From our simulations, we also made some important novel observations. We found that the total variance in haplotypes for k causal variants was approximately k times the variance explained by 1 causal variant, implying that within a 100 kb block with mutations arising randomly and allowing for recombination at the usual rate of 1cM/Mb that the causal variants were in approximate linkage equilibrium. We found that the mean R 2 between haplotype values and the presence/absence of a genotyped SNP allele for the most associated genotyped SNP was ∼0.09 for 1 rare causal variant and ∼0.08 for 9 rare variants or more. In other words, for this particular model, the proportion of variance attributable to the causal variants explained by the most associated genotyped SNP was insensitive to the number of rare variants simulated, so that on average the most associated SNP explains less than 10% of the genetic variation contributed by the locus. Furthermore, we found that compared to this average, very common associated SNPs explained a smaller proportion of the total variance than less common variants, implying that
