An Introduction to mrandom 3.0

We o er a set of public-domain tools to aid the development and test of pseudorandom Clanguage computer programs running under 4.3bsd Unix. Our tools make it easier to reproduce experimental results, to report them in a standardized format, and to check their validity by using a di erent random number generator (RNG). Furthermore, the tools should help users avoid some of the common pitfalls in randomized experimentation. These pitfalls include the initial seeding of the RNG, the generation of random integers in a restricted range, and the use of RNGs with known weaknesses. Our tool suite also includes several of the standard tests on the \randomness" of an RNG. All the tests we implement are based on examining a Pearson's chi-square statistic for a process that, under the null hypothesis, independently places N items into k equiprobable bins. We provide a routine to calculate an approximate p-value for such test statistics, applying novel corrections for \small N ." Research supported by the National Science Foundation, through its Design, Tools and Test Program under grant number MIP 9023238.