You are here

Familial variation

4 posts / 0 new
Last post
rebecca.barron's picture
Offline
Joined: 07/01/2016 - 11:21
Familial variation

Hi there,

I am working with twin data. My current analysis involves data that has a lot of variation, and I am using small numbers n=90 MZ twins and n=48 DZ twins. I am working off a univariate ACE script and I am getting the CE model as the best fit for variables where the MZ correlation is significantly higher than the DZ, I would expect some heritable component.

I have seen a paper (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3202796/) which has similar numbers and looking at similar data. They say they are not powered to present A and C individually, and instead present A and C combined as one familial component. They also incorporate D into their model.

I have two questions.

Firstly, should I be combining A and C into familial component?

Second question can D be incorporated into this model to make and ADCE model. I am working with twins reared together and have never seen this before.

Thanks,

Rebecca

AdminRobK's picture
Offline
Joined: 01/24/2014 - 12:15
variance components
I am working off a univariate ACE script and I am getting the CE model as the best fit for variables where the MZ correlation is significantly higher than the DZ, I would expect some heritable component.

That's a reasonable thing to expect from the correlations, but it's not guaranteed that the results of formal model-fitting will support it.

Firstly, should I be combining A and C into familial component?

The method described in the Nicholson et al. paper you linked parameterizes a model in terms of (1) familial variance common to DZ co-twins, (2) additional familial variance common to MZ co-twins, and (3) individual-specific variance. Shared-environmental variance goes into component (1), whereas additive-genetic variance is split in half between components (1) and (2). Dominance variance is split 25% to component (1) and 75% to component (2). I've not seen that done before, but it's a valid parameterization. But, is it actually informative to your research questions?

Second question can D be incorporated into this model to make and ADCE model. I am working with twins reared together and have never seen this before.

Yes, an ADCE model can be fitted to data from the classical twin design, albeit by less-conventional methods.

rebecca.barron's picture
Offline
Joined: 07/01/2016 - 11:21
Hi there, Thanks for the

Hi there,

Thanks for the response and the paper is helpful.

Is it possible to get a variance estimate for A and C combined using OpenMx?

We altered our univariate ACE script to combine A and C matrices to present it as a familial component. I've attached the script and an example of the output we're getting. Would you mind looking at the script to see if it is this valid analysis?

Thank you again for your help,

Rebecca

AdminRobK's picture
Offline
Joined: 01/24/2014 - 12:15
your script

At a glance, your script looks reasonable to me.