Figure 2.
Dates of Mixture
We pool samples based on linguistic affiliation (Indo-Europeans [n = 175] and Dravidians [n = 144]) and run rolloff (with the merged Illumina-Affymetrix data set of 86,213 SNPs) to measure the LD resulting from mixture between ANI and ASI. To obtain weights proportional to the allele frequency differences between ANI and ASI at each SNP (needed to run rolloff), we use SNP loadings obtained from a PCA of Basque and a pool of groups from the linguistic cluster whose admixture is not being dated (e.g., we run PCA with Indo-European and Basque when we are dating Dravidian admixture). The output of rolloff is represented as points and the line shows the exponential fit used for estimating the time in generations (n) since mixture. The nonzero constant c allows for variability in the mixture proportion among the groups we pooled and d is the genetic distance in Morgans. Standard errors are computed via a weighted block jackknife (see Material and Methods).