对数正态分布环境因子的统计推断

Statistical Inferences of Environmental Factors for the Lognormal Distribution

为了对不同环境下的试验数据进行折算与综合，本文在失效机理不变的前提下，系统地讨论了对数正态分布环境因子的定义；给出了判别两种环境下产品失效机理是否发生了显著变化的统计检验方法；并推导了对数正态分布环境因子的极大似然估计（ＭＬＥ）以及经典的Ｂａｙｅｓｉａｎ与Ｆｉｄｕｃｉａｌ区间估计。最后用数值例说明了这些方法。结果表明该定义及这些方法是恰当的。

In order to convert and integrate the test data in different environments, this paper. under a premise of constant failure mechanism, discusses the definition of the environmental factors for lognormal distribution systematically. proposes the approach of statistical test method to discriminate if a significant transformation has taken place in the failure mechanism of a product under two different environments. derives the MLEs, and the classical. Bayesian and Fiducial confidence intervals for envitonmental factors of the lognormal distribution. and at last illustrates these methods by numerical examples. The results indicate that the definition and these methods are feasible.