Chunk #34 — Results — A single effective sample size does not capture all variance inflations in genotype frequency estimations, but it provides a good approximation
The genotype-specific sample size reductions in Eq.(12) can be applied in the following way: (1) the allele frequency p is estimated from the data; (2) three αGs are calculated by Eq.(12); (3) each genotype count is discounted by the genotype-specific αG, then these modified genotype counts can be used for further genotype-based association. Notice that αG’s in Eq.(12) are confined to the range (2/3, 4/5). One can also obtain an averaged ESS by averaging over three genotypes: αavg,sibpair(p) = p2α1 + 2pqα2 + q2α3, and αavg,sibpair (p is estimated from the data first) can be used to discount all three genotype counts by the same factor. In yet another approach, αavg,sibpair can be averaged over p: α¯≡∫01αa(p)dp. This leads to ᾱ = 0.7096. One can use ᾱ to discount all three genotype counts without the need to estimate p first. Note that this sample size reduction is less severe than that to account for variance inflation in allele frequency estimation, α = 0.6667.
