A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering, School of Engineering and Technology

Thesis (M.Eng.) - Asian Institute of Technology, 1984

In this study, the Generalized Gamma (GG) distribution was analyzed. The properties of the distribution were first considered. It was shown that the GG distribution has different forms depending on the values of its parameters, and many useful distributions can be treated as special cases of it. Specifically, the Powered Gamma distribution is obtained when the power parameter is positive, while the Kritskii-Menkel distribution corresponds to the case where the expected value of the GG variable is equal to unity. Seven important methods for parameter estimation were then evaluated by using many sets of actual data on annual floods and annual maximum rainfall amounts along with the following criteria : percent standard error of the T-year event, percent standard errors of the parameters and the resiliency of under- or over- designed values. Related computer programs were developed, incorporating new algorithms for computing useful functions such as the incomplete gamma, digamma and trigamy functions. From the data sets and criteria used, it was found that the method based on the mean of the GG variable X, and the mean and variance of Y + in X turns out to be the best in most cases. The next desirable method is the maximum likelihood method.