Estimates of mixture proportions and date of mixture.

Note: Estimates of the proportions and dates of mixture for all populations that give statistically significant evidence of mixture in Table 1 (4 Population Test Z<−3). Regions are abbreviated as: I – Northwest Europe, SE – Southern Europe and L – Levant. Mixture proportion estimates are based on f4 Ancestry Estimation using San, Yoruba, CEU and Papuan as the reference populations. The ROLLOFF estimated date of mixture uses CEU and YRI as the proposed ancestral populations (in the supplementary materials, we show that very similar inferences are obtained when the analysis is repeated with other ancestral populations, such as East Africans Luhya instead of Yoruba). Standard errors are computed using a Block Jackknife.
*Our simulations show that ROLLOFF produces a bias in the date estimates for small sample sizes, small mixture proportions, and old mixture dates. For each row of this table, we carried out a simulation to assess the expected bias for the inferred parameters (Table S12) and we computed the bias as (average - true date) in generations. Based on the simulation results, we have corrected the estimate in the last column as (estimated date - bias). We do not report a correction for the two rows marked “n/a” because our simulator cannot accommodate this large sample size.