Open JindrichSoukup opened 4 years ago
Current dataset contain first date of outage, not the decision date.
New version of school outage takes dates from https://en.unesco.org/themes/education-emergencies/coronavirus-school-closures
atm i am not sure that confirmed cases is the best measure from when to start filtering dataset from the left, because in different countries the scaled of testing could be very different. Relationship between testing and cases you can see here: https://ourworldindata.org/covid-testing .I think that looking to deaths would be more accurate in longer term. Most warm countries in the world are more poor and therefore they might not have enough tests for everyone at least for bad or deadly cases would be more believable. To have enough deaths data i think it will take at least few more weeks, because many countries especially warm ones (poorer=less connections to get imported cases) are still at early stages of epidemic.
States reaction to coronavirus modulates the virus spread. We have to be careful to choose time period where the spread rate was not affected by these actions. On the other hand we need to focus on the period where community spread started to dominate over imported cases. I propose define the timeframe by confirmed cases>K from left and by date of full school outage +-L days from the right for each country individually.