Moment-method estimation based on censored sample (Q2574496)

In reliability theory and survival analysis, the problem of point estimation based on censored samples has been discussed in many papers. However, most of them are focused on MLE, BLUE, etc; and little work has been done on the moment-method estimation in the censoring case. To make the method of moment estimation systematic and unifiable, in this paper the moment-method estimators (abbr. MEs) and modified moment-method estimators (abbr. MMEs) of the parameters based on type I and type II censored samples are put forward involving mean residual lifetime. The strong consistency and other properties are proved. To be worth mentioning, in the exponential distribution, the proposed moment-method estimators are exactly MLEs. By a simulation study, in the view point of bias and mean square of error, we show that the MEs and MMEs are better than MLEs and the ``pseudo complete sample'' technique introduced by \textit{B. J. Whitten} et al. [Commun. Stat., Theory Methods 17, 2239--2258 (1988)]. And the superiority of the MEs is especially conspicuous, when the sample is heavily censored.