On the stability of the Bayenv method in assessing human SNP-environment associations

Table 3 Empirical p value tail comparisons between indicated runs

		.05	.01	.005	.001
(a) All populations (100,000 iterations) - Hancock2 vs. Blair1	.05	0.55	0.79	0.83	0.88
	.01	0.75	0.40	0.51	0.66
	.005	0.78	0.49	0.34	0.54
	.001	0.83	0.62	0.51	0.24
(b) All populations (500,000 iterations) - LongBlair1 vs. LongBlair3	.05	0.64	0.85	0.89	0.93
	.01	0.88	0.51	0.62	0.78
	.005	0.91	0.65	0.46	0.67
	.001	0.94	0.80	0.70	0.37
(c) Without Siberia (150,000 iterations) - W/O_Sib1 vs. W/O_Sib3	.05	0.60	0.83	0.86	0.92
	.01	0.81	0.45	0.56	0.73
	.005	0.85	0.56	0.40	0.61
	.001	0.90	0.71	0.59	0.30
(d) Without Siberia (150,000 iterations) vs. Hancock2 - Hancock2 vs. W/O_Sib1	.05	0.51	0.72	0.75	0.80
	.01	0.75	0.36	0.44	0.56
	.005	0.81	0.46	0.30	0.44
	.001	0.87	0.62	0.48	0.20

The table reports the fraction of SNPs from the smaller tail of one run that is present in the larger tail of the second run, averaged over climate variables. Two identical runs produce values of 1. Within each part (a to d), all pairs of runs (see Table 1) have similar fractions of overlap, and representative tables are shown here. For a given part, the first run's empirical tails are shown as rows (e.g., Hancock2 in part a), and the second run's are shown as columns (e.g., Blair1 in part a). Also, 623,318 SNPs are included in this analysis.

ISSN: 1479-7364