A comparison of least squares and least absolute deviation regression models for estimating Weibull parameters

Abstract

This paper studies alternative methods for simple estimation of the parameters of a Weibull distribution. Results from an extensive series of simulation experiments indicate that L₁ estimation can provide a useful alternative to L₂ estimation over a range of sample sizes and parameter values. Different methods for approximating the cumulative Weibull distribution have also been investigated.

Keywords:

• least absolute deviation
• least squares
• robust
• Weibull