具有结构变化的非线性回归模型的阶段异方差检验

对于具有结构变化的非线性回归模型,两阶段的随机误差同时具有方差齐性是一个基本假设,但是该假设未必正确．本文研究该模型阶段异方差的检验问题．首先探讨了两阶段异方差的同时检验,然后构造了两阶段异方差的两个单个检验,分别得到了同时检验和单个检验的score统计量以及相应的调整形式．然后应用得到的检验统计量分析了南澳大利亚洋葱数据的阶段异方差性(Ratkowsky,1983),并用AIC,SBC进行模型比较,得到的结果与检验结果非常吻合．最后,用Monte Carlo模拟方法研究了统计量的检验功效．     

具有AR(1)误差的线性随机效应模型中方差齐性和自相关性的检验

在回归分析中，方差齐性是一个很基本的假设．本文对具有AR(1)误差的线性随机效应模型，研究了方差齐性和自相关性的检验问题．我们分别讨论了随机误差异方差、随机效应异方差、多元异方差以及自相关性的检验问题，并用score检验方法给出了三种方差齐性和自相关性的检验统计量．随机模拟的结果表明，当样本容量较大时，检验的功效较好．本文还给出一个数值例子说明检验方法的实用性．另外，模型的结果也可以推广到非线性情形．     

方差分析与参数估计

方差分析用Ｆ统计量进行检验，Ｆ＝Ｓ２ＡＳ２ｅ，Ｓ２Ａ为因素Ａ的组间方差，Ｓ２ｅ为剩余方差．实际上，Ｓ２Ａ、Ｓ２ｅ都是方差σ２的估计量．本文简介组内观测值个数相等情形单因素方差分析的参数估计方法．（双因素方差分析的估计方法类似）１．组间方差等于样本平均...     

正态总体位置参数移动的似然比检验统计量的分布

原假设（ｙ１，ｙ２．．．，ｙｎ）是正态独立随机量时间序列，其均值和方差分别为μ和σ＾２备选假设为均值μ在某一时刻（未知）发生变化，本文对σ＾２为已知的情况，导出了由Ｈａｗｋｉｎｓ（１９７７）提出的似然经检验统计量Ｕ的简明且便于计算的分布函数表达式，并建立了分布函数的数值表。     

具有线性趋势回归信度模型异方差的score检验

回归信度模型在保险研究中具有重要的作用.本文分险种内部,险种之间,险种内部及之间三种情形讨论了具有线性趋势回归信度模型异方差的score检验问题.首先推导了异方差存在性检验的score检验统计量,然后利用Monte-Carlo方法模拟了这几种检验统计量的功效,功效模拟结果显示：这几种检验统计量都有很好的检验效果.最后利用文中所得到的检验方法对旅客意外身体伤害保险数据进行了实例分析.     

两样本队列半数生存期检验:反射和检基统计量

依据反射或检基原理，本文提出用于截尾资料的两样本队列半数生存期（ＣＨＬ）检验．两样本合并ＣＨＬ经指数内插取自Ｋａｐｌａｎ－Ｍｅｉｅｒ或Ｂｅｒｋｓｏｎ－Ｇａｇｅ估计值．连续性校正，经以有效样本容量取代样本容量，扩展自Ｙａｔｅｓ校正．合并标准误来自同源性生存率方差估计值，后者经有效样本容量扩展自二项分布方差．无截尾时，这些统计量还原为经典中位数检验．与反射统计量相比，检基统计量具有更高的功效．附有工作实例描述其临床应用．     

单指标模型异方差检验（英文）

在可加回归模型中,高维回归分析一般采用单指标模型.该模型与参数模型相比更加灵活,同时避免了维数灾难,因为单指标将标准变量向量的维数降低为单变量指标.本文构建了一个带有函数型误差项的单指数回归模型用于检验单指标模型的异方差性.由于回归模型的有效推断要求在存在异方差的情况下考虑异方差,本文提出了检验单指标模型方差不变性的假设.将Levene检验和无限因子水平的方差分析理论结合得到检验统计量用来评估方差同质性.模拟研究显示与已有方法相比,所提检验统计量适用于多种情形.最后将本文的方法应用于分析一组实际数据.     

基于巴斯卡数据成败型元件串联系统可靠性经典精确置信限

设Ａ为由Ｋ个相互独立的成败型元件组成的串联系统，第ｉ个元件的可靠性ｐｉ，ｐｉ未知，ｉ＝１，２，…，Ｋ．设对第ｉ个元件，对于给定的ｍｉ，有ｎｉ个巴斯卡试验数据：Ｘｉ１，Ｘｉ２，…，Ｘｉｎｉ，其中Ｘｉｊ表示对第ｉ个元件进行试验，试验进行到ｍｉ次成功时所需要的试验次数ｊ＝１，２，…，ｎｉ，ｉ＝１，２，…，Ｋ．记Ｔｉ＝Ｘｉ１＋Ｘｉ２＋…＋Ｘｉｎｉ，ｉ＝１，２，…，ｋ．本文研究基于统计量（Ｔ１，Ｔ２，…，Ｔｋ）求串联系统Ａ的可靠性经典精确最优置信下限．     

多元正态的两个均匀性检验

本文提出两个均匀性统计量用于检验多元正态性．该检验建立在多元正态分布的一个特征性质基础上，模拟研究和实例分析显示该均匀性统计量可以帮助解释来自已有正态性检验统计量的结论．     

多元威布尔分布形状参数相等的似然比检验

将多元威布尔分布形状参数相等的检验转化为多元极值分布尺度参数相等的检验,利用Logistic模型的似然比统计量,给出相关参数为0．3,0．5,0．8时,检验统计量的模拟分位数和功效,指出相关参数越小,似然比统计量的功效越大。     

无重复因析试验中多个散度效应检验的一种改进方法

在无重复因析试验的多个散度效应分析中,常常出现错误识别的现象,即两个显著的散度效应可能在它们的交互列上产生一个错误的散度效应,并且现有的许多方法都存在这样的问题.为了解决这种模棱两可性,McGrath和Lin(2001)提出了一种基于残差样本方差几何平均的检验方法(ML方法),但是这个方法不能应用于零残差样本方差的情形.鉴于此,提出了一种基于修改残差的改进方法,适用于零残差样本方差的情形,并且通过实例验证了方法的合理性.最后,通过模拟和ML方法做了比较.     

随机估计和VDR检验

众所周知统计推断有三种理论:普遍承认的Neyman理论(频率学派),Bayes推断和信仰推断(Fiducial)。Bayes推断基于后验分布,由先验分布和样本分布求得。信仰推断是基于信仰分布(Confidence Distribution,简称CD),直接利用样本求得。两者推断方式一致,都是用分布函数作推断,称为分布推断。从分析传统的参数估计、假设检验特性来看,经典统计推断也可以视为分布推断。通常将置信上限看做置信度的函数。其反函数,即置信度是置信上界的函数,恰是分布函数,该分布恰是近年来引起许多学者兴趣的CD。在本文中,基于随机化估计(其分布是一CD)的概率密度函数,提出VDR检验。常见正态分布期望或方差的检验,多元正态分布期望的Hoteling检验等是其特例。VDR(vertical density representation)检验适合于多元分布参数检验,实现了非正态的多元线性变换分布族的参数检验。VDR构造的参数的置信域有最小Lebesgue测度。     

Numerical approximation of conditional asymptotic variances using Monte Carlo simulation

We consider in this paper the use of Monte Carlo simulation to numerically approximate the asymptotic variance of an estimator of a population parameter. When the variance of an estimator does not exist in finite samples, the variance of its limiting distribution is often used for inferences. However, in this case, the numerical approximation of asymptotic variances is less straightforward, unless their analytical derivation is mathematically tractable. The method proposed does not assume the existence of variance in finite samples. If finite sample variance does exist, it provides a more efficient approximation than the one based on the convergence of finite sample variances. Furthermore, the results obtained will be potentially useful in evaluating and comparing different estimation procedures based on their asymptotic variances for various types of distributions. The method is also applicable in surveys where the sample size required to achieve a fixed margin of error is based on the asymptotic variance of the estimator. The proposed method can be routinely applied and alleviates the complex theoretical treatment usually associated with the analytical derivation of the asymptotic variance of an estimator which is often managed on a case by case basis. This is particularly appealing in view of the advance of modern computing technology. The proposed numerical approximation is based on the variances of a certain truncated statistic for two selected sample sizes, using a Richardson extrapolation type formulation. The variances of the truncated statistic for the two sample sizes are computed based on Monte Carlo simulations, and the theory for optimizing the computing resources is also given. The accuracy of the proposed method is numerically demonstrated in a classical errors-in-variables model where analytical results are available for the purpose of comparisons.     

Comparison of two variances under inequality constraints by using empirical likelihood method

In this article, the empirical likelihood introduced by Owen Biometrika, 75, 237-249 （1988） is applied to test the variances of two populations under inequality constraints on the parameter space. One reason that we do the research is because many literatures in this area are limited to testing the mean of one population or means of more than one populations; the other but much more important reason is： even if two or more populations are considered, the parameter space is always without constraint. In reality, parameter space with some kind of constraints can be met everywhere. Nuisance parameter is unavoidable in this case and makes the estimators unstable. Therefore the analysis on it becomes rather complicated. We focus our work on the relatively complicated testing issue over two variances under inequality constraints, leaving the issue over two means to be its simple ratiocination. We prove that the limiting distribution of the empirical likelihood ratio test statistic is either a single chi-square distribution or the mixture of two equally weighted chi-square distributions.     

基于Jeffreys验前的Bayes方差估计修正

分析了基于Jeffreys验前的经典Bayes方差估计以及考虑验前信息可信度情况下Bayes方差估计存在的问题,在一般情况下,其方差估计要大于验前子样和验后子样的方差,这显然是不合理的.这是采用Jeffreys验前和正态共轭分布假设时存在的固有问题.为了解决这一问题,提出了方差估计的修正公式,经过计算验证,其值在验前子样和验后子样方差之间,说明修正公式是合理的.     

Triad count estimation in digraphs

One may use information about a random sample of network members to estimate quantities related to the triad census of a network. Various kinds of information about the graph may be observable from the sample; four observation schemes involving the local networks of the sampled vertices are considered here. Unbiased triad count estimators are defined, and their variances (and unbiased estimators of these variances) are derived. A main result is that under one of the observation schemes, the estimator can be written as a sum of vertex attributes; standard estimation formulas for various sampling designs, such as stratified sampling, are therefore effortlessly applied. The estimator properties are compared in a simulation study.     

Study of weighted Monte Carlo algorithms with branching

Various weighted algorithms for numerical statistical simulation are formulated and studied. The trajectory of an algorithm branches when the current weighting factor exceeds unity. As a result, the weight of an individual branch does not exceed unity and the variance of the estimate for the computed functional is finite. The unbiasedness and finiteness of the variance of estimates are analyzed using the recurrence “partial“ averaging method formulated in this study. The estimation of the particle reproduction factor and solutions to the Helmholtz equation are considered as applications. The comparative complexity of the algorithms is examined using a test problem. The variances of weighted algorithms with branching as applied to integral equations with power nonlinearity are analyzed.     

Degeneracy in Heteroscedastic Regression Models

The maximum likelihood estimation in a regression model with heteroscedastic errors is considered. When the design matrices in the model are inappropriately specified, the maximum likelihood estimates of the variances of certain observations are found to be zero irrespective of the observed values, resulting in degeneracy. Necessary and sufficient conditions for degeneracy are given and used for its avoidance.     

Accuracy of <Emphasis Type="Italic">p</Emphasis>-values of approximate tests in testing for equality of means under unequal variances

Seemingly, testing for fixed effects in linear models with variance-covariance components has been solved for decades. However, even in simple situations such as in fixed one-way model with heteroscedastic variances (a multiple means case of the Behrens-Fisher problem) the questions of statistical properties of various approximations of test statistics are still alive. Here we present a brief overview of several approaches suggested in the literature as well as those available in statistical software, accompanied by a simulation study in which the accuracy of p-values is studied. Our interest is limited here to the Welch’s test, the Satterthwaite-Fai-Cornelius test, the Kenward-Roger test, the simple ANOVA F-test, and the parametric bootstrap test. We conclude that for small sample sizes, regardless the number of compared means and the heterogeneity of variance, the ANOVA F-test p-value performs the best. For higher sample sizes (at least 5 per group), the parametric bootstrap performs well, and the Kenward-Roger test also performs well.     

General projection-pursuit estimators for the common principal components model: influence functions and Monte Carlo study

The common principal components (CPC) model for several groups of multivariate observations assumes equal principal axes but possibly different variances along these axes among the groups. Under a CPCs model, generalized projection-pursuit estimators are defined by using score functions on the dispersion measure considered. Their partial influence functions are obtained and asymptotic variances are derived from them. When the score function is taken equal to the logarithm, it is shown that, under a proportionality model, the eigenvector estimators are optimal in the sense of minimizing the asymptotic variance of the eigenvectors, for a given scale measure.