Allstaters, A colleague has presented me with a problem. What is the expectation of the number of trials for success on all of n iid Bernoulli processes of probability p. Of course, the problem is easy if n is 1, because the process is simply modelled by the geometric distribution which has mean 1/p but I cannot see how to generalise beyond n=1. Does anyone know the answer? Allan -- Dr. Allan White, Statistical Advisory Service, University of Birmingham Tel. 0121-414 4750 or 44750 (internal), Email [log in to unmask]