A Computational Technique For Maximum Likelihood Estimation With Weibull Models

Publisher Website
Google Scholar
Abstract
An improved computational technique has been developed for use in the maximum likelihood estimation of Weibull parameters for complete, censored or grouped data, for the 3-parameter Weibull model. To demonstrate the technique, a limited set of Monte Carlo results are given which compare the actual covariance matrix for the Weibull parameter estimates with an approximation to this matrix using the negative inverse of the empirical information matrix. The approximation, which has been suggested by several authors, is examined for both the 3-parameter and the standard 2-parameter Weibull in particular cases of complete, censored, and grouped samples of several sizes.