I am interested in determining if, for a certain type of trait, there is just as much variation between twins as there is between unrelated individuals. However, I am having difficulties in figuring out how I can combine equivalence testing (as covered here and here) with the kind of paired observations that comes from twin studies.
To be more specific, I have four sets of twins, and for each of them 400,000 measurements were made (representing 400,000 different genomic loci). My hypothesis is that the general level of variation at each locus between each twin is not significantly higher than the level of variation between unrelated individuals at this same locus.
While I am far from my usual comfort zone, I have come to the following partial conclusions:
- If I consider each locus separately, I cannot use any variance based methods, such as ANOVA, since I only have 2 measurements per group and it is impossible to obtain a variance without at least 3 measurements (and even then there are 0 degrees of freedom left)
- I cannot use the squared differences of measurements between twins as 4 sets of 400,000 observations, since then I lose the opportunity to evaluate the intergroup variation of the absolute measurements.
Which techniques/methods/models should I be looking into to make statistical sense of such data?
