It’s been a while since I posted anything in the Bad Statistics folder. That’s not as if the present Covid-19 outbreak hasn’t provided plenty of examples, it’s that I’ve had my mind on other things. I couldn’t resist, however, sharing this cracker that I found on Twitter:
COVID-19 is less contagious at higher temperatures — good news for lower transmission come summerhttps://t.co/hsAIJbDxDX pic.twitter.com/yFMYQsxFrs
— Bjorn Lomborg (@BjornLomborg) March 16, 2020
The paper concerned can be found here from which the key figure is this:
This plots the basic reproductive rate R against temperature for Coronavirus infections from 100 Chinese cities. The argument is that the trend means that higher temperatures correspond to weakened transmission of the virus (as happens with influenza). I don’t know if this paper has been peer-reviewed. I sincerely hope not!
I showed this plot to a colleague of mine the other day who remarked “well, at least all the points lie on a plane”. It looks to me that if if you removed just one point – the one with R>4.5 – then the trend would vanish completely.
The alleged correlation is deeply unimpressive on its own, quite apart from the assumption that any correlation present represents a causative effect due to temperature – there could be many confounding factors.
P.S. Among the many hilarious responses on Twitter was this:
Follow @telescoperFitting a 36 degree polynomial rather than a straight line, it seems indeed that contagiousness decreases somewhat with temperature, but only until 23.5 ºC, after which it explodes!
In other words, your data show that we must take immediate action to avoid global warming! pic.twitter.com/57gBNavREe
— Peter Laursen (@anisotropela) March 18, 2020
