Possible Duplicate:
A Question About Dice
I was reading " Expected time to roll all 1 through 6 on a die " and it got me thinking...
There are various ways to generalize the coupon-collector's problem (for example, by requiring certain tickets to be collected a certain number of times minimally... I think you can abuse exponential generating functions for that one).
The generalization I had in mind is this: The coupons have an arbitrary distribution: e.g. some are rarer than others.
Thus, given a discrete distribution $p_{coupon=x}(x)$, is there a solution that is more elegant than doing a massive calculation over the entire tree of outcomes with factorially-many branches? Also does the solution generalize nicely to continuous distributions, and if so, what does it mean to have the "coupon collection" problem on a continuous distribution?
A simple case might be, for example, how many rolls it takes to see all outcomes (2-12) of rolling a pair of dice. Of course there's a nice solution, but if I were to give an arbitrary distribution...