寿命分布的参数Bootstrap拟合优度检验方法 孙权,周星,冯静,潘正强 (国防科学技术大学 信息系统与管理学院,湖南 长沙 410073) 摘要:拟合优度检验在统计和可靠性等领域具有非常重要的地位,基于参数Bootstrap重采样的思想,对未知参数的常用寿命分布进行拟合优度检验。数值仿真结果表明,相对于传统的EDF检验,这种基于参数Bootstrap的拟合优度检验具有更高的功效,特别在小样本的情况下,优势明显。 关键字:参数Bootstrap;拟合优度检验;寿命分布 中图分类号: 文献标志码:A Goodness-of-fit test for life distributions based on parametric bootstrap Quan Sun, Xing Zhou, Jing Feng, Zhengqiang Pan (College of Information system and Management, National University of Defense Technology, Changsha 410073, China) Abstract: Goodness-of-fit test plays a vital important role in the statistics and reliability fields. Parametric bootstrap is applied in the goodness-of-fit test for common life distributions. Through simulation, the power of the test based on parametric bootstrap is compared to that of the common EDF tests, such as Liefors test, CvM test and AD test against several alternative distributions. Generally, the parametric bootstrap test has more power, especially in the situation where the sample size for the life data is relatively small. Key words: Parametric Bootstrap; Goodness-of-fit test; Life distribution 在进行产品可靠性分析中,我们常常需要根据实际的样本数据对产品的寿命分布做出假设,以便进行随后的研究分析。如果假设的寿命分布根本不能反映或者不能完全反映出实际寿命数据的特性,那么基于假设分布的后续研究结果显然是不可靠的。因此,在可靠性等领域,对产品的寿命分布进行拟合优度检验是一项尤为必要的工作。 在选择合适的分布类型时,研究者需要做寿命分布的拟合优度检验。传统建立在经验分布函数上的EDF(Empirical Distribution Function)拟合优度检验方法,如Kolmogorov-Smirnov检验[1][2][3]、C