Table 2.

Predictive performance statistics of different models evaluated on the training data for the same time period (months 23−60) to reduce the potential bias

Model Name	RMSE	SRMSE	R-sq.(adj)	Deviance Explained	Δ AIC
A: Seasonal Naïve	10.22	0.62	0.16	0.16	0
B: Meteorology Optimal	8.83	0.54	0.28	0.32	-492.64
C: Optimal Lag Surveillance Model	7.32	0.45	0.49	0.49	-2420.39
D: Optimal Met and Lag Surveillance Model	6.30	0.39	0.62	0.64	-2725.62
E: Optimal Representation Model	6.12	0.37	0.64	0.66	-2718.90
F: Social-economic data Included	6.10	0.37	0.64	0.73	-2713.86

The performance is measured on different metrics. The best model should have the lowest errors (RMSE, SRMSE) and have the best fit (measured in R-sq.(adj).), high deviance and low Δ AIC