TABLE 5

Neutrality tests under alternative models

	Tajima's D		D/D_min		Fu and Li's D		Fay and Wu's H
Models	P_avg	P_var	P_avg	P_var	P_avg	P_var	P_avg	P_var
Expansion (IS)^a	0.0000***	0.9891*	0.0001***	0.2163	0.0419	0.9977**	0.1415	0.9160
Expansion model^b	0.4297	0.9992**	0.6196	0.1170	0.5767	0.9952**	0.0002***	0.9978**
Refugia model^c	0.0427	0.1750	0.2151	0.0000***	0.5190	0.9920*	0.0000***	1.0000***
Bottleneck model^d	0.8680	0.7834	0.9819	0.0329	0.7072	0.9825*	0.0001***	1.0000***

↵* P ≤ 0.025 or P ≥ 0.9750,
↵** P ≤ 0.005 or P ≥ 0.995,
↵*** P ≤ 0.0005 or P ≥ 0.9995.
↵^a N₁/N₀ = 0.57, t₀ = 0, and t₁ = 0.6 in N generations.
↵^b N₁/N₀ = 0.21, t₀ = 0.07, and t₁ = 0.25 in N generations.
↵^c Two refugia, N_anc/N₀ = 0.1, N_refugia/N₀ = 0.001; time of merging refugia was set to 5.6 × 10⁻² and the duration of the refugia was 5 × 10⁻³ expressed in N generations.
↵^d Bottleneck, N_anc/N₀ = 0.15, N_bottleneck/N₀ = 0.003; time of bottleneck was set to 4 × 10⁻² and the duration of the bottleneck was 10⁻² expressed in N generations.
P is the probability of having a smaller value than that observed. P ≤ 0.0250 and P ≥ 0.9750 are considered significant. For Tajima's D and D/D_min, n = 195 loci and for Fu and Li's D and Fay and Wu's H, n = 43. Levels of variation were the same as used in Table 4.