Multivariate continuous ACE/ADE with sex as moderator: Example script?

5 replies [Last post]
trinewa's picture
Offline
Joined: 11/30/2009

Anyone who has a well-functioning OpenMx script for running multivariate ACE/ADE with sex as a moderator (general heterogeneity) (including constrained Cholesky, independent pathways and common pathways models) that they would be willing to share as an example script?

I find it hard to adapt existing univariate scripts and be sure I do everything right. It would be very reassuring to have a mulitvariate script to validate one's own against.

neale's picture
Offline
Joined: 07/31/2009
Modeling multivariate

Modeling multivariate sex-limitation is not nearly as simple as it sounds. Weird things happen with a sex-limited multivariate Cholesky model. For example, if one changes the order of the variables in the analysis, then the fit of the model changes. See, e.g., Neale, M. C., Røysamb, E., and Jacobson, K. (2006c). Multivariate genetic analysis of sex limitation and g x e interaction. Twin Research and Human Genetics 9(4):481–9. Issues also arise with the factor models; some insights can be gained by considering the method a variant of testing for measurement invariance between males and females.

I don't have a set of these models in OpenMx. Perhaps most useful would be to translate the relevant scripts from the Boulder/Amsterdam/VCU websites, and to verify that the OpenMx versions give the same answers as the Classic Mx one. Coordination of such efforts as part of an Open source project would be advantageous.

tbates's picture
Online
Joined: 07/31/2009
Hi Guys, Boulder and

Hi Guys, Boulder and Amsterdam have univariate sex-lim scripts, but nothing multivariate, AFAICT. Could be a Behavior Genetics or TRHG paper in the making?

neale's picture
Offline
Joined: 07/31/2009
Perhaps. I am not sure that

Perhaps. I am not sure that there is much more to say than was covered in the article in BG, cited earlier in this thread. Obviously, the here's-how-to-do-it-in-OpenMx piece would be novel. Maybe if there was a unified interface for sex-limitation, say entirely via definition variables, which would permit scalar and non-scalar varieties as well as regressing out sex as a covariate on the means, it would be a useful generalization.

Yet I think either one would probably best compiled into book form as you, Hermine and I have previously discussed... and which we should revisit.

trinewa's picture
Offline
Joined: 11/30/2009
I agree! It would be of

I agree! It would be of invaluable help to have a kind of quality assured script library including also the more advanced models. Trying to build these models 'at the kitchen table' for each and every new OpenMx user is time consuming, and one may well end up with a script that seems to work while in fact it is seriously wrong. Unfortunately, my knowledge of OpenMx at the moment is too limited to be of any use, but I hope it will be possible for someone with more competence to organize what you suggest.

Steve's picture
Offline
Joined: 07/30/2009
Excuse me if I am speaking

Excuse me if I am speaking out of turn. But I believe that what Mike was, ever so gently, suggesting is this.

1. You have interest in this particular model.

2. You have already demonstrated sufficient expertise in OpenMx to translate an example from Mx Classic, which has a known answer, into OpenMx (NB we would still be here to help you if you get stuck).

3. If you were to perform this exercise, you would likely learn a good deal.

4. In this way you could contribute back to the community by contributing a verified model to a library of OpenMx scripts.

5. You would also set a good example for other people who use this board for help. We all need to pitch in if this is going to be a successful community!