Autocorrelation criterion for quality assessment of random number sequences
Еміль Віталійович Фауре, Ірина Валеріївна Миронець, Artem Lavdanskyi · 2020
The authors analyze different approaches to forming estimates of autocorrelation coefficients of random and pseudorandom number sequences.An integral estimate of normalized autocorrelation coefficients is theoretically obtained.Estimates of some statistical properties of normalized autocorrelation coefficients have been improved.The autocorrelation criterion for quality assessment of time series based on simultaneous analysis of several autocorrelation coefficients has been further developed by adapting it to uniformly distributed random variables.The technique of its implementation is presented.Applying the criterion revealed statistical deviations for some pseudorandom number generators that successfully pass all TestU01 autocorrelation tests.