Genotypes from dense sets of SNPs are partially, and locally, correlated (Sachidanandam et al., 2001). The pattern of correlation is nonrandom, since recombination does not occur uniformly across the genome but is localized to hotspots (McVean et al., 2004), giving rise to blocks of linkage disequilibrium. The extent of linkage disequilibrium (that is to say the degree of correlation between markers) is one determinant of the ability of a set of markers on a genotyping array to detect genetic signal. An important consequence is that genotyping only a subset of loci captures most of the common variation in the genome. Conversely, if a causative variant is not correlated with any markers on a genotyping array, it cannot be detected. The degree to which genotyping arrays capture genomic information is partly population specific, because population history affects the extent of linkage disequilibrium. Thus, linkage disequilibrium tends to increase the further away a population is from Africa (Conrad et al., 2006), consistent with the hypothesis that humans migrated out of Africa, experiencing severe population bottlenecks on the way to colonizing the rest
