Non-parametric Analysis of Some Reliability Functions
by Panlop Zeephongsekul
Abstract: The term reliability is familiar to most of us as something which is good to have but how we achieve it can be elusive. We live in an age where complex mechanical and software systems play a pervasive role in many spheres of human activities. Due to design and other faults, few man-made systems are completely free of faults and there are many examples of reliability disasters in recent history. Therefore, it is of utmost importance that the designers of systems and software be made aware of the reliability of their systems by being able to measure them accurately. There are many reliability functions currently in use such as the hazard rate, reversed hazard rate, mean residual life, expected inactivity time and mean time to failure. All these measures can easily be calculated once the underlying lifetime distributions of the systems are known. Unfortunately, this knowledge is not always available and reliability functions derived from an inappropriate choice of the underlying lifetime distributions can be hazardous when they are applied.
In this talk, I will provide a non-parametric approach to estimate these reliability functions. The approach I will be discussing is a two-stage one where an empirical estimator is first obtained through binning and followed by an improved estimator obtained by smoothing the empirical estimator using local polynomial estimation.
For More Information: contact: Mihee Lee. email: firstname.lastname@example.org