In reliability theory and survival analysis, the problem of point estimation based on the censored sample has been discussed in many literatures. However, most of them are focused on MLE, BLUE etc; little work has bee...In reliability theory and survival analysis, the problem of point estimation based on the censored sample has been discussed in many literatures. However, most of them are focused on MLE, BLUE etc; little work has been done on the moment-method estimation in censoring case. To make the method of moment estimation systematic and unifiable, in this paper, the moment-method estimators(abbr. MEs) and modified momentmethod 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 in Whitten et al.(1988). And the superiority of the MEs is especially conspicuous, when the sample is heavily censored.展开更多
基金This research is partially supported by National Science Foundation of China (No. 69971016).
文摘In reliability theory and survival analysis, the problem of point estimation based on the censored sample has been discussed in many literatures. However, most of them are focused on MLE, BLUE etc; little work has been done on the moment-method estimation in censoring case. To make the method of moment estimation systematic and unifiable, in this paper, the moment-method estimators(abbr. MEs) and modified momentmethod 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 in Whitten et al.(1988). And the superiority of the MEs is especially conspicuous, when the sample is heavily censored.
