search

epiforecasts / simplified-forecaster-evaluation

Evaluation of a simplified forecasting model designed to emulate the performance of the ECDC forecasting hub.
https://epiforecasts.io/simplified-forecaster-evaluation/paper.pdf
MIT License
4 stars 0 forks source link

Improve discussion of visualised forecasts #14

Closed seabbs closed 2 years ago

seabbs commented 2 years ago

At the moment we conclude that performance is comparable on a visual scale (at a very high level) in both the results and discussion. As presented on the log scale this may actually mask a large amount of variability so could be a misleading statement. Revist this. (From @sbfnk).

Interesting that Germany/Greece seem to have had a larger BA.2 peak than BA.1 and Slovakia/Poland had not much of a noticeable BA.2 peak.

I'd suggest to 1) reword the text below a bit to note that the behaviour is across Europe was quite heterogeneous (the statements as written are really only true for the UK out of the 6 shown) and 2) add tics at the beginning/end of the timeline to Fig. 1b (or at least one for January).

seabbs commented 2 years ago

Onee option here is to only visualise forecasts at the 1 and 4 week horizons and then use this spare real estate to visualise the same data but on the natural scale.