Previous studies have explored the impact of LD on GWAS. Most have done so empirically, and only for multiplicative models at single SNPs [e.g. Spencer et al., 2009]. At least two studies go further: Bhangale et al. [2008] considered recessive and dominant models empirically; Zheng et al. [2009] studied nonmultiplicative models analytically assuming the same allele frequency at the causal and marker SNP. Our study is more extensive: we characterize the effect of LD on power analytically, we do not impose restrictions on allele frequencies, and we study interactions as well as single-SNP models.
