Testing the quality of random number generators.

INTRODUCTION A good source of "random numbers" is important in genetics research and in many other applied sciences. There are many algorithms being used today for a wide variety of statistical procedures such as testing model effectiveness via simulation studies, estimating distribution functions and model parameters such as variances and covariances via Gibbs sampling and other techniques, understanding and studying a mathematical or statistical procedure that is intractable mathematically, and assessing the reliability of software. In this paper results are presented when 18 objective tests were applied to three familiar algorithms for the uniform istribution implemented in C. d