I find the desire to plot Covid cases/deaths/R0 vs Brexit vote fairly unimpressive.
1) *So* many epidemiologically significant factors correlated w/ Brexit vote that u need *loads* of other vars before results worth looking at. Rural/urban, pop dens, age, grads. All confounders.
1) *So* many epidemiologically significant factors correlated w/ Brexit vote that u need *loads* of other vars before results worth looking at. Rural/urban, pop dens, age, grads. All confounders.
2) Geography. Some people are using the large regions (London, North East, Midlands etc) for this kind of analysis.
No thanks.
With regions that large you’re basically opening the door to the ecological fallacy and saying "Come on in! Make yourself at home".
No thanks.
With regions that large you’re basically opening the door to the ecological fallacy and saying "Come on in! Make yourself at home".
3) At local authority level it’s not much better. As @MattSingh_ has shown, you’re still immediately and acutely exposed to the ecological fallacy
A good model can mitigate this (see e.g @benwansell’s work) but most of what I’m seeing ain’t that.
A good model can mitigate this (see e.g @benwansell’s work) but most of what I’m seeing ain’t that.
4/4) I also strongly suspect that in most cases where this is being done on a whim (rather than anything approaching a pre-registered analysis) we’re seeing a lot of:
"matches my beliefs — tweet it!"
vs
"hmmm not much of a pattern — guess I won’t share it"
"matches my beliefs — tweet it!"
vs
"hmmm not much of a pattern — guess I won’t share it"
