Robust statistical inference for one-shot devices based on divergences

  1. Castilla Gonzalez, Elena Maria
Supervised by:
  1. Nirian Martín Apaolaza Director
  2. Leandro Pardo Llorente Director

Defence university: Universidad Complutense de Madrid

Fecha de defensa: 30 June 2021

Committee:
  1. Pedro Miranda Menéndez Chair
  2. Juana María Alonso Revenga Secretary
  3. Isabel Molina Peralta Committee member
  4. Ayanendranath Basu Committee member
  5. María Angeles Gil Alvarez Committee member
Department:
  1. Economía Financiera, Actuarial y Estadística

Type: Thesis

Abstract

A one-shot device is a unit that performs its function only once and, after use, the device either gets destroyed or must be rebuilt. For this kind of device, one can only know whether the failure time is either before or after a speci c inspection time, and consequently the lifetimes are either left- or right-censored, with the lifetime being less than the inspection time if the test outcome is a failure (resulting in left censoring) and the lifetime being more than the inspection time if the test outcome is a success (resulting in right censoring). An accelerated life test (ALT) plan is usually employed to evaluate the reliability of such products by increasing the levels of stress factors and then extrapolating the life characteristics from high stress conditions to normal operating conditions. This acceleration process will shorten the life span of devices and reduce the costs associated with the experiment. The study of one-shot device from ALT data has been developed considerably recently, mainly motivated by the work of Fan et al. [2009]...