Hi, everyone

Some questions about the GxE model:

From Wikipedia,

The basic ACE model relies on several assumptions, including the absence of assortative mating,[4] that there is no genetic dominance or epistasis,[5] that all genetic effects are additive, and the absence of gene-environment interactions.[3] In order to address these limitations, several variants of the ACE model have been developed, including an ACE-β model, which emphasizes the identification of causal effects,[3] and the ACDE model, which accounts for the presence of dominant genetic effects.[6]

So why is the GxE model still valid?

Another question: Because rGE can result in spurious detection of G E, it is advisable to incorporate SES into the moderation model as a fixed regressor, which will partial out any phenotypic variance due to correlation between SES and A (Purcell, 2002; van der Sluis, Posthuma, & Dolan, 2012).

I just can’t understand how the rGE is out in GxE model?

The third questions：

Mr Rob, I'm so confused by the GxE model. Maybe a little silly.

You said the following in your paper (file:///C:/Users/officestaff/Desktop/nihms671114.pdf)

However, we can estimate more than two sources of familial variance in our sample. One possibility would be twin-specific environmental effects ( twin effects ), which would contribute to between-family variance among twins but not among non-twin siblings. Another possible source of variance is assortative mating. We need not assume that the additive-genetic correlation between full siblings is 0.5 we can estimate it from the data, because an ACE model would be identified by MZ twins and adoptees alone.

One possibility would be twin-specific environmental effects ( twin effects ), which would contribute to between-family variance among twins but not among non-twin siblings.

What do you mean for this sentence?

We need not assume that the additive-genetic correlation between full siblings is 0.5 we can estimate it from the data, because an ACE model would be identified by MZ twins and adoptees alone.

How to estimate the correlaton between full siblings? Can estimate using MZ and DZ data?

an ACE model would be identified by MZ twins and adoptees alone.

Do you mean the additive-genetic correlation are 1 and 0 for MZ twins and adoptees? Can they be included in the model together?

thaks!

For members of a twin pair, the total phenotypic variance would be Va + Vc + Ve + Vt

The MZ-twin covariance would be Va + Vc + Vt.

The DZ-twin covariance would be 0.5Va + Vc + Vt.

The covariance for non-twin full sibs would be 0.5Va + Vc.

The covariance for unrelated sibs reared together would be Vc.

An unorthodox thing I did in that paper was to define the total phenotypic variance for non-twins to be Va + Vc + Ve. I did that because the non-twins had lower phenotypic variance than the twins in that sample (see the SDs in the descriptive statistics table). Usually, Vt is included in the phenotypic variance for non-twins as well.

You would need to either have an additional kinship group (such as adoptive siblings) or have collected DNA from your participants.

Yes.

Edit: non-twins had lower variance, not higher

Thank you for your explanation.

My math and statistics are too poor to understand well.

Thank you for your explanation.

My math and statistics are too poor to understand well.

test post, please disregard.

When one of the limits of the ACE model you list includes not modelling interactions, it seems odd to then ask "why is the GxE model still valid?" The answer is it is used precisely to include moderation effects.

Purcell's paper is well worth reading closely. It goes into detail about what happens when assumptions are not met. Much study has also been devoted to what happens to parameters when, e.g. A and C are modelled, but D is also occurring.

It also explains why, when the twins share the moderator, the mandatory modelling of moderator effects on the means allows GxE effect estimation to avoid confounding by moderator effects on the trait of interest. Other extensions have been developed to allow for non-shared moderators (e.g. umxGxEbiv) moving the modelling out of the means into latent trait effects.

PS: Assortative mating is readily modelled by varying the correlation between genetic factors in twins from its default of .5. see the rA and rC parameters in umx's umxACE model.

Assuming "no genetic dominance" and "all genetic effects are additive" are the same point, and not the case. One can include dominance, but not dominance AND shared environment at the same time, without, as Rob notes, and additional degree of freedom from a new relative type or, now, a measured polygenic score, for instance.