I'm risk-averse myself, when it comes to evidence. For example, I'd been tweeting to be wary of the HCQ harms arguments from that observational Lancet study (long before it was retracted) because.. Exactly. Not highest quality evidence, given they unblinded RCTs & found no harms.
-
-
Look at this new preprint. They find, quite unsurprisingly, that masks *increase* distancing. I've been arguing this for months based on social science evidence! This is what you'd expect, not false sense of security nonsense. But all this was disregarded. https://arxiv.org/pdf/2005.12446.pdf …pic.twitter.com/k7f0ZiQQ4s
2 replies 4 retweets 11 likes -
I don't think problems was lack of high-quality evidence—though sure, we can't have RCTs. Rather: the evidence that was there was dismissed; evidence free hand-waving was allowed for harms; and unresonable evidentiary standards were only held for masks, not for say, hygiene.
2 replies 0 retweets 5 likes -
An aside: why can't we have an RCT, with an encouragement design and IV specification? Eg provide free masks, cluster-randomised by workplace? This seems doable and important.
1 reply 0 retweets 0 likes -
I had this exact discussion early on with people on how to try but I couldn't see an ethical context or a workable framework during a pandemic. Rich countries? People will mask up, cost little issue. Poor countries? Given evidence "for" side, how do you withhold? Plus, not blind.
1 reply 0 retweets 1 like -
Here in Toronto lots of employees are not wearing masks, incl shops. Maybe pay people to wear them? No need to withhold from the control group in an encouragement design. Agree not blind, but that means you capture the full effect incl behavior change which is also interesting.
2 replies 0 retweets 0 likes -
The problem is masks for source-control is a community-wide intervention, controlling for egress, not ingress. So the RCT design wouldn't make sense with individual level measurement (and why a lot previous health-care research isn't applicable despite WHO being stuck on them).
2 replies 0 retweets 0 likes -
But if you managed to introduce enough random variation in treatment intensity at the workplace-level, you could look at workplace outbreaks as an outcome. Power likely a problem
1 reply 0 retweets 0 likes -
Maybe. But also note the k overdispersion that's become clearer. You're going to get a lot of confounding by super-spreader events. Plus evidence that masks increase distancing (will overstate mask effects). Very hard. But we can look at countries that masked up early probably.
1 reply 0 retweets 3 likes -
I can see a lot of causal inference being done on this in coming months/years
1 reply 0 retweets 3 likes
Absolutely!
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.