Table 3. Out-of-sample (leave-one-out) root mean-squared error (ORMSE) for each Poisson generalized additive model.

Values are normalized by the population size of each ZIP Code quartile and then multiplied by 10⁶ to obtain ORMSE per one million residents. The rightmost column gives aggregate ORMSE across all ZIP Codes included in our study area. The quartiles contained: [0-8) (1st quartile), [8-12) (2nd quartile), [12-21) (3rd quartile), and >21 (4th quartile) percent of residents below the poverty line.

Surveillance Data Sources	1st quartile	2nd quartile	3rd quartile	4th quartile	Combined
ILINet	1.45	1.81	2.63	4.04	2.20
BioSense 2.0	1.69	2.05	2.66	4.23	2.52
GFT	1.33	1.64	2.61	3.63	2.00
ILINet+ BioSense 2.0	1.55	1.81	2.49	3.88	2.18
ILINet+ GFT	1.33	1.59	2.65	3.29	2.00
BioSense 2.0 + GFT	1.39	1.91	2.59	4.03	2.35
ILINet+ BioSense 2.0 + GFT	1.39	1.91	2.52	4.03	2.35