The easiest way to do this is I think to use Bayes Theorem. N is Poisson(lambda), and R|N is binomial(N,p). So, using Bayes Theorem: P(N-r = s | R = r) = {P(R = r| N-r = s) P(N-r = s)} / P(R=r) The conditional probability is binomial. The second probability is Poisson. The probability in the denominator is obtained as: Summation {P(R = r| N=n) P(N = n)} Michael Hosking
Oops! When the forums were updated the links to the old posts broke - the correct link is now https://www.acted.co.uk/forums/index.php?threads/sept-2007-question-3.1332/ and I have updated it above.