Thread by @GidMK

You may have seen the massive viral tweets about how staying at home doesn't prevent COVID-19 deaths

These were based on a paper with what we think are quite significant flaws

Our full critique now preprinted here:

osf.io/63efj/

The basic explanation here is that the original article looked at whether Google "residential" mobility data was correlated with COVID-19 death rates, and found no association

There are significant drawbacks with that methodology, some of which I outlined in a thread

https://twitter.com/GidMK/status/1369078568844595200?s=19

However, together with @lonnibesancon @RaphaelWimmer and @FLAHAULT we have identified quite a few further problemsin

In particular, @RaphaelWimmer has found that repeating the author's methodology on simulated data where an effect of lockdowns is clearly visible results in no effect

Even using the dataset that the authors sent through, which does show an effect, slight perturbations in the data seem to remove it

What this means is that it is likely that the methodology in the study would not identify a benefit of lockdowns even unless they were unrealistically effective, and even then almost never

Forgot to mention - the original article has an Altmetric of 7,000 already. Just one of the tweets about it got 3,500 RTs. It's been all over the news already and has 150k accesses 😕