I recently ran a bivariate cholesky script in OpenMX and am attempting to write up my results.

The results were a bit odd - I found that the genetic correlation was negative while the shared environmental correlation was positive. The phenotypic correlation is positive. If I am understanding correctly, this means that common genetic influences that serve to increase one trait will decrease the other trait, whereas the common shared environmental influences increase both traits.

My question is - I am tying to report the proportion of the phenotypic correlation due to genetic factors and the proportion due to environmental factors. I'm getting a value of -1 something for the proportion due to genetic factors and a value over 1 for the proportion due to shared environment. A colleague was looking at my results, and I found it very difficult to explain the idea of genetic influences contributing negatively to a proportion and shared environment contributing to a proportion over 1. Is there anything I'm missing?

I don't *think* there's a problem with my script but just wanted to check that this profile of results seems possible. From reviewing other posts, it seems like it is (I looked over this post - http://openmx.ssri.psu.edu/thread/3969), but I just want to make sure since I'm still learning about BG and also ask specifically about reporting the proportion of the phenotypic correlation due to genes/environment in a conceptually clear manner when the genetic and environmental correlations are opposite signs.

The interpretation as "contribution to a proportion" comes unraveled when some of the values involved are negative. See this thread (especially Professor Neale's post) and this thread.

Thanks! I looked over these threads - very helpful.

The main source that I relied on for developing my script was this powerpoint:

http://ibg.colorado.edu/cdrom2010/posthuma/MultivariateGeneticAnalysis/Multivar_Intro_Posthuma_2010.ppt

I see here in the slides (I've attached an image of the two slides I'm referring to within this presentation) that the author is referring to it as a "proportion of the phenotypic correlation due to genetic effects," which I understand from Professor Neale's post is not appropriate terminology (the word "proportion").

When I looked at the second thread you posted, the individual was wanting to calculate - "contribution of genes to phenotypic correlation between two phenotypes in the OpenMx script (rpha, not the genetic correlation rg)" -- same as me. She reported that her rpha was over 1, which was concerning to her. Even with Dr. Neale's interpretation of the "contribution" to the phenotypic correlation - "("a phenotypic correlation might be made up of .2 due to A, .2 due to C and -.3 due to E, giving a phenotypic correlation of .1") - I don't understand how A could contribute OVER 1.

You responded to this poster by questioning the formula she used. I am following the guidance of the slides and using the value that is circled (in my code, ACE.StandardizedA[2,1]). I just wanted to double check if I am indeed using the right value.

I guess in summary: in my mind, one conceptual issue is the idea of genes or environment contributing a negative value to the phenotypic correlation. I think I understand that we shouldn't be thinking of it as a proportion, and that if the genetic correlation is negative, the shared environmental correlation is positive, and the phenotypic correlation is positive, it actually makes sense to me to think about the genetic contribution being negative, as Dr. Neale said.

A separate conceptual issue is the idea of this contribution being greater than 1 or -1 - that I am still having trouble wrapping my mind around, hence the desire to make sure my script is correct!

Thanks again, this has already been really helpful.

A big big oops. I'm realizing that my question about the contribution being "greater than 1 or -1" doesn't make any sense and wasn't necessary, because I was indeed looking at a PROPORTION that was greater than 1. My value was the entity in this attached formula divided by the phenotypic correlation.

I agree that it makes more sense to report it as the contribution to the phenotypic correlation rather than the proportion. So I will not divide by the phenotypic correlation!

Thanks for the resources for this board to allow me to talk it all out!