The Inverse Power Law-Normal Model for Right-Censored Data With Application to Life Prediction of Organic Light-Emitting Diodes

Abstract

This work generalizes the inverse power law-normal (IPL-normal) model for complete data to right-censored data, assuming that the coefficient of variation remains constant and free of stress. The maximum likelihood (ML) estimating equations of the model’s accelerating parameters and the general coefficient of variation are derived using new trivial but fundamental identities. The ML estimating equation of the general coefficient of variation is explicit and generalizes its counterpart for complete data, which was previously introduced. The ML method is compared with the classical least squares (LS) technique. Although the ML method is laborious and numerically sensitive, this article favors ML over LS for a drastic reason that only ML can estimate the general coefficient of variation, but it still recommends using both the methods for some other reasons. The generalized IPL-normal model is used to precisely specify the life model of organic light-emitting diodes based on a standard real data of complete samples of lives which was discussed in several previous works but censored in this work.