OK, let me spell out the first problem. (Calling MegMeg for the others!)
[quote=Hsu]
where G and E are, again for simplicity, independent Gaussian random variables (normally distributed) with similar standard deviations (SDs).
This seems like an awfully big assumption to make. He gives no justification for why we should assume G & E are independent and normally distributed, with similar standard deviations. Much less how we would or could go about actually measuring these values in the first place.
In addition, I would also point out that the children would have their own E, thus their "X" is not fully dependent on their parents X anyways. So even if you took his calculations as accurate, if you could calculate the parents' X, at most this could only give you the child's G (and that's being generous) but would not be related to their E, thus their outcome could be significantly different than their parents.
I agree with Colinsmom, he should stick to physics.