Table 3.

Calibration and forecasting performance metrics obtained from the fits of the SEIR model and the exponential growth model with a negative binomial error structure (i.e., <method1>=3, <dist1>=3) based on the first 17 days of the fall wave of the 1918 influenza pandemic in San Francisco. The SEIR model yielded a better forecast in terms of the MAE and MSE metrics. In contrast, the exponential growth model yielded a better forecast regarding the coverage of the 95% prediction interval and the WIS.

Calibration performance
Model	Calibration period	MAE	MSE	Coverage 95% PI	WIS
SEIR model	17	5.74	61.43	94.12	3.67
Exponential growth	17	5.99	67.47	100.0	3.81
Forecast performance
Model	Forecasting period	MAE	MSE	Coverage 95% PI	WIS
SEIR model	10	65.83	8430.51	70.0	49.77
Exponential growth	10	80.18	9384.90	90.0	46.93