> What I am saying is their adamant declarations of "masks don't work for the public", which note is an affirmative declaration, were false, and never had any supporting evidence.
There's no such thing as an affirmative declaration of the null hypothesis. You either have proof that something works, or you do not. If you do not have evidence that X works, or the evidence is ambiguous, your conclusion is the null hypothesis (that X doesn't work)...but that doesn't mean that you have to scream it from the rooftops. You can just say "we don't know; the evidence is poor."
The only accurate thing you can say in the "mask debate" is that strong declarations both ways are wrong. The original declarations were wrong, and the declarations now are also wrong. If you look at the data you can't judge either way (with the exception of cloth masks, which are looking quite poor), and so we must equivocate.
The only way you can possibly go on this issue and still be correct is to use nuance. And if you do that, then it's a question of how you use medical evidence to advocate for interventions.
I strongly disagree, epistemologically speaking. You can run repeated tests and then conclude an intervention is successful. You can also run repeated tests and conclude an intervention is not successful - as you put it, no better than the null hypothesis. Or, finally, you can have just not run tests at all. There is a difference between the second and third states, and health authorities implied the second state when the third state was far more accurate.
No??? This is what having a prior is for. In this case, the prior was the mechanistic model which told you with reasonable confidence that masks would work in the slices of worlds where the primary mode of transmission was one mitigated by mask-wearing. We do not in fact live in a state of helplessness absent a double-blind peer-reviewed RCT; you will actively come to incorrect conclusions if you refuse to use your existing models & knowledge of the world to draw conclusions about the likelihoods of various outcomes.
Yes. I'm incorporating a prior. Read what I wrote at the top of the thread: every medical intervention that has ever failed a trial has had a biologically plausible justification for doing the trial. Nearly all trials fail.
In the history of medicine, literally every failed medication, surgery, treatment or intervention has had an explanation that seemed plausible at the time. Just as we're seeing with masks, the vast majority of interventions have little to no effect. Many make things worse. From bloodletting to thalidomide to failed cancer drugs, medicine is littered with examples of people who "knew" that their preferred treatment would work based on "priors" or "plausible mechanisms", and ended up doing great harm.
As a bayesian and someone who is knowledgable of science and medicine, my prior probability of any medical intervention working is almost zero.
I don't disagree. You test to reject the null hypothesis. If you do not reject, you must accept. Eventually, hopefully, you give up on the failed alternative hypothesis and move on.
> Or, finally, you can have just not run tests at all. There is a difference between the second and third states, and health authorities implied the second state when the third state was far more accurate.
We agree completely here.
Yes, this is reasonable, in the general case, absent a specific example to examine. However, masks do in fact work to prevent the spread of certain kinds of disease, based on both obvious mechanistic reasoning and on actual experimental evidence to that effect. Making the affirmative claim that masks would not work against covid (at the time) would have been ignoring or denying the non-trivial possibility that they would work quite well (or work poorly, but working poorly is still working on the margin).