News for hackers. Stuff that matters.
Re:Alien hunters have spent 60 years finding new solutions f... by derbOac on Tuesday November 30, 2021 @10:45PM (#29400425)
It's sort of an interesting problem from a purely inferential perspective: for most problems, if you want to estimate the probability of X, you just go observe and see how often X appears.

There's issues about the estimate to place with zero observations from a sample (e.g., https://en.wikipedia.org/wiki/Rule_of_succession), but basically you'd have 0 or some very small number.

So why don't we just take it as 0, or ϵ?

The reason is because we doubt our observational ability. So then the question becomes "is the zero because nothing is out there, or because we can't observe it?"

I guess it's interesting to me because it's a salient case of where our confidence in inferential paradigms breaks down in a very obvious way. Either that, or we are uncomfortable with its implications. Or both.

Re:Alien hunters have spent 60 years finding new solutions f... by WoahNoun on Tuesday November 30, 2021 @11:04PM (#29400526)
This problem also exists in manufacturing. If you make 1,000,000 widgets and randomly sample 1,000 for quality control, how can you be sure if your processes are reliable if all of the samples pass inspection.

Luckily with most physical processes, you can assume it's >0 and go about some estimations using your linked article (a very basic one is rule of 3[0]). This particular problem where 0 vs >0 is seems impossible to make an actual inference about until some piece of data tells us it's >0.

https://en.wikipedia.org/wiki/Rule_of_three_(statistics)

Re:Alien hunters have spent 60 years finding new solutions f... by whatshisface on Tuesday November 30, 2021 @11:02PM (#29400519)
Zero? I would say we've observed at least one. :)