New study on MLE for censored lifetime data

This paper presents, a new study for censored lifetime data. Question arises frequently in engineering practice regarding the quality of reliability data analysis using the maximum likelihood estimates for the data with many suspensions such as field data. With a large number of suspensions, the predicted reliability resulted from the existing MLE is sometimes too optimistic. It is found out that, when suspensions are present, the MLE resulted from the existing sample likelihood function may not have all excellent statistical properties as claimed. A new sample likelihood function for random censored data is then proposed in this paper. The resulted ML estimators shall have all excellent statistical properties that MLE are supposed to have. A heuristic proof is given and numerical substantiation is presented.