Because I'm certain this will get tons of media attention - it's bad research done on masks - here's a very simple explanation of why it's nonsense
-
-
Show this thread
-
This is the main results graph. You see that before masks (red) the slope was increasing, during masks (green) the slope started decreasing, and then after it decrease and finally leveled offpic.twitter.com/Er73GWIrSF
Show this thread -
But actually, the date chosen for the start of the study period WASN'T the first day of masks It was 5 days AFTER the start of universal masking regulations, because, well, why not? They could've allowed 1 day, 10 days, it's entirely arbitrarypic.twitter.com/neu7iMogAd
Show this thread -
And this poses a HUGE problem I quickly recreated the 7-day avg graph in Excel. Here's what the 'intervention' period looks like with that simple linear regression line if you move the start date to the day when universal masking STARTEDpic.twitter.com/0Z0kNJjSat
Show this thread -
The effect doesn't entirely disappear, but it's far less noticeable and certainly not statistically significant Moving the start date even one or two days almost completely changes the result! Not great
Show this thread -
Even worse, almost the entire relationship is basically based on just two days This is a problem!pic.twitter.com/rS6YWcsbfs
Show this thread -
Remove those two days, and the graph becomes flatter still, with an almost imperceptible gradient downwards that is almost certainly meaninglesspic.twitter.com/Rpbj1k3QDx
Show this thread -
And remember, these days are based on when people REPORTED, not technically on when they got the disease! In other words, the entire relationship could just be down to reporting error
Show this thread -
This isn't the only issue with the study, but it does show you how it fails even some of the more basic sense-checks (removing outliers, shifting study dates), making it mostly useless as evidence for/against masks
Show this thread
End of conversation
New conversation -
-
-
Those graphs are atrocious. You can’t even compare the two even if you fixed the x axis, because the y axis is so off. Why not put them on a log scale so you could actually visualize the two simultaneously?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.