Monte Carlo Error and how to decrease it

Dec 7, 2012 at 10:06 PM
Edited Dec 7, 2012 at 10:09 PM

From probability theory, the error in a Monte Carlo simulation is given by the formula: calculation error = SQRT((D/N)), where D= a constant and  N = the number of trials

So, if you want to decrease the error by a factor of ten, to get another significant figure of accuracy (say from 17% to 17.3%), then you must increase the number of trials, or iterations, by 100.  So to go from 17% with 1000 iterations to 17.X% in accuracy, you must increase the iterations by a factor of 100, to 100000 iterations.