Table 3.

Non-spatial correlation (Pearson) and spatial correlation (Bivariate Moran’s I) between pairs of statistically significant covariates for districts in Ghana

Pairwise risk factors	#Non-spatial Pearson's correlation		Spatial Bivariate correlation
	r	p-value	Moran's I	p-value
Basic edu-Intermigration	0.001	0.992	0.032	0.064
Basic edu-Intramigration	-0.039	0.610	-0.003	0.373
Basic edu-Urban	0.969**	0.000	0.113*	0.023
Basic edu-Emp.-to-pop. ratio	0.925**	0.000	0.145*	0.018
Basic edu-No/other religion	0.861**	0.000	0.094*	0.025
Basic edu-Agric household	-0.014	0.854	0.013	0.198
Intermigration-Urban	-0.028	0.719	-0.008	0.404
Intermigration- Intramigration	0.612**	0.000	0.028	0.062
Intermigration-Emp.-to-pop. ratio	-0.011	0.888	-0.0005	0.240
Intermigration-No/other religion	-0.046	0.552	-0.046*	0.043
Intermigration-Agric household	0.867**	0.000	0.030*	0.047
Intramigration-Urban	-0.052	0.497	-0.0209	0.288
Intramigration-Emp.-to-pop. ratio	-0.040	0.607	-0.015	0.418
Intramigration-No/other religion	-0.084	0.275	-0.069*	0.007
Intramigration-Agric household	0.915**	0.000	0.115*	0.025
Urban-Emp.-to-pop. ratio	0.941**	0.000	0.145*	0.019
Urban-No/other religion	0.860**	0.000	0.098*	0.026
Urban-Agric household	-0.050	0.521	-0.034	0.059
Emp.-to-pop. ratio- No/other rel.	0.835**	0.000	0.104*	0.022
Emp.-to-pop.ratio-Agric household	-0.030	0.693	-0.015	0.388
No/other rel.-Agric household	-0.075	0.331	-0.128*	0.001

** Non-spatial correlation is significant at p< 0.01 (2-tailed)

*spatial correlation is significant at pseudo p < 0.05 for conditional 999 permutations

# Spatial correlation is most appropriate but the spatial correlation was much smaller than the non-spatial correlation and the departure from independence is consistent but weak. The assumption of spatial independence may have affected the non-spatial correlation result