Table 5  Three-population test in HGDP
Source1Source2Targetf3Z-scoreαLαUZSanZHanαL > αR
JapaneseItalianUygur−0.0259−74.790.4840.573−46.08−42.31
JapaneseItalianHazara−0.0230−74.050.460.615−45.19−42.22
YorubaSardinianMozabite−0.0211−56.950.2880.304−40.65−31.16
MozabiteSuruiMaya−0.0149−19.670.1650.408−11.51−9.40
YorubaSanBantu-SA−0.0107−31.390.6770.839−24.67−16.70
YorubaSardinianPalestinian−0.0107−36.700.070.157−25.64−18.35
YorubaSardinianBedouin−0.0104−33.730.070.185−23.37−14.24
DruzeYiBurusho−0.0090−27.620.5580.731−15.94−13.59
SardinianKaritianaRussian−0.0086−20.680.6940.923−10.07−10.98
DruzeKaritianaPathan−0.0084−22.250.5470.922−10.68−9.37
HanOrcadianTu−0.0076−20.640.8750.926−12.38−8.98
MbutiOrcadianMakrani−0.0076−19.560.0380.151−11.87−6.61
HanOrcadianMongola−0.0075−19.210.8790.916−12.63−8.16
HanFrenchXibo−0.0069−16.920.8880.922−9.52−8.19
DruzeDaiSindhi−0.0067−21.990.4670.877−12.25−8.40
SardinianKaritianaFrench−0.0060−18.360.8160.964−9.55−9.33
DaiItalianCambodian−0.0060−13.160.8460.928−6.78−6.43
SardinianKaritianaAdygei−0.0057−13.030.6350.956−5.60−5.59
BiakaSardinianBantu-Kenya−0.0054−13.420.4050.834−9.65−7.15
SardinianKaritianaTuscan−0.0052−11.260.8030.962−5.12−4.76
SardinianPimaItalian−0.0045−12.480.840.97−7.48−5.66
DruzeKaritianaBalochi−0.0044−11.580.4830.96−6.96−6.30
DaurDaiHan−0.0026−13.200.6640.26−7.89−6.31*
HanOrcadianHan-NChina−0.0025−7.090.9580.97−4.16−2.74
HanYakutDaur−0.0025−9.050.60.588−6.91−5.78*
DruzeKaritianaBrahui−0.0025−6.430.470.964−2.23−2.41
HezhenDaiTujia−0.0021−6.970.4520.39−4.36−3.94*
SardinianKaritianaOrcadian−0.0019−4.310.8030.952−2.18−3.24
SheYakutOroqen−0.0017−5.130.4220.296−4.99−2.44*
  • This table only lists the most significantly negative f3-statistics observed in HGDP samples. For each target population, we loop over all possible pairs of source populations, and report the pair that produces the most negative f3-statistic. Here we only print results for target populations for which the most negative f3-statistic is significant after correcting for multiple hypothesis testing; that is, the Z-score is more than 4 standard errors below zero. For the line with Bantu-SA as target, we used HGDP Han as an outgroup. In four cases indicated by an asterisk in the last column, the lower bound on the admixture proportion αL is greater than the upper bound αR, suggesting that our proposed three-population phylogeny is not feasible. We suspect that here the admixing (source) populations are themselves admixed. The 2 Z-score columns are with San and Han het ascertainment respectively.