Skip to main content
. 2018 Jul 25;15(144):20180174. doi: 10.1098/rsif.2018.0174

Figure 3.

Figure 3.

Cumulative sum of log score of the three methods, by variant and target. In each sub-panel, the better scoring variant would have a higher cumulative score, i.e. closer to y = 0. For example, with DYN, the one-week-ahead scores for Baseline, Real-time and Baseline with post-processing have very similar scores. Removing nowcast degraded the scores and the availability of stable ILI improved the scores.