@kravietz This graph feels really misleading. You're comparing time series data to non-timeseries (average) data and you're also comparing an ongoing event. America is up to >30k fatalities, but heart disease and cancer are >600k/year. The US fatality numbers recently spiked, but that's also misleading because New York and some others backported previously non-COIVD19 deaths to covid19, fucking up the numbers and making in impossible to see if we're approaching the inflection point.
This: https://news.yahoo.com/york-city-revises-coronavirus-death-205328540.html
Where do those numbers fit in? If they get tacked onto the end of the time-series data, that's fucked because that's not how time series worked and it makes it seem like a bunch of people just died. John Hopkins github data, appears to have just tacked it on to the end.
Dude, how is it not misleading? You're comparing unrelated things. You cannot compare time series to yearly averages that way and it be meaningful at all.
🇳🇱
Its very comparable. the flu makes sense to need a week rolling average to be valid for several reasons. 1) it behvaes differently early on in a new pandemic 2) it has a seasonal component so it will fluctuate up and down significantly.
However with heart disease or cancer there isnt a whole lot of variation one week to the next. The weekly average calculated from yearly data will be close to the weekly rolling average in this case.
Keep in mind the data there is NOT the number of deaths per year, it is the yearly deaths divided by 52 to approximate a weekly rolling average.
@freemo @kravietz I did a full write up on why graphs like this are really misleading:
https://battlepenguin.com/tech/fighting-with-the-data/
The TL;DR is that 3 months of time series data != 12 months of some other non-time series data. But real the post for a more detailed explanation.