Weeks in and I still see nearly anyone using data poorly. In the sake of the general interest here's a recap of all the common #loserthink I see regularly correct by a cartoonist who never hated science. Here we go:
1/ log vs linear. As spread is slowing down everywhere this one isn't as bad as it was but still going. The first linear graph reflects of how the reality of deaths country wide but hides that it's actually slowing down, as shown by log graph scale. If you already knew, good4U! 
2/ adjusted start vs unadjusted start. I see a lot of "there's less/little happening here", incorrectly conflating the fact that virus spread is subjected to time&space and assuming every other places are, or could be like that place they mention. Illustrated there:
We did have a fair share of "what's the problem with Italy" early March that looks (predictabilly) silly in retrospect. Italy did more or less like (or better) France, Spain, UK, etc. The tweets from NYC medics are very close in substance to what we had from similar Italian pros.
An illustration with death rate per capita. Do pay attention to the provision at the top.
3/Indeed, don't compare figures country to country because "this is not a pipe"
A representation isn't reality. Those figures are measured made with a mess of criteria.
A representation isn't reality. Those figures are measured made with a mess of criteria.
One country will count a COVID death if patient had COVID + other diseases while another will not. One country will mass test while another will not. One country is big. One has a lot of old people. A zillion factors make any comparison a methodological trap. Don't walk in.
What can be done though is measuring how virus fares by the slope of a relevant metric provided its measurement are sticking to the same criteria all along. To that regard not all metrics are equal.
*contaminated cases* sucks because in most places, way to count drastically changed over time. There were no testing, then some, then many, tests changed, etc. Read this from NYC to see how figures are shaped by medical infrastructure rather than spread.
*death rate* based on those cases vs death is faulty for the same reason. But death rate can also be per capita or per ICU cases - those two are more useful to evaluate how gov and treatment do respectively. But with provision as...
...*death count* itself isn't a perfectly solid metric. As we've seen above the way to count change at least from places to places and sometimes in times. I believe some work can be made to account for it and the effort has started. But be wary of juicy headlines here.
4/ "my leader is terrible bc". If the data above have their issue, it's nothing compared to how our bias taint everything. We read our media who cherry pick for us info supporting what we want to believe. Reality doesn't care. But!
As imperfect as they are, the data above do allow us to make some form of comparison that are at least partially based on real world data. While we should be wary of conclusion, anyone who's flaming a leader in blatant opposition to data isn't part of the adult effort to get this
For instance if you portray Macron's leadership as abnormally bad, you have to account for the fact that AFAWCT, France is doing more or less like its neighbors.
(save for Germany... that doesn't do post morterm test though.)
(save for Germany... that doesn't do post morterm test though.)
Similarly if you want to impeach Trump for mismanaging the COVID situation, you need to show there's a case of USA doing bad.
Actually there's a nagging question if you're from EU about why it grew earlier in EU than in the US. Trump's early travel ban with China sticks as the obvious difference.
5) Yet again... If you were pro Trump and cheered at the above presentation of a facts supporting your view - see how easy it is to forget that
In that case there are solid indications that COVID was in CA and possible in a lot of other places way earlier than thought. That tells you how wrong all of our data likely are.
justthenews.com/politics-polic…
justthenews.com/politics-polic…
Early on I made a reference to , a concept developped by @ScottAdamsSays to describe the methodological errors we're *all* doing all the time. I'll conclude with one he mentionned as of late: if you're expressing certainty on a complex topic, you're experimenting #loserthing.
