zlacker

Sure,

> The prior probability of a lab escape is low

but the prior probability of "a pandemic from natural causes envelops the world" is also low, say below 1% a year. But now, given that "a pandemic does envelop the world", the posterior probability of lab escape and natural cause of course rises dramatically... and then saying "we didn't see much evidence for it" on a superficial tour organised by those responsible for the lab does not strike me as a very powerful rebuttal.

It's akin to "the prior probability of labour camps is low", and "on our tour of North Korea (accompanied by several North Korean tour guides showing us just those places they wanted to show us when they wanted to show us) we didn't see any labour camps, therefore we conclude that the probability of labour camps in North Korea is low".

While we're invoking statistics: Take the null H_0 = "the virus emerged naturally in any large Chinese city with > 1m inhabitants". In 2017, there were 102 of those. Thus, the probability under H_0 that the virus emerged in the one city in China with a level 4 biolab has p < 0.01.

>>FabHK+(OP)
Many low probability events happen every day, without contradicting the prior that they remain unlikely.

It is akin to saying "Pliny the Younger didn't die in Pompeii overtaken by a volcano because there is so many other cities he could have died in and other causes of death." Statistically, that would have been true, up until we learned that it happened. Then the low probability ceases to matter, since it becomes an observation.

Statistically, this is due to both priors rising, proportional to their original values, such that now the sum P(lab escape)+P(natural cause) == 1 and their average is a coin-toss (50-50). So without outside information, the prior statistical probability isn't low: it is exactly neutral between the options.

We've also seen viruses arise in small villages. This is even more unlikely to be predictable in advance which village with p << 0.0001.

>>FabHK+(OP)
Except there have been leaks for this specific lab and it has been known to have flawed training on handling of biohazardous material before.

>>oyashi+wg
And in this article it's explicitly covered they were doing work on these at biosafety level 2 and 3.

>>FabHK+(OP)
It is incorrect to consider only one year. If the probability of an event is 1% each year, you expect the event to happen every 100 years or so. In fact with probability 1 it will happen at some point.

>>manwe1+Ma
> Statistically, this is due to both priors rising, proportional to their original values, such that now the sum P(lab escape)+P(natural cause) == 1

Exactly, that's what I was getting at - the prior for lab escape is low, but given that the pandemic happened, it the posterior for lab escape is 1 - posterior for natural.

And we had about as many examples for lab escapes as for natural epidemics in the last few decades, it seems to me.

>>agnost+4A
Yes... and the last pandemic (Spanish Flu) was about 100 years ago. That's why I chose that number.