Machine calculation of moments of a probability distribution
A method is presented for the calculation on a machine of the moments of a probability distribution, necessitating little more than n additions and n references to memory for each moment, instead of the minimum of n multiplication, 2n additions, and 2n references to memory required by the most straightforward method (where n is the number of entries in the probability distribution). The method is directly applicable when a tabulated distribution exists, as when it has been computed by repeated convolution; but in this case it conserves both time and accuracy.