On a possible error of type II in statistical evaluation of pseudo-random number generators (Q1909596)

The paper is a continuation of a sequence of articles by both authors. The quality of a random number generator \(G\) can be tested via tests of the statistical properties of the so-called renewal process (RP) simulated with the help of \(G\). RP's are often very sensitive to the deviation of \(G\) from uniformity. There are, however, distributions (called critical) for which the properties of RP are arbitrary close to the properties of RP corresponding to the uniform distribution case. Two such distributions are presented. A method (subinterval renewal process) is presented allowing to detect whether the distribution of \(G\) is critical. No application of the results on the widely used random number generators are presented.