Math help please?
I am doing weighted reservoir sampling (without replacement) from a list of probabilities. If I grab `n` samples, how do I calculate the probability of each sample being chosen?
Conversation
Obviously I can just do it by sampling. But this seems like something that I should be able to calculate directly...
1
Replying to
Probably easiest to calculate 1-probability a given sample is not chosen after n.
1
Different weights + without replacement makes that really tough. Maybe someone has something clever but I'll be a bit surprised if there's a nice closed form.
3
1
I ran into a similar problem in grad school ~2003, this seemed to offer some unusual tacks. Proved too challenging for me but luckily I found a simpler hack and moved on amazon.com/Urn-Models-The
Yeah using simulation is a perfectly reasonable hack, so I'm moving on too. :)
1



