Environmental conditions and Puumala virus transmission in Belgium

Table 3 Results from regression models.

Statistical model	Linear regression		Logistic regression		Negative binomial regression
Dependent variable	Number of bank voles		PUUV prevalence		Number of human NE cases
	2004	2005	2004	2005	1994–2005
Land-surface attributes
Greenness of the vegetation	1.75**	-	-	-	-
Landscape configuration
Area of forest patch	-	-	-	-	0.02*
Proximity index	-6.39**	-	-	-	-
Proportion of build-up areas around the forest patch	-	-145.73*	-	-	-
Soil
Proportion of thin particles (< 10 μm)	-	-	-	-	0.20***
Climate
Maximum temperature during the previous winter	-	-	-0.22***	-0.19***	-
Rainfall during the previous autumn	-	-	-	-0.01*	-
Rainfall during the previous winter	-0.30*	-	-	-	-
Rainfall during the previous spring	-	-0.37*	-	-	-
Absolute number of bank voles	-	-	0.02*	-	-
Goodness of fit
N	17	17	17	17	17
Events/Trials	-	-	35.5/502.5	79.5/548	-
R²	0.71	0.39	-	-	-
Adj. R²	0.64	0.30	-	-	-
Degrees of freedom	-	-	14	14	14
Residual deviance	-	-	14.47	20.67	10.51
Deviance value/DF	-	-	1.03	1.48	0.75

Parameter estimates of significant variables and goodness of fit statistics for: (i) linear regressions on the absolute number of bank voles captured, (ii) logistic regression with logit link function and binomial distribution on the PUUV prevalence of bank voles, and (iii) negative binomial regression on NE cases per postal code area. For southern Belgium, data used in statistical analyses correspond to the average number of bank voles captured in summer and autumn. The dispersion parameter for the negative binomial family was taken to be 1. *** P-value < 0.001; ** P-value < 0.01; * P-value < 0.1.

ISSN: 1476-072X