Results from statistical analysis of popular pseudorandom number generators for simulation

Assuming that the number of replicates r and the number of random numbers n_i for each input variate sample are known to reach the required accuracy on simulation outputs, and that goodness-of-fit techniques test results are observed, we investigate possible interpretations of these results based on known bounds in probability and statistical analysis, and explore the relationship between input and output significance levels. Therefore, we focus on testing if each stream n_i and the aggregated stream rn_i are good candidates under the null hypothesis to be samples from the parent variate distribution with alternatives unspecified. Finally, we discuss a possible heuristic for selecting statistically acceptable aggregated random number streams for input variates in simulation