Studies on Methods of Random Number Generation for Computer Simulation

摘要

Results of studies to date indicate that the multiplicative or power residue method is generally satisfactory. The results employing the method on the GE 225, however, indicate that the sampling distribution of the means and variances have dispersions less than those expected from random samples. This is not a serious fault in many applications. However, when an important feature of a study employing random number generators involves empirical estimates of confidence intervals, the multiplicative generator leaves much to be desired.