Table 1.

Accuracy statistics for the dengue models shown in figure 1 with (a) one, (b) five and (c) a maximum of 10 variables. Kappa, kappa index of agreement; %correct, overall correct predictions (%); %PPV, positive predictive value (=consumer's accuracy for presence sites) (%); %NPV, negative predictive value (=consumer's accuracy for pseudo-absence sites) (%); %Flse +ves, false-positive predictions (%); %Flse −ves, false-negative predictions (%); sens., sensitivity; spec., specificity; TSS, true skill statistic; AUC, area under curve (ROC); AICc, corrected Akaike information criterion; nvar, mean number of variables used in models (AICc-dependent); n, total number of models in each series. In brackets after each mean = 1 standard deviation.

no. vars	kappa	%correct	%PPV	%NPV	%Flse + ves	%Flse -ves	sens.	spec.	TSS	AUC	AICc	nvar	n
1	0.306 (0.049)	65.30 (2.44)	54.60 (7.78)	75.97 (4.98)	12.07 (2.82)	22.62 (4.39)	0.543 (0.089)	0.754 (0.056)	0.297 (0.105)	0.735 (0.018)	721.2 (17.4)	1 (0)	100
5	0.455 (0.042)	72.73 (2.10)	77.93 (4.28)	68.69 (3.43)	15.93 (1.84)	11.33 (2.13)	0.768 (0.043)	0.678 (0.037)	0.446 (0.057)	0.80 (0.017)	714.9 (36.2)	5 (0)	100
10	0.716 (0.029)	85.78 (1.44)	88.74 (2.18)	84.02 (2.50)	8.43 (1.38)	5.78 (1.15)	0.880 (0.023)	0.827 (0.028)	0.707 (0.036)	0.911 (0.014)	502.5 (51.0)	9.76 (0.64)	100