|
Abstract:
|
Monte Carlo methods are used to estimate the mean of a random variable by the sample mean. Quasi-Monte Carlo methods solve essentially the same problem but by equi-distributed sampling, rather than IID sampling. For all such methods there remains the question of how many samples are needed to approximate the population mean to within a specified tolerance. This talk describes recent efforts to develop stopping rules that guarantee the accuracy of the answer produced and that determine the number of samples required adaptively, i.e., based on statistics of the initial samples. Various examples will be provided that demonstrate the success of our stopping rules.
|
