Failure Prediction from Censored Data with Uncertain Population Size

Abstract

Listen

Translate article

To obtain prediction intervals for the future number of failures from a population with unknown size, the simulationbased calibration approach for a known population size can be adapted, if a prediction interval for the population size exists. The boundaries of the population size prediction interval are treated as known population sizes. Then, the lower bound of the final interval is the lower bound from the smaller population and the upper bound from the larger population is used as the upper bound of the final interval. This is possible as a larger population size leads to a smaller lower bound and a larger upper bound.