计数型最小样本量截尾值的序贯检验

对具有"高成本、破坏性"计数型产品的可靠性验收试验,序贯检验方案的样本量截尾值对试验的成本预算起着决定性作用.为降低试验的成本预算,本文对最小样本量截尾值序贯检验(minimum truncated sample size sequential test, MTST)进行研究,给出MTST的定义、性质及求解方法.通过与目前广泛采用的国际标准IEC1123及序贯网图检验进行比较,结果表明MTST显著地减少了样本量截尾值.为进一步减少序贯检验的样本量截尾值,当已知产品质量的先验信息时,本文研究了Bayes最小样本量截尾值序贯检验方案(Bayesian minimum truncated sample size sequential test, BMTST),与MTST的比较表明, BMTST极大地减少了序贯检验的样本量截尾值及平均试验次数,能更好地缩减"高成本、破坏性"产品的试验成本预算及平均试验费用.     

基于Koopman-Darmois分布族的序贯网图检验方法

通常高成本、破坏性的抽样检验需采用具有最大样本量限制的序贯检验方法．立足于设计出最大样本量尽量小的序贯检验方案,本文基于Koopman-Darmois分布族建立了序贯网图检验方法．与目前广泛采用的截尾序贯概率比检验方法相比,序贯网图检验方案具有更小的样本量上界,更适合高成本、破坏性的抽样检验．     

关于与ISO2859一次方案所对应的截尾序贯方案之优良性的实例分析

本文通过典型实例对BS-6001、NFX 06-024、ISO/DIS8422及与ISO2859正常和加严检查一次方案所对应的截尾序贯方案进行了分析,并讨论了这些截尾序贯方案在优良性方面的差异。     

多元Beta分布特性分析

利用小样本截尾序贯检验理论,在武器系统对空中目标的命中精度检验问题中,遇到了一类多元Beta概率分布函数,讨论分析了多维Beta概率分布函数的特性并给出了概率计算表.结果对武器精度检验具有重要意义和实用价值.     

计数型二次序贯网图检验

序贯概率比检验(SPRT)是应用非常广泛的抽样检验方法, 序贯网图检验在控制最大样本量方面很好地改进了SPRT, 但其结果还有进一步改进的余地, 为此, 我们建立了二次序贯网图检验, 计算结果表明, 它比原先的计数序贯网图检验有更好的效果.     

试验设备误差的统计假设检验法

本文给出了固定样本检验的子样 n 与方差比 λ~2的关系式λ~2=x_α~2;(n-1)/(x_(1-β)~2);n-1,介绍了固定样本检验法、序贯检验法,并把用固定样本检验法计算的子样容量 n 与序贯检验法计算的平均子样容量 M_(σ~2)(n)作了实际比较;为了确保在给定子样容量 M_(σ~2)(n)=n_0时有明确的统计假设检验的结论,介绍了序贯截尾法.还介绍了,有实用价值的验前误差可预估时的序贯检验法.     

一类渐近最优的截尾序贯检验

本文研究了一类截尾序贯检验的性质,发现在一定条件下,检验的平均样本量与检验显著性水平的对数的绝对值之比E_θτ_b/|logα(b)|有一个渐近下界。对多参数指数族,我们证明了重复显著性检验使其E_θτ_b/|logα(b)|的极限(当b→∞)达到上述下界,是渐近最优的。     

计数累积和控制图的原理

本文介绍基于近代序贯分析原理设计的一类计数型累积和控制技术.它是利用 前面样本统计量的累计信息判断生产过程是否异常,对信息利用充分,可以节省检验量,能更经济地保证产品质量.     

几何分布的一类贝叶斯停止判决法则

1 引言 以节约样本为目的序贯方法在数理统计中占有重要地位.从实际工作的角度出发,人们往往更强调时间的价值,希望当有足够的证据做出推断时应尽早停止试验,这样就提出了时间序贯计划.近年来,时间序贯方法得到了迅速发展(见[1—6]).[3]和[6]讨论了指数分布的时间序贯检验问题. [5]讨论了单试验平台情形,几何分布的时间序贯检验问题,适合于受试样品比较昂贵的情形.本文讨论多试验平台,受试品比较廉价而试验时间(次数)比较宝贵情形的几何分布的检验问题.     

“系统可靠性鉴定及验收方法”的原理及其应用

本文着重讨论从时间观点、经济观点来看,具有极大优越性并已得到广泛应用的序贯截尾寿命试验的计量抽样方案,即序贯概率比检验(SPRT),并证明了投试多台系统或设备进行独立的有替换及无替换试验时统一适用的判别准则。这样可针对给定的可靠性要求及风险值。根据总试验时间内失效的台数就可作出接收批、拒收批或尚需继续试验的结论。而且讨论了方案的作业持性曲线及期望试验时间。     

Controlled sequential factorial design for simulation factor screening

Screening experiments are performed to eliminate unimportant factors efficiently so that the remaining important factors can be studied more thoroughly in later experiments. This paper proposes controlled sequential factorial design (CSFD) for discrete-event simulation experiments. It combines a sequential hypothesis testing procedure with a traditional (fractional) factorial design to control the Type I error and power for each factor under heterogeneous variance conditions. We compare CSFD with other sequential screening methods with similar error control properties. CSFD requires few assumptions and demonstrates robust performance with different system conditions. The method is appropriate for systems with a moderate number of factors and large variances.     

关于序贯检验的Bayes分析

本文将充分利用验前信息的Ｂａｙｅｓ方法与序贯检验相结合,推导出了Ｂａｙｅｓ序贯检验方法。文章从表面形式上的决策方案推导到实质上真正的决策方案;并对飞行器落点密集度进行了模拟计算的具体分析,最后得出结论,Ｂａｙｅｓ序贯检验方法对具有昂贵成本的产品试验和我国航天试验有较大的应用价值．     

Sequential-analysis Based Randomized-regret-methods for Lot-sizing and Scheduling

Lot-sizing and scheduling comprises activities that have to be done repeatedly within MRP-systems. We consider the proportional multi-item, capacitated, dynamic lot-sizing and scheduling problem that is more general than the discrete lot-sizing and scheduling problem, as well as the continuous set-up lot-sizing problem. A greedy randomized algorithm with regret-based biased sampling is presented. We partition the parameter space of the stochastic algorithm and choose subspaces via sequential analysis based on hypothesis testing. The new methods provided in this paper, i.e. the randomized-regret-based backward algorithm, as well as the controlled search via sequential analysis, have three important properties: they are simple, effective and rather general. Computational results are also presented.     

���������ݻع��������

With the advance of computer storage capacity and online observation technique, more and more data are collected with curves and images. The most two important feature of curve and image data are high-dimension and high correlation between adjacent data. Functional data analysis has more advantage in deal with these data, which can not be treated by traditional multivariate statistics methods. Recently, a variety of functional data methods have been developed, including curve alignment, principal component analysis, regression, classification and clustering. In this paper, we mainly introduce the origins,development and recent process of functional data. Specifically, we firstly introduce the notion of functional data. Secondly, functional principal component analysis has been presented. Then, this paper is devoted to introduce estimation, variable selection and hypothesis testing of functional regression models. Lastly, the paper concludes with a brief discussion of future directions.     

联合监测过程位置参数和尺度参数的非参数Shewhart控制图

用于检测生产服务过程的传统控制图多数都假定过程的分布是已知的。这些控制困经常是在正态分布的假设下构建的,然而在服务质量实时监控中数据往往是非正态的。在这种情况下,基于正态分布假设的控制图的结果是不可靠的。为了解决这个问题,通常考虑非参数方法,因为在过程分布未知情况下,非参数控制图比参数图更加稳健有效。本文提出一个新的基于Van der Waerden和Klotz检验的Lepage型非参数Shewhart控制图(称为LPN图)用于同时检测未知连续过程分布的位置参数和尺度参数。文中给出了LPN图在不同参数下的控制限。依据运行长度分布的均值,方差和分位数,分析了LPN图在过程受控和失控时的性能,并与其他一些现有的非参数控制图进行比较。基于蒙特卡洛的模拟结果表明,LPN图对非正态分布具有很好的稳健性,并且在不同的过程分布下对检测位置参数和尺度参数,尤其对检测尺度参数的漂移都具有很好的性能。最后通过监控出租车服务质量说明LPN图在实际中的应用。     

Optimal testing strategies in overlapped design process

Testing is an important activity in product development. Past studies, which are developed to determine the optimal scheduling of tests, often focused on single-stage testing of sequential design process. This paper presents an analytical model for the scheduling of tests in overlapped design process, where a downstream stage starts before the completion of upstream testing. We derive optimal stopping rules for upstream and downstream stages’ testing, together with the optimal time elapsed between beginning the upstream tests and beginning the downstream development. We find that the cost function is first convex then concave increasing with respect to upstream testing duration. A one-dimensional search algorithm is then proposed for finding the unique optimum that minimizes the overall cost. Moreover, the impact of different model parameters, such as the problem-solving capacity and opportunity cost, on the optimal solution is discussed. Finally, we compare the testing strategies in overlapped process with those in sequential process, and get some additional results. The methodology is illustrated with a case study at a handset design company.     

抛散落点的均匀性检验

讨论了抛散落点的均匀性检验,给出了一种排序法检验,并将它与传统的两种检验方法进行比较．     

Sequential Variational Testing Hypotheses on the Wiener Process Under Delayed Observations

This paper proposes a sequential procedure for testing the null hypothesis absence of a drift in the Wiener process against the hypothesis that the drift is present. The decision is taken on the basis of current and randomly delayed observations. For any given error probabilities, the optimal decision rule minimizes a certain convex combination of the mean durations corresponding to these hypotheses. By a special geometric result, a convex characterization of the risk is obtained, and the initial problem is reduced to the optimal stopping problem.     

利用两步群测模型确认运动员违禁物质

运动员体内违禁物质的检测属于低发生率问题,介绍了利用两步群测模型解决这一问题的方法.通过分析运动员尿样(血样)检测结果,找出服用违禁物质的运动员,为运动员服用违禁物质的检测提供了一种新的分析方法.     

Type I Error Probability Spending for Post-Market Drug and Vaccine Safety Surveillance With Poisson Data

Statistical sequential hypothesis testing is meant to analyze cumulative data accruing in time. The methods can be divided in two types, group and continuous sequential approaches, and a question that arises is if one approach suppresses the other in some sense. For Poisson stochastic processes, we prove that continuous sequential analysis is uniformly better than group sequential under a comprehensive class of statistical performance measures. Hence, optimal solutions are in the class of continuous designs. This paper also offers a pioneer study that compares classical Type I error spending functions in terms of expected number of events to signal. This was done for a number of tuning parameters scenarios. The results indicate that a log-exp shape for the Type I error spending function is the best choice in most of the evaluated scenarios.