Procedures for generating gamma variates with non-integer parameter sets

Abstract

Abstract One of the most useful distributions in stochastic modeling is the two parameter gamma distribution. This paper presents a technique for generating random gamma variates from any two parameter gamma form with either integer or non-integer parameters. Only two techniques are presently available for generating gamma variates with a non-integer scale parameter, those being a variate averaging technique proposed by Naylor, et al. and a rejection scheme proposed by Johnk. The three generation schemes are compared with respect to (1) statistical goodness of fit, (2) computer running time and (3) random number calls. The author's scheme is shown to be statistically comparable with respect to goodness of fit, and generaliy superior relevant to computer running times and random number calls.