On Analyzing Non-Monotone Failure Data

Abstract

A new two-parameter distribution is defined for modeling non-monotone lifetime data. It is constructed based on the logistic-G family and the exponential distribution. Its hazard rate properties are different than those of the well-known distributions. Some of its statistical properties are presented. An extended regression based on the logarithm of the random variable of the introduced distribution is defined. The new regression can provide better fits than other special regressions for analyzing real data. The performance of the maximum likelihood estimates is investigated from a simulation study. Three lifetime data sets are used to prove empirically the usefulness of the new models.