Table 4.

Kruskal-Wallis test for equality of frequency of the slow acetylation phenotype across geographical regions, linguistic families, subsistence modes, and climatic zones

Dataset1	Categories grouping2	Number of populations	Number of groups	Kruskal-Wallis H statistic	P-value3
AFR	Geo4	37	3	2.38	0.3039
	Lang	38	4	7.13	0.0680
	Subsist	38	4	15.62	0.0014 (0.0056)
	Clim	38	2	13.17	0.0003 (0.0012)
FP5	Geo4	28	3	2.52	0.2843
	Lang	29	3	3.61	0.1649
	Subsist	29	3	8.25	0.0161 (0.0644)
	Clim	29	2	7.74	0.0054 (0.0216)
FPLS	Geo	13	3	0.33	0.8480
	Lang5	12	2	1.15	0.2827
	Subsist	13	2	5.90	0.0152 (0.0608)
	Clim	13	2	6.43	0.0112 (0.0448)

¹ As described in Methods, the three population data subsets are AFR: 38 samples, excluding the 1KG Americans of African ancestry (ASW, see Table 1); FP: 29 samples of African food-producing populations; FPLS: 13 samples of African food-producing populations with sample size ≥ 20 individuals. Thus, the ASW sample was not considered in any of the Kruskal-Wallis tests

² Categories as in Table 1 : classification according to geographical region (Geo), subsistence mode (Subsist), linguistic affiliation (Lang), and ecoregion (Clim), namely climatic zone and biome, which defines the fourth categorization criterion that considers whether populations live within the dry savanna biome or outside of it

³ Significant P-values (i.e., <5 %) are shown in bold, and adjusted P-values after Bonferroni correction for multiple testing (here, four tests) are provided in brackets

⁴ Only three geographical regions are considered here (Western, Central and Eastern, Table 1) because the fourth region (Northern) is represented by one population sample only (EGY)

⁵ Only two linguistic families are considered here (Niger-Congo and Nilo-Saharan, Table 1) because the third family (Afro-Asiatic) is represented by one population sample only (SOM)