The above results apply for any given, fixed, marker loci. To study the impact of distortion on actual GWAS outcomes, we perform a simulation study with empirical patterns of LD. We find that loci highlighted by GWAS will often be highly correlated with the causal SNP, limiting the amount of distortion observed. When this is the case, there will be reasonable power to detect substantial departures from the multiplicative model, such as for recessive and dominant models. Therefore, there is value in testing for such departures routinely.
