Subsampling Unit Root Tests for Heavy-Tailed Observations

It is shown that the subsampling methodology can be used to develop unit root tests when the noise sequence is heavy-tailed with infinite variance. Using least-squares residuals, we construct processes which approximately satisfy the null hypothesis and then, using subsampling, we approximate the null distribution of test statistics. We establish the asymptotic validity of this method and demonstrate its applicability in finite samples by means of a simulation study and a data example.     

Asymptotic expansions of the distributions of some test statistics

Summary A modified Wald statistic for testing simple hypothesis against fixed as well as local alternatives is proposed. The asymptotic expansions of the distributions of the proposed statistic as well as the Wald and Rao statistics under both the null and alternative hypotheses are obtained. The powers of these statistics are compared and its is shown that for special structures of parameters some statistics have same power in the sence of order . The results obtained are applied for testing the hypothesis about the covariance matrix of the multivariate normal distribution and it is shown that none of the tests based on the above statistics is uniformly superior. Research supported by the National Science Foundation Grant MCS 830149.     

Design of component reliability test plan for a series system having time dependent testing cost with the presence of covariates

Consider a series system with n different components. Assume that the lifetime of \(i\text {-}th\) component follows exponential distribution with parameter \( \lambda _{i}(\mathbf {x}), \ \mathbf {x} \in \mathbb {R}^{k} \) is a covariate vector, \( 1\le i \le n \). For example, \( \mathbf {x} \) may be (temperature, pressure, humidity), so that \( \mathbf {x} \in \mathbb {R}^{3} \) represents a three dimensional covariate vector. Assume that each \( \lambda _{i}(\mathbf {x}),1\le i \le n \), is distinct and depends upon \( \mathbf {x} \) through linear relationship. Prior information available in the form of upper bounds on \( \lambda _{i}(\mathbf {x}) \) are also incorporated in the design. We propose to obtain optimal reliability test plan based on maximum likelihood estimator of system reliability. A non-linear integer optimization problem is formulated for minimizing the maximum of total expected cost involved in testing satisfying usual probability requirements (Type-I and Type-II error constraints). In addition, it is also established through simulation that the derived sampling plan meets the specified producer’s and consumer’s risks as well. A sensitivity analysis is carried out to study the effect of various input parameters on maximum total expected testing cost. Finally, a qualitative analysis is presented at the end to discuss the nature of sampling plan derived. Several numerical examples are discussed to illustrate our test plan, and it is observed that the proposed plan has significant potential to reduce the total number of components to be tested for failure. It is noted that the number of components to be tested for failure is reduced by about 96% as compared to the existing test plans in the literature.     

On limit laws for sums of Pfeifer records

The problem of determining limiting distributions for sums of records has been studied by several authors who have considered a variety of assumptions sufficient to ensure that sums of records properly normalized will converge to a non-degenerate distribution. As a parallel to these endeavors, it is of interest to establish conditions under which the sum of Pfeifer records, properly normalized, converges. Pfeifer records are defined under the assumption that initial observations are i.i.d. with common survival function and following the (n−1)-th record value the observations are assumed to have survival function ,n=1,2,.... The study of the asymptotic behavior of sums of Pfeifer records constitutes a natural generalization of work on sums of classical records. The present paper introduces conditions under which the limit distribution of sums of Pfeifer records is non-degenerate.      

A Weighted Goodness-of-Fit Test for GARCH(1,1) Specification

Suppose that (j) is the lag-j autocorrelation of the squared residuals computed from a realization of length n under the assumption that the observations follow a GARCH(1,1) model. We study the asymptotic distribution of the statistics of the form , where the _j are nonnegative summable weights and the matrix , can be estimated from the data. We show that, under weak assumptions on model errors, the statistic Q _n converges in distribution to , where the N _i are iid standard normal. We discuss choices of the weights _j for which the distribution of Q is tabulated. Our results lead to and provide a rigorous justification for Portmanteau goodness-of-fit tests for GARCH(1,1) specification.     

A Note on Extremes of Concomitants of Order Statistics

Let be a sequence of identically distributed variables. We study the asymptotic distribution of , where Y _[r:n] denotes the concomitant of the rth order statistic X _r:n , corresponding to , and is held fixed while . Conditions are given for the and to have the same asymptotic behavior as that we would apply if were i.i.d. The result is illustrated with a simple linear regression model , where is a stationary sequence with extremal index .     

Robust tests in group sequential analysis: One- and two-sided hypotheses in the linear model

Consider the linear modelY=X+E in the usual matrix notation where the errors are independent and identically distributed. We develop robust tests for a large class of one- and two-sided hypotheses about when the data are obtained and tests are carried out according to a group sequential design. To illustrate the nature of the main results, let and be anM- and the least squares estimator of respectively which are asymptotically normal about with covariance matrices ²(X ^t X)^–1 and ²(X ^t X)^–1 respectively. Let the Wald-type statistics based on and be denoted byRW andW respectively. It is shown thatRW andW have the same asymptotic null distributions; here the limit is taken with the number of groups fixed but the numbers of observations in the groups increase proportionately. Our main result is that the asymptotic Pitman efficiency ofRW relative toW is (²/²). Thus, the asymptotic efficiency-robustness properties of relative to translate to asymptotic power-robustness ofRW relative toW. Clearly, this is an attractive result since we already have a large literature which shows that is efficiency-robust compared to . The results of a simulation study show that with realistic sample sizes,RW is likely to have almost as much power asW for normal errors, and substantially more power if the errors have long tails. The simulation results also illustrate the advantages of group sequential designs compared to a fixed sample design, in terms of sample size requirements to achieve a specified power.     

Parameter Estimates in Random Intercept Mixed Effects Model for Repeated Measures

In this article the following random intercept mixed effects model will be considered： yij = vi =v^τijβ＋ εij,i=1,…,m;j=1,2,…,ni, where {vi} are i.i.d, random effects with mean α 2. 2 and finite variance σ^2 v, {εij} are i.i.d, random errors with finite variance ε^2 ε. Here we will estimate α,σ^2 v,σ^2 ε,β and study their large sample properties, such as strong consistency, strong convergence rates and asymptotic normality.     

基于主Hessian方向的中央均值子空间边际坐标检验

本文给出了基于两种相近的主Hessian方向方法的边际坐标检验. 这种检验方法能够非常有效的识别自变量对于回归均值中央子空间的贡献. 此外, 与利用切片逆回归和切片平均方差估计的检验方法不同的是, 本文中主Hessian方向的检验方法可以避免对切片数目的选择. 我们证明了检验统计量在原假设下的渐近分布, 并且通过模拟, 证实了检验的有效性.     

Asymptotic distributions of the likelihood ratio test statistics for covariance structures of the complex multivariate normal distributions

In this paper, the authors derived asymptotic expressions for the null distributions of the likelihood ratio test statistics for multiple independence and multiple homogeneity of the covariance matrices when the underlying distributions are complex multivariate normal. Also, asymptotic expressions are obtained in the non-null cases for the likelihood ratio test statistics for independence of two sets of variables and the equality of two covariance matrices. The expressions obtained in this paper are in terms of beta series. In the null cases, the accuracy of the first terms alone is sufficient for many practical purposes.     

The Asymptotic Distributions of Sums of Records

LetX ₁,X ₂... be a sequence of i.i.d. random variables and let be the associated (upper) record sequence. Resnick (1973) identified the class of all possible limit distributions for . Here we focus on sums of records, . We describe three cases in which T _n can be normalized to have a non-trivial limiting distribution. The problem of identifying all possible limit laws for normalized sums of records remains open.     

Conditional limit results for type I polar distributions

Let (S ₁,S ₂) = (R cos(Θ), R sin(Θ)) be a bivariate random vector with associated random radius R which has distribution function F being further independent of the random angle Θ. In this paper we investigate the asymptotic behaviour of the conditional survivor probability when u approaches the upper endpoint of F. On the density function of Θ we impose a certain local asymptotic behaviour at 0, whereas for F we require that it belongs to the Gumbel max-domain of attraction. The main result of this contribution is an asymptotic expansion of , which is then utilised to construct two estimators for the conditional distribution function . Furthermore, we allow Θ to depend on u.      

On the Relative Approximation Error of the Generalized Pareto Approximation for a High Quantile

Let F be a distribution function in the domain of attraction of an extreme value distribution . In case 0 and F has an infinite end-point, we study the asymptotic behaviour of the relative approximation error of a high quantile such that , where the order tends to 0. We use the approximation of the excesses over a high threshold u by a Generalized Pareto distribution. We give sufficient conditions under which tends to 0.AMS 2000 Subject Classification Primary—60G70, Secondary—62G20, 62G32     

Bounds for the Characteristic Functions of the System of Monomials in Random Variables and of Its Trigonometric Analogue

Using a multidimensional analogue of Vinogradov's inequality for a trigonometric integral, the upper bounds are constructed for the moduli of the characteristic functions both of the system of monomials in components of a random vector with an absolutely continuous distribution in and of the system where is uniformly distributed in [0, l]^s.     

A unifying theory of a posteriori error control for nonconforming finite element methods

Residual-based a posteriori error estimates were derived within one unifying framework for lowest-order conforming, nonconforming, and mixed finite element schemes in Carstensen [Numer Math 100:617–637, 2005]. Therein, the key assumption is that the conforming first-order finite element space annulates the linear and bounded residual ℓ written . That excludes particular nonconforming finite element methods (NCFEMs) on parallelograms in that . The present paper generalises the aforementioned theory to more general situations to deduce new a posteriori error estimates, also for mortar and discontinuous Galerkin methods. The key assumption is the existence of some bounded linear operator with some elementary properties. It is conjectured that the more general hypothesis (H1)–(H3) can be established for all known NCFEMs. Applications on various nonstandard finite element schemes for the Laplace, Stokes, and Navier–Lamé equations illustrate the presented unifying theory of a posteriori error control for NCFEM. Supported by DFG Research Center MATHEON “Mathematics for key technologies” in Berlin and the German Indian Project DST-DAAD (PPP-05). J. Hu was partially supported by National Science Foundation of China under Grant No.10601003.     

Thresholds of moving average of stationary processes for given on target significant levels

A Markov chain is associated to a finite order one-sided moving average of a discrete time stationary Gaussian process. A method is developed to specify thresholds for given on target significant levels in the sense that in the long run the probability that the moving average process lies in [L _i , L _i+1), will be π _i , i = 0,. . . ,m. Special inputs, AR(1) and MA(1) are treated in details. This article extends the work of Soltani et al. in (Commun Stat Theory Methods 36(14):2595–2606) where the inputs were assumed to be i.i.d.; and a single threshold was considered. This research was supported by Kuwait University, Research Administration, Research Grant No.[SS08/06].     

Robust <Emphasis Type="Italic">U</Emphasis>-type test for high dimensional regression coefficients using refitted cross-validation variance estimation

This paper aims to develop a new robust U-type test for high dimensional regression coefficients using the estimated U-statistic of order two and refitted cross-validation error variance estimation. It is proved that the limiting null distribution of the proposed new test is normal under two kinds of ordinary models. We further study the local power of the proposed test and compare with other competitive tests for high dimensional data. The idea of refitted cross-validation approach is utilized to reduce the bias of sample variance in the estimation of the test statistic. Our theoretical results indicate that the proposed test can have even more substantial power gain than the test by Zhong and Chen (2011) when testing a hypothesis with outlying observations and heavy tailed distributions. We assess the finite-sample performance of the proposed test by examining its size and power via Monte Carlo studies. We also illustrate the application of the proposed test by an empirical analysis of a real data example.     

Asymptotic expansion in the central limit theorem for quadratic forms

We consider statistics of the form

Second-order Linearity of Wilcoxon Statistics

The rank statistic , with R _i (t) being the rank of and e ₁ , . . . , e _n being the random sample from a distribution with a cdf F, is considered as a random process with t in the role of parameter. Under some assumptions on c _i , x _i and on the underlying distribution, it is proved that the process converges weakly to the Gaussian process. This generalizes the existing results where the one-dimensional case was considered. We believe our method of the proof can be easily modified for the signed-rank statistics of Wilcoxon type. Finally, we use our results to find the second order asymptotic distribution of the R-estimator based on the Wilcoxon scores and also to investigate the length of the confidence interval for a single parameter β _l .