Metodologia gráfica para dados de eventos recorrentes via bootstrap.

Abstract

Experiments related to recurrent events provide information about the number of events, time to their ocurrence and their costs. Nelson (1995) presents a methodology to obtain confidence intervals for the cost and the number of cumulative events. Apart from this, it is possible to construct confidence intervals via computer-intensive methods, where the bootstrap is a particular case. In this dissertation we present these two procedures and make a comparison, checking the coverage probability and the sample size influence in the precision of the intervals provided by the two methods. One of the advantages of the methodology presented in this dissertation is the possibility for its application in several areas and its easy computational implementation. An example from engineering illustrates the methodology.