Early warning signal for dengue outbreaks and identification of high risk areas for dengue fever in Colombia using climate and non-climate datasets

Table 2 Regression outputs of the CRF index on DIP

Department	Number	Selected model	P-value for Z-score test (overdispersion)	Constant (α)		CRF (β)		AIC^b	AIC (comparison)^c
Antioquia	108	Poisson	0.47	−2.89	***	0.31	***	3.84	4.26
Arauca	108	NB^a	0.00	2.00	***	0.06	***	8.46	30.37
Boyaca	108	NB	0.05	−0.59		0.51	***	4.26	4.19
Cauca	108	Poisson	0.19	−0.45	*	0.08	***	3.53	3.58
Cundinamarca	108	NB	0.03	−5.35	***	0.52	***	2.55	2.26
Guaviare	108	NB	0.02	3.23	***	−0.01		7.69	16.72
Huila	108	NB	0.00	0.49		0.05	***	7.87	14.33
Magdalena	108	NB	0.00	0.92		0.02		5.44	7.05
Norte de Santander	108	NB	0.00	1.61	***	0.05	**	7.68	10.54
Quindio	108	NB	0.02	−3.01	***	0.11	***	7.41	21.83
Risaralda	108	Poisson	0.11	−0.62	*	0.07	***	4.56	4.69
Santander	108	NB	0.00	1.07	*	0.06	**	7.12	9.89
Valle del Cauca	108	NB	0.00	−2.75	***	0.12	***	6.20	8.68

^aNegative Binomial
^bAkiake Information Criterion
^cAICs for non-selected count models were presented for comparison. The AIC fit test was consistent with the Z-score test in terms of choosing a better model fit except Boyaca and Cundinamarca. Since the AIC differences were trivial for the two departments, the Bayesian Information Criterion (BIC) was further assessed, and NB was preferred over Poisson
* Significance at the 10% level, ** at the 5% level, *** at the 1% level

ISSN: 1471-2334