Table 2.

P-values from randomization tests of anthrax co-occurrences among species that commonly succumbed to disease (>50 total cases and case detection in most years)

	Buffalo	Impala	Wildebeest	Zebra	Livestock	Humans
Buffalo		0·021*	0·700	0·293	0·661	0·318
Impala	0·009**		0·559	0·198	0-072†	0·115
Wildebeest	NS	0·554		0·010	NS	NS
Zebra	0·341	0·165	0·012*		0·015*	0·120
Livestock	0·780	0-055†	0·008**	0·017*		NS
Humans	0·111	0-055†	0-061†	0·362	0·446

Values above the diagonal correspond to permutation tests where all positions in the time series were permuted. Values below the diagonal correspond to tests where only non-zero months were permuted (i.e. months where at least one anthrax case occurred). P-values have not been adjusted for multiple comparisons. If adjusted, the significant P-value would be 0·01.

^**indicates significance at the adjusted P-value of 0·01

^*indicates significance at the non-adjusted P-value of 0·05

^†indicates close to significance. NS indicates non-significance because the real test statistic was equal to the permuted test statistic (i.e. because of small numbers of occurrences). Shaded cells correspond to co-occurrences of anthrax that were significant (P < 0·05) using just presence / absence data.