[gregwebs]

Great to see statistical thinking. There have been a lot of Covid shouting matches because people are trying to reduce things to binary. Masks don’t work! Actually, they work as advertised- only 95% effective at best (and a cloth mask may be 30%), not a binary protected or not. That’s effective at blocking particles. You are still guaranteed to get some virus particles. They are going to be in your respiratory tract trying to replicate, and maybe some are replicating-does that mean you are infected in a clinically meaningful way? Depends.

Researchers will try to measure if some intervention like that works reporting binary PCR positive or negative- but we should want to know what was the severity of the illness. PCR positive with minor illness can be a good outcome indicating that the intervention helped lower the innoculum.

[blurbleblurble]

I don't know if I'd call this "statistical thinking", at least not in the conventional sense of linear statistics with simple parametric models. If anything I'd call this "systems thinking", which unfortunately is not the kind of thinking I typically associate with "statistical thinking".

[PaulKeeble]

Actually N95 is AT LEAST 95% effective. 95% is its worst case performance on a PM2.5 sized particle, at all other particle sizes it is better than 95% usually considerably. We also have N99 masks and even P100 respirators which do an even better job. We just need to start using them. We also need to educate ourselves how to fit and fit test a mask because if its leaking its not giving you that level of protection.

[hamburglar]

Well, the actual mask's performance is 95% effective, but you can't really make a guarantee about the overall worst case effectiveness when deployed by humans. N95 masks are not 95% effective on someone who has a big fluffy beard preventing a proper seal, for example. Or with people who use their gut feeling to decide where and when to put them on (like deciding if there's nobody in the elevator, you don't need it, ignoring the fact that someone may have gotten out of the elevator 20 seconds ago and left particulates in the air).

[avereveard]

I think that's about the material itself, then at the mask face interface you get all kinds of sealing issues

[weaksauce]

you're absolutely correct and furthermore smaller particles and larger particles are easier for the masks to catch to state it differently if it wasn't clear.

[ryanjshaw]

If you want to be truly scientific you also need to discuss the risks of mask wearing. Then, in relation to COVID response, you have to weigh up the costs vs. benefits of a mask wearing policy.

Some of those costs are hard to quantify scientifically or intangible e.g. restrictions on individual freedom, and even the scientifically quantifiable data may have large variability so you have to take an opionated position. Different cultures will obviously take different positions due to different weights they place on risks and values.

Therefore, at this point you leave the realm of science and enter the realm of politics. For whatever reason, people like to pretend it's entirely a scientific discussion but it's not.

[lesuorac]

> Therefore, at this point you leave the realm of science and enter the realm of politics. For whatever reason, people like to pretend it's entirely a scientific discussion but it's not.

I mean discussions are typically both science and politics. Masks lower the amount of people that get sick is scientific fact; why do you think they wear them in the operating room? Whether or not that is "worth" wearing a mask is politics though. However, if you want to convince people of something (aka politics) you may want to appeal to them using logic and for that you'll want to use facts.

Having large variability doesn't make something not a fact. Plenty of males 20-24 do not get in car accidents but that doesn't mean that none of them will or that if you were to insure say 100k males 20-24 and 100k females 20-24 that the males wouldn't in aggregate have higher claims. But given a specific male and specific female its entirely possible that the female gets into an accident first.

[avidiax]

> if you want to convince people of something (aka politics) you may want to appeal to them using logic and for that you'll want to use facts

I used to think this way, but this advice needs to be conditional.

Dispassionate or distant parties can sometimes use facts to drive a decision. It's not even clear that it's the "correct" decision, since nobody has all the facts, and presentation, ordering, accessibility all matter tremendously.

Once you add an emotional reaction, well, people will logically or rationally pick or ignore facts to justify the emotion. Surely you've had someone tell you that you logically shouldn't be angry, and you greatly appreciated that insight and recalculated your emotions, right?

https://news.harvard.edu/gazette/story/2022/01/logic-or-emot...

[lesuorac]

If you argue purely with logic you won't convince everybody but that doesn't mean you won't ever want to appeal to somebody with logic; just that not _every_ time you won't want to.

The different words in "Pathos, Logos, and Ethos" all have their place and if somebody is acting emotionally it'd be better to use a Pathos arguement.

[AvocadoPanic]

They're worn in theater to prevent the surgical team from contaminating the sterile field they're all huddled together around.