Abstract :
Define the cumulative probability Pn of occurence of some event as the probability that it occurs at least once in n trials. If the single trial probabilities are not fixed, but are drawn from a distribution of their own, or are dependent on some other nonfixed variable, then it would be convenient to have an ¿effective single trial probability (esp)¿ for use in the simple Bernoulli model which would give the same cumulative probability of occurence as actually observed. We show here that the esp can be interpreted as a kind of average, and that its value is given by 1 minus the geometric mean of 1-d(p), where d(p) is the probability density function of the single trial probabilities, the p´s. We further define this geometric mean for both discrete and continuous distributions and evaluate the esp for several cases. These results are compared with earlier ones which suggest that the esp is given by the (arithmetic) average or expectation of the single trial probabilities, and we determine under what conditions this simpler result can be used.
