Maximum Likelihood Estimation for Two Parameter Decreasing Failure Rate Distributions Using Censored Data

摘要

Problems of maximum likelihood estimation, for shape and scale parameters from certain decreasing hazard rate distributions which are typically mixed-exponential or work-hardened, are discussed. Sufficient conditions on the mixing distribution are given that guarantee regular behavior of the hazard rate; this ensures, even with highly censored data, that the MLE's exist whenever the sample satisfies a certain condition quite likely in samples from DHR distribution; otherwise a constant hazard rate is assumed. Some computational methods are discussed and applications made. (Author)