Hello,

I am planning to do bivariate ACE modelling using IQ and brain imaging derived measures and I would be very grateful if you could help me answer a few questions regarding the general assumptions for this type of modelling.

Are there any initial assumptions about the variables (or correlations between them) that should be met in order to apply for bivariate ACE? For example, is it meaningful to perform bivariate ACE when the correlation between variables is relatively low, or, in other words, should the initial correlation exceed a particular threshold for a bivariate ACE to be meaningful?

Thank you very much for your help.

Best,

Aurina

I would say not. It's possible for the phenotypic correlation to be rather modest, but for (as an example) the genetic correlation to be substantial but mostly counterweighted by sizeable environmental correlations with the opposite sign.

The probability that counterbalancing components of covariance (positive rA, negative rE, e.g.) exist increases as the phenotypic correlation rP approaches zero. That most of the time rA rC and rE have the same sign reflects the usual choice to investigate large rather than small rP pairs of variables.