# Variance estimate of small “meta-set”?

I've got two sets $A$ and $B$, both with $n=2$ elements / measurements. The variance of both can be assumed to be similar / depend on the same measurement error-function. I need to compare (with a t-test) a third set $C$ (also $n=2$) of independent measures to the set $D$. $D$ should reflect the sum of the true value (without measurement errors) of $A$ and $B$ tries to measure, but also have a variation that could be expected of $D$ based on the variance of $A$ and $B$.

My question is simply: How to construct $D$?

-
There's a stats Q&A site - would this fit better there? It sure ain't set theory, so I'm retagging. –  Gerry Myerson Aug 10 '12 at 9:32
Did find any on stackexchange.com/sites#science –  deinonychusaur Aug 10 '12 at 9:40
Check out cross-validated. –  Gerry Myerson Aug 10 '12 at 9:48