A Comparison of the Lieberman-Ross and Man-Grubbs Methods
比较了两种计算串联系统可靠性置信下限的方法，发现Lieberman-Ross方法在利用全部失效数据时不会产生比Mann-Grubbs方法更低的界限。
This article compares two procedures for computing reliability confidence bounds for series systems whose components have independent, exponentially distributed failure times. The LiebermanRoss method is statistically exact, whereas the Mann-Grubbs method approximates an exact optimum bound. For test plans that involve simultaneous testing of all samples of each component type, it is shown that when the Lieberman-Ross method uses all of the observed failure data in computing the confidence bound, the resulting bound will not be practically lower than that produced by the Mann-Grubbs method. Simulation is used to infer that only when the observed data are ordered so that the observed failure times of a significant fraction of the least reliable samples are not actually used in computing the Lieberman-Ross bound can that method be expected to produce a bound superior to that produced by the Mann-Grubbs procedure, using the full data set.