivariate AC model with moderator.
Is my understanding correct?
Log in or register to post comments
In reply to Many thanks for your reply! by Qiuzhi Xie
sign indeterminacy
I am not completely sure that's correct, because the moderated variance component is quadratic in the moderation coefficient. So, the sign on the moderation coefficient might be indeterminate.
Log in or register to post comments