COMPARISON OF SOME SIMULATION ALGORITHMS ON BASIS OF DISTRIBUTION

The statistical distribution of the time series generated by three sim ulation algorithms of Gaussian stochastic processes are derived and co mpared. In all three methods, the time histories are modeled as weight ed linear combinations of terms of the form cos(omega(k)t + phi(k)) wh ere the phases phi(k) are independent random variables uniformly distr ibuted in [0,2 pi]. The frequencies omega(k), however, are either dete rministic parameters (the spectral representation algorithm), or indep endent random variables either uniformly distributed in a very small i nterval (the random frequencies algorithm). The results of the present investigation show that, from the standpoint of normality of the gene rated time series and irrespectively and irrespectively of computation al aspects, the random frequencies algorithm performs always (consider ing first-order distributions) and generally (considering second-order distributions) better than or as well as the spectral representation technique and its randomized version, which yield almost identical pro bability density functions. Finally, if the spectral representation al gorithm or its randomized version is used, it is recommended that the frequencies omega(k) be selected so that the energy associated with ea ch term cos(omega(k)t + phi(k)) be the same.