A Comparison between MLE and BLUE for a Decreasing Failure Rate Distribution when Using Censored-Data

摘要

Estimation procedures for the shape and scale parameters of a mixed exponential, decreasing failure rate distribution are compared and the more advantageous methods are determined for various situations. The maximum likelihood estimate, MLE, is slightly more accurate than Best Linear Unbiased Estimator of the scale parameter based on the first K order statistics. In the joint case the MLE's are much superior to other possible estimates of the distribution. (Author)