Table 1 . Behavior of f- and D-statistics for a simulated scenarios of admixture.

Scenario	Fst(C, B)	Fst(O, B)	D(A, B; C, O)	D(A, X; C, O)	f3(B; A, C)	f3(X; A, C)	f4-ratio
Baseline	0.10	0.14	0.00	−0.08	0.002	−0.005	0.47
Vary sample size
n = 2 from each population	0.10	0.14	0.00	−0.08	0.002	−0.005	0.47
Vary SNP ascertainment
Use all sites (full sequencing data)	0.10	0.13	0.00	−0.11	0.001	−0.002	0.47
Polymorphic in a single B individual	0.10	0.16	−0.01	−0.06	0.003	−0.006	0.47
Polymorphic in a single C individual	0.10	0.16	0.00	−0.13	0.003	−0.007	0.46
Polymorphic in a single X individual	0.11	0.16	0.00	−0.11	0.003	−0.007	0.49
Polymorphic in two individuals: B and O	0.10	0.16	−0.01	−0.08	0.002	−0.005	0.46
Vary demography
NA = 2,000 (vs. 50,000) pop A bottleneck	0.10	0.14	0.00	−0.08	0.002	−0.005	0.48
NB = 2,000 (vs. 12,000) pop B bottleneck	0.14	0.17	0.00	−0.08	0.011	−0.004	0.48
NC = 1,000 (vs. 25,000) pop C bottleneck	0.16	0.14	0.00	−0.08	0.002	−0.005	0.46
NX = 500 (vs. 10,000) pop X bottleneck	0.10	0.14	0.00	−0.08	0.002	0.004	0.47
NABB′ = 3,000 (vs. 7,000) ABB′ bottleneck	0.14	0.17	0.00	−0.09	0.002	−0.007	0.47

We carried out simulations for populations related according to Figure 4 using ms (Hudson 2002) with the command: ./ms 110 1000000 -t 1 -I 5 22 22 22 22 22 -n 1 8.0 -n 2 2.5 -n 3 5.0 -n 4 1.2 -n 5 1.0 -es 0.001 5 0.47 -en 0.001001 6 1.0 -ej 0.0060 5 4 -ej 0.007 6 2 -en 0.007001 2 0.33 -ej 0.01 4 3 -en 0.01001 3 0.7 -ej 0.03 3 2 -en 0.030001 2 0.25 -ej 0.06 2 1 -en 0.060001 1 1.0. We chose parameters to produce pairwise F_ST similar to that for A = Adygei, B = French, X = Uygur, C = Han and O = Yoruba. The baseline simulations correspond to n = 20 samples from each population; SNPs ascertained as heterozygous in a single individual from the outgroup O; and a mixture proportion of α = 0.47. Times are in generations with the subscript indicating the populations derived from the split: t_admix = 40, t_BB′ = 240, t_ABB′ = 400, t_CC′ = 280, t_ABB′ = 400, t_ABB′CC′ =1,200, t_O = 2,400. The diploid population sizes are indicated by a subscript corresponding to the population to which they are ancestral in Figure 4 and are: N_A = 50,000, N_B = 12,000, N_B′ = 10,000, N_BB′ = 12,000, N_C′ = 25,000, N_X = N_C′= 10,000, N_CC′ = 3,300, N_O = 80,000, N_ABB′ = 7,000, N_ABB′CC′ = 2,500, N_ABB′CC′O = 10,000. All simulations involved 10⁶ replicates except for the run involving 2 samples (a single heterozygous individual) from each population, where we increased this to 10⁷ replicates to accommodate the noisier results.