A note on convergence problems in numerical techniques for accelerated life-testing analysis

Numerical solutions to determine maximum-likelihood estimators of parameters use a variety of techniques such as Newton-Raphson, Hook-Jeeves, and steepest ascent. All of these methods are subject to convergence difficulties that frequently occur in simulation studies. A tested and tried transformation for a variety of likelihood functions, and for the above mentioned numerical techniques, is introduced. The transformation is recommended for situations in which convergence difficulties are encountered.<
>