Comparison of Several Alternatives to Maximum Likelihood for the Weibull Distribution

摘要

Five methods of parameter estimation for the two parameter Weibull distribution are investigated using simulation. They are: Non-Linear Regression, Log-Linearization, Weighted Least Squares, Generalized Least Squares, and Method of Moments. Effects of contamination and censoring of data as well as robustness of the methods were investigated. Weighted Least Squares seemed to be the most cost-effective method and Nonlinear Regression was the most robust. (Author)