Every new strain that establishes itself in the population will initially appear to have a higher transmissability. If one of those gets lucky (by appearing in a capital city, under lower restrictions, in the run up to Christmas) this effect could appear quite drastic... https://twitter.com/BBCNews/status/1341763451568480257

If you examined areas with more of this 'new mutant' & compared to those with less, it'd appear more transmissabile! But this approach is dangerous because you're not studying a random sample! This exposes you to a selection bias known as 'conditioning on the outcome'.

Two new 'super transmissible' strains of #SARSCoV2 are not implausible. But the emergence of this 'second strain' has greatly increased my concern that we're actually seeing an artefact of 'conditioning on the outcome' (a variety of the Texas Sharpshooter Fallacy).

Few people recognise the importance of understanding data lineage. Is this sample random? If not, how has it arisen? Epidemiologists are quite good at this. Virologists less so (they're more familiar with experimental data). We need epidemiologists looking at these data!

I'm not saying the UK isn't experiencing a serious spike in infections. Nor that we shouldn't take this seriously. But I AM worried that we're attributing these spikes to 'new strains', when this is very difficult to prove. If we've got this wrong, it will hurt trust in science!

Conditioning on the outcome is a particularly troublesome fallacy because it arises from a very natural instinct: to look at 'unusual' signals in data and try to explain them. But we have to be agnostic to data. Not draw the target around the cluster...

This fallacy is highly resistant to null hypothesis significance testings. 'Differences' will appear genuinely (& highly) significant. That means nice small p-values & narrow confidence intervals. That might give you confidence you've 'accounted for randomness'. You'd be wrong.

We're dealing with a huge, complex dynamic system. There are so many dice being rolled that 'unlikely' things can and WILL sometimes happen. Like the emergence of one or more 'new strains' that *appear* much more transmissible. When, really, a butterfly flapped its wings.

Which brings me back to this. Anyone who says they're certain about what's going on doesn't understand what's going on. They might be right. But they CANNOT be certain. Maybe ask them to explain the Texas Sharpshooter Fallacy. If they can't, be worried! https://twitter.com/PWGTennant/status/1341493036203819010