Hello,

is it possible to estimate a NTFD without zygosity information? I know from other twin models that they can be adapted to this type of data. Is it also the case for the NTFD? If so, how would one adapt the script in OpenMx?

Best,

Mirko

Wed, 04/19/2023 - 12:06

#1
NTFD without zygosity information

is it possible to estimate a NTFD without zygosity information? I know from other twin models that they can be adapted to this type of data. Is it also the case for the NTFD? If so, how would one adapt the script in OpenMx?

Best,

Mirko

What do you mean by NTFD? Nuclear Twin and Family Design? You don't estimate the design; you estimate parameters of a model using data from a design. Many parameters of a typical ACE model are identified by knowing the zygosity of the twins in a nuclear twin design (by which I mean MZ and DZ twins raised together).

I really don't understand what you are asking. Please elaborate.

It's ambiguous, could be several things. Assuming that it means Nuclear Twin Family Design, yes, I believe it could be estimated as a mixture distribution if zygosity information is missing. It would likely work best if some estimate of the proportion of MZ pairs in the sample was available.

Hi, sorry for the confusion, with NTFD I refer to the Nuclear Twin Family Design as discussed, e.g. by Keller here (presented in Fig. 1 and discussed before the Stealth and Cascade model) or at the Boulder workshop here. I do not have any experience in twin models without any zygosity information but I have heard that it is possible to estimate models, such as e.g. the univariate ACE model even without zygosity information by adjusting the genetic correlations based on assumptions made based on the twins' sex composition.

You may find the mixture distribution model for twin data helpful here, see https://vipbg.vcu.edu/vipbg/Articles/TwinRes-finite-2003b.pdf There are corresponding scripts in the OpenMx git directory - beginning Acemix, see https://github.com/OpenMx/OpenMx/tree/master/inst/models/passing

I think it should work fine with nuclear family, twins & their parents data. It is also possible to estimate the proportion of MZ pairs as (TotalNumberofPairs - 2 x NumberofOppSexDZpairs)/ TotalNumberofPairs

HTH

Yes, it should be possible to fit an NTFD model without zygosity information. Opposite-sex twins are obviously DZ. You would then need to search the literature for some estimate of what proportion of same-sex twins are monozygotic in the population from which your sample was drawn, and fix the relatedness coefficient for same-sex twin pairs to an appropriate value between 0.5 and 1.0.

If you have unrelated siblings reared together (e.g., stepsiblings or adoptees), you could actually estimate the relatedness coefficient for the same-sex twins, as a free parameter.