多元时间序列GARCH型模型的诊断检验

多元时间序列GARCH型模型已被证实在理论和实际中具有重要作用.该文对这一类模型的拟合优度提出了一组得分型检验统计量.这些检验在零假设模型下渐近服从卡方分布,计算简单,临界值容易得到.检验对备择模型比较敏感,能侦察到以1/n~(1/2)的速度收敛到零假设的备择模型.对于可能的多个备择,构造了渐近分布自由的Maximin检验;而对于饱和备择情形,基于得分型检验的思想提出了一个构造Omnibus检验的可能性.值得指出的是构造的这组检验能检测到零假设模型的条件协差阵的每一部分可能的偏离,从而当模型被错误指定时,该检验能提供相关信息进行模型修正.模拟结果表明该文的检验表现理想.     

分位数回归的光滑经验似然

对线性分位数回归模型参数的检验问题,文献证明了用最小绝对距离法(LAD法)及光滑经验似然法(SEL法)构造的检验统计量在原假设下都依分布收敛到x~2M.本文证明在局部备择假设下这两个检验统计量都依分布收敛到非中心卡方分布.文中用随机模拟比较了两种方法在局部备择假设下的表现,模拟结果表明在合理控制犯第一类错误的前提下,用SEL法构造的检验比用LAD法构造的检验更有效.     

时间序列中向量自回归模型的拟合优度检验

本文提出并研究了一些诊断检验工具,用于一般参数型的时间序列向量自回归模型的拟合优度检验．该检验在零假设下渐近服从卡方分布,并能侦察到以参数速度收敛到零假设模型的备择模型．检验涉及到权函数,因此可以灵活地选择权函数以提高检验功效,尤其是在可能的偏离方向已知情形．如果备择不是方向型的,而只知道其属于某一个模型类中,此时可构造一个渐近分布自由的极大极小（maximin）检验．对于饱和备择,基于得分型思想给出了构造万能（omnibus）检验的可行性构想．本文对提出的检验从理论上进行了功效研究．另外,为提高检验在小样本情形的功效,本文把非参数Monte Carlo检验方法推广到相依数据情形．最后,通过模拟研究和实际数据分析进一步表明检验的有用性．     

上界型拟合优度检验

对简单零假设情况,构造出一类上界型拟合优度检验.取不同的参数λ和不同的权函数,这类检验不仅包含许多已存在的检验,如Kolmogorov-Smirov检验,Berk-Jones检验等,而且还给出一些新的检验.众所周知,对不同的问题,"最优"的检验是不同的,有必要对这类检验的性质进行讨论.该文对任意给定的λ和较一般的权函数q(·),在较弱的条件下,导出了相应上界型检验统计量在零假设下的渐近分布,研究了它们的局部渐近功效;在若干固定备择假设下,对该类检验的功效进行了模拟研究.模拟结果表明,在不同的备择假设下,功效较优的检验是不同的,不存在对所有情况一致最优的检验.     

函数型二次模型的假设检验

本文研究了函数型二次回归中二次参数函数的显著性检验问题。采用函数型主成分分析将预测变量函数进行投影降维,利用零模型和全模型的残差平方和构造F型检验统计量。在一定的正则条件下证明了检验统计量在原假设下渐近于F分布,在备择假设下检验统计量依概率趋于无穷,从而表明该检验方法是相合的。进一步证明了在一定收敛速度的局部备择假设下,检验统计量渐近于非中心F分布。最后通过数值模拟研究了该检验方法在有限样本下的表现,并给出了一个实际例子进一步验证所提方法的有效性。     

多项分布最大概率的检验

考察了对多项分布最大概率p_[1]的单边假设检验. 将Ethier (1982)的一个小样本检验推广到了一般情形. 基于一个对p_[1]的估计, 给出了一类大样本检验. 求出了在局部备择假设下以上这些检验的渐近功效. 最后给出了一个实例分析.     

长相依序列均值变点检测

研究了长相依序列均值变点检测问题.首先给出变点检验的Ratio统计量其次分别推导了原假设和备择假设下统计量的极限分布,最后用蒙特卡洛方法模拟出检验的临界值,并通过数值模拟和实例分析说明方法的有效性.     

浅谈假设检验中原假设和备择假设的建立

建立原假设和备择假设是进行假设检验的第一步.如何设立假设检验中的原假设和备择假设,是学生和初学者普遍困扰的问题.而根据奈曼—皮尔逊准则,原假设和备择假设的地位不对等.本文对如何建立假设检验中的原假设和备择假设进行了讨论,给出了几个建立原假设和备择假设的原则.     

生长曲线模型中的球性检验问题

本文给出了生长曲线模型中球性检验似然比准则在原假设相接近的两类备择假设下的非零渐近分布。     

一类相合的分位点型检验

广义非参数似然比检验统计量是一类很广的统计量,包含了众多重要的检验统计量,如Anderson-Darling(AD)等.利用Rubin的随机经验分布函数替代经验分布函数的方法,得到了广义非参数似然比检验统计量的新版本,构造了新的检验统计量.由于新的检验统计量在给定样本下仍然是随机变量,选择了它的分位点和期望作为检验统计量,分别称之为分位点型检验统计量和期望型检验统计量.在简单假设情况下,证明了分位点型检验统计量和期望型检验统计量在固定备择下的相合性.模拟结果显示,在某些备择下,新的检验的功效明显高于原有的基于经验分布函数的检验的功效.     

Invariant Tests for Covariance Structures in Multivariate Linear Model

The null hypothesis that the error vectors in a multivariate linear model are independent is tested against the alternative hypothesis that they are dependent in some specified manner. This dependence is assumed to be due to common random components or autocorrelation over time. The testing problem is solved by classical invariance arguments under multinormality. The most powerful invariant test usually depends on the particular alternative and may even lack a closed form expression. Then the locally best test is derived. The power is maximized at the null hypothesis in the direction of some alternative. In most applications the direction where the maximization is performed does not enter the test. Then the locally uniformly best test exists. Several applications are outlined.     

A novel trace test for the mean parameters in a multivariate growth curve model

A trace test for the mean parameters of the growth curve model is proposed. It is constructed using the restricted maximum likelihood followed by an estimated likelihood ratio approach. The statistic reduces to the Lawley-Hotelling trace test for the Multivariate Analysis of Variance (MANOVA) models. Our test statistic is, therefore, a natural extension of the classical trace test to GMANOVA models. We show that the distribution of the test under the null hypothesis does not depend on the unknown covariance matrix Σ. We also show that the distributions under the null and alternative hypotheses can be represented as sums of weighted central and non-central chi-square random variables, respectively. Under the null hypothesis, the Satterthwaite approximation is used to get an approximate critical point. A novel Satterthwaite type approximation is proposed to obtain an approximate power. A simulation study is performed to evaluate the performance of our proposed test and numerical examples are provided as illustrations.     

Asymptotic power comparison of the chi-square and likelihood ratio tests

Summary A Hoeffding-type power comparison is made between the likelihood ratio and chi-square tests for a simple hypothesis in a multinomial. The power comparison is based on the fact that under an alternative hypothesis the distribution of the test statistic can be approximated by a normal distribution. The theory of large deviations is used to match the significance levels. This research was partially supported by the National Science Foundation Grant No. GP-33697X2, U.S. Energy Research and Development Agency Grant 7064100, Environmental Protection Agency Grant R805379-01-0 and U.S. Public Health Service Grant USPHS ES01299-15, at the University of California-Berkeley.     

A contribution to the statistical theory of linear graduation

Moving (weighted) average techniques provide a flexible tool for the graduation (smoothing) of vital rates. As is shown in the present paper, one may construct linear graduation methods which are superior with respect to some formal evaluation criteria, but such ‘improved’ methods lack an important robustness property which the moving average graduation techniques possess and which is a main motivation for their use.A linear graduation technique is an estimation method, and an important consideration in the evaluation of its properties is the class for which it is unbiased, i.e., the class of functions which it reproduces unchanged. A classical test usually said to be a chi-square test of the goodness-of-fit of a graduation is revealed to be a test of the hypothesis that the graduation method is indeed unbiased for the set of theoretical rates actually estimated, and its properties are investigated along with those of alternative chi-square tests.     

Empirical likelihood-based subset selection for partially linear autoregressive models

Based on the empirical likelihood method, the subset selection and hypothesis test for parameters in a partially linear autoregressive model are investigated. We show that the empirical log-likelihood ratio at the true parameters converges to the standard chi-square distribution. We then present the definitions of the empirical likelihood-based Bayes information criteria （EBIC） and Akaike information criteria （EAIC）. The results show that EBIC is consistent at selecting subset variables while EAIC is not. Simulation studies demonstrate that the proposed empirical likelihood confidence regions have better coverage probabilities than the least square method, while EBIC has a higher chance to select the true model than EAIC.     

检验两个部分线性模型中非参函数相等

We propose the test statistic to check whether the nonparametric func-tions in two partially linear models are equality or not in this paper. We estimate the nonparametric function both in null hypothesis and the alternative by the local linear method, where we ignore the parametric components, and then estimate the parameters by the two stage method. The test statistic is derived, and it is shown to be asymptotically normal under the null hypothesis.     

Incorporating historical information in testing for a trend in Poisson means

An exact conditional test is developed for testing a trend in Poisson means when the historical control information is incorporated into the concurrent control data. An asymptotic conditional test is also developed as an alternative to the Tarone test. Asymptotic gains by the incorporation of the historical information is evaluated.     

New families of estimators and test statistics in log-linear models

In this paper we consider categorical data that are distributed according to a multinomial, product-multinomial or Poisson distribution whose expected values follow a log-linear model and we study the inference problem of hypothesis testing in a log-linear model setting. The family of test statistics considered is based on the family of ?-divergence measures. The unknown parameters in the log-linear model under consideration are also estimated using ?-divergence measures: Minimum ?-divergence estimators. A simulation study is included to find test statistics that offer an attractive alternative to the Pearson chi-square and likelihood-ratio test statistics.     

Measurement‐error detection: international evidence on industrial production

The existence or non‐existence of measurement error (ME) in observed time series is examined not by hypothesis testing but by model selection using the values of information criteria of a battery of alternative models with and without ME. Whether the time series contains ME is determined as a result of model selection. This method is compared with recently proposed hypothesis‐testing method. Simulation results suggest that the performances of the proposed method are usually comparable to and sometimes better than those of the hypothesis‐testing method. The proposed method is applied to monthly time series of industrial production for fifteen developed countries. Obtained results indicate that MEs are detected for at least one and at most seven countries. Copyright © 2006 John Wiley & Sons, Ltd.