In genetic case-control studies, the association signal originates from the allele or genotype frequency difference in the diseased and the normal group. The estimation of allele or genotype frequency is very much like the estimation of heads proportion in the tossing coin example given above. We will show that for allele frequency, effective sample size captures the effect of sample correlation exactly. Even for situations where the effective sample size does not provide an exact solution, for example, in estimating genotype frequencies, an averaged parameter usually leads to good approximation. Because the calculation of test statistics X2, p-value, and power all directly involve sample size, replacing the apparent sample size with the effective sample size is a quick and convenient solution to the problem of correlated samples without the need to use a custom program.
