Concerns About Maximum Likelihood Estimation for the Three-Parameter Weibull Distribution: Case Study of Statistical Software.

The writers examined 12 statistical packages for four real-world datasets including oil spill data from the Gulf of Mexico. The different methods utilized by the packages resulted in fairly major differences in the estimated parameters between the statistical packages. Some statistical packages did...

Full description

Bibliographic Details
Published in:	American Statistician Vol. 65; no. 1; pp. 44 - 55
Main Authors:	Harper, William V., Eschenbach, Ted G., James, Thomas R.
Format:	Article
Published:	American Statistical Association February 2011
Subjects:	Application software Weibull distribution Maximum likelihood statistics
Online Access:	View this record in EBSCOhost

Description
Summary:	The writers examined 12 statistical packages for four real-world datasets including oil spill data from the Gulf of Mexico. The different methods utilized by the packages resulted in fairly major differences in the estimated parameters between the statistical packages. Some statistical packages did not offer three-parameter Weibull estimation. Other software tried the approximation in some cases. A subset of the 12 packages was subsequently tested in detail based on an experimental design using pseudo-random Weibull datasets. The writers also discussed the most prevalent estimation method used, which was maximizing a profile log-likelihood function.