对数正态分布下基于截断数据的统计分析

加速寿命试验是在短时间内获得高可靠产品的失效数据的一种有效办法,因而得到普遍应用。但在低应力下,要在短时间内得到失效数据较为困难,常常进行到一段时间被迫停止,因而得到的是截断数据。本文考虑在有截断数据情形下,对数正态分布未知参数的估计。具体地讲,试验一共有k个加速应力S_1,S_2,…,S_k,有R个产品进行试验,在应力S_i下有n_i个产品。进行加速寿命试验,但是我们得到的不全是寿命失效数据y_j~i,i=1,2,…,k,j=1,2,…,n_i,n_1+n_2+…+n_i=n,而是Z_j~i=min(y_j~i,τ_j~i是预先给定的截断时间,记δ_j~i=I(y_j~i≤τ_j~i),则我们只能观察到(Z_j~i,δ_j~i),i=1,2,…,k,j=1,…,n_i。本文给出了用(Z_j~i,δ_j~i)来估计未知参数的一种方法,证明了这种参数估计的存在性,无偏性及相合性。

Parametric Estimation from Incomplete Observations under the Log-Normal Case

Buckley and James (1979) proposed an estimation method for linear regression methods with unspecified residual distribution and right-censored response variable. In this paper we propose a parametric estimation method trom incomplete observations in accelerated life test under Log-Normal distribution. This estimator is explicitly defined and easily computed. We have derived that the iteration algorithm is convergent and proved that the estimator a,b is undi-ased and consistent under weak conditions. Moreover,these estimators are maximum likelihood ones. A numerical example is also given to illustrate the proposed method.