Deviation metric for peak week in forecasts of several past influenza seasons.

<p>Dynamic (blue) and static (red) forecasts were considered here assuming the specific infectiousness profile function and scalar of 0.50 for pre-existing immunity. See legend of Fig. 4 for further details about x-axis. Deviation in peak week was calculated as difference in peak week of simulated and observed epidemic with 95% confidence intervals. Positive values should be interpreted as overestimation, by the forecast model, of the observed peak week. Negative values, similarly, indicated underestimation of the observed peak week. Metric values closer to zero indicated better predictive ability of the forecasting methodology. Peak week for each season (green line) was plotted to allow comparisons under each method of forecasting with regards to the timeliness of the forecast.