The estimation of the parameters for the generalized gamma distribution is treated in many articles. Moments and
Maximum Likelihood Methods do not provide a unique and simple solution. These methods were derived by some
authors but the results are very complex. Some others methods, not based on the moment or the Maximum Likelihood
methods, were proposed: Stacy proposed a graphic method and Cohen and Whitten a routine that loops on the parameter
c. We use this last idea to define our routine. We use also the power transformation in order to have a gamma distribution.
Then, we use the chi-square test between the sample to be fitted and the gamma distribution.
The new algorithm proposed is powerful, provides the expected values when it is applied on particular cases of
the generalized gamma distribution. The only constraint is in the values of the sample we want to fit. Indeed, the
generalized gamma distribution is defined for positive values. Consequently, the algorithm designed in this article
requires also positive values in the sample to be fitted. To remove this inconvenient, we can use the four-parameter
generalized gamma distribution