Statistical inferences for the generalized gamma distribution

Procedures for handling statistical problems with nuisance parameters are considered with special reference to problems in the three parameter generalized gamma distribution. Maximum likelihood estimation of the parameters of this density has been investigated. Properties of these estimates are established which make it possible to make inferences about the parameters. Discrimination between various models for life testing problems is discussed and the robustness of the Weibull model is advanced. The question of the existence of the maximum likelihood estimates of the parameters for all samples is raised. Empiric evidence is presented indicating that they may not exist for all small samples --Abstract, page ii