Results unexpectedly depend on time span about covid19_scenarios HOT 4 CLOSED

Klaus Peper has explained on twitter that ,"Changing time span automatically changes dates of interventions which has massive effect. Not a bug in a narrower sense but an effect that easily could be overseen." I can see why you might decide to do this but it's not the behaviour I expected. Three of us (all mathematical biologists) concluded this was a bug. Perhaps at least a note could be added to explain this feature.

from covid19_scenarios.

Thanks Rob. I agree its unexpected. We noticed this yesterday and added a note to alert users to this. We'll try to change this today.

from covid19_scenarios.

One option would be to integrate the tools for setting timespan and mitigation schedule by modifying the interactive plot with the moveable black dots. First, fix the time range of this plot to the maximum permitted values (e.g. 1st March 2020 to 1st March 2021). Second, add sliders for setting the start and end dates of the model output. A bonus is that then you no longer need a separate calendar webpage.

from covid19_scenarios.

this merge b553df9 should fix this. We now resample the linear interpolation of the previous mitigation curve. that still might result in small differences due to interpolation errors. But should be a lot closer to what you expect.

Thanks again for bringing it up. This was definitely unexpected behavior and needed fixing.

from covid19_scenarios.