THE EFFECT OF THE PRESENCE OF CENSORSHIP: AN APPLICATION IN LIFETIME DATA OF ELECTRICAL LAMPS

Abstract

The Survival Analysis is the set of statistical techniques that evaluate the time until the occurrence of a certain event, even with the presence of elements that do not provide complete information about their respective times, called censored data. The objective of this work is to present the effects of the insertion of censored data in the reliability analysis of electric lamps using three non-parametric estimators, Kaplan-Meier, Nelson-Aalen and Life Table as comparative methods. The results showed that the difference between the survival function with the censored and uncensored data is more intense in the case of the presence of type I censorship. It was verified that the estimators tend to overestimate the survival with the censorship increase, independently of the Type of this, however, they maintain the same downward trend of survival over time. In the overestimation question, the estimator Nelson-Aalen showed ahead of the others.