线性模型中F-检验的稳健性

本文讨论了一般线性模型中关于均值参数β的线性假设基于广义最小二乘估计的F-检验统计量的稳健性问题.主要研究了当误差的协方差矩阵含有参数时,设计阵可以列降秩情况下的F-检验统计量的稳健性,得到了F(V(θ))为该假设下F-检验统计量的误差协方差矩阵的最大类.并讨论了分块线性模型中,关于分块参数的线性假设的F-检验统计量的稳健性.     

The general linear model and the generalized singular value decomposition

A new derivation is given for the generalized singular value decomposition of two matrices X and F having the same number of rows. It is shown how this decomposition reveals the structure of the general Gauss-Markov linear model (y, Xβ, σ²FF′), and exhibits the structure and solution of the generalized linear least squares problem used to provide the best linear unbiased estimator for the model. The decomposition is used to prove optimality of the estimator and to reveal the structure of the covariance matrix of the error of the estimator.     

The asymptotic validity of invariant procedures for the repeated measures model and multivariate linear model

Fairly general sufficient conditions are given to guarantee that invariant tests about means in the multivariate linear model and the repeated measures model have the correct asymptotic size when the normal assumption under which the tests are derived is relaxed. These conditions are the same as Huber's condition which guarantees asymptotic validity of the size of the F-test for the univariate linear model.     

Growth Curve Model with Hierarchical Within-Individuals Design Matrices

This paper deals with some inferential problems under an extended growth curve model with several hierarchical within-individuals design matrices. The model includes the one whose mean structure consists of polynomial growth curves with different degrees. First we consider the case when the covariance matrix is unknown positive definite. We derive a LR test for examining the hierarchical structure for within individuals design matrices and a model selection criterion. Next we consider the case when a random coefficients covariance structure is assumed, under certain assumption of between-individual design matrices. Similar inferential problems are also considered. The dental measurement data (see, e.g., Potthoff and Roy (1964, Biometrika, 51, 313-326)) is reexamined, based on extended growth curve models.     

On the Gaussian approximation of vector-valued multiple integrals

By combining the findings of two recent, seminal papers by Nualart, Peccati and Tudor, we get that the convergence in law of any sequence of vector-valued multiple integrals F_n towards a centered Gaussian random vector N, with given covariance matrix C, is reduced to just the convergence of: (i) the fourth cumulant of each component of F_n to zero; (ii) the covariance matrix of F_n to C. The aim of this paper is to understand more deeply this somewhat surprising phenomenon. To reach this goal, we offer two results of a different nature. The first one is an explicit bound for d(F,N) in terms of the fourth cumulants of the components of F, when F is a R^d-valued random vector whose components are multiple integrals of possibly different orders, N is the Gaussian counterpart of F (that is, a Gaussian centered vector sharing the same covariance with F) and d stands for the Wasserstein distance. The second one is a new expression for the cumulants of F as above, from which it is easy to derive yet another proof of the previously quoted result by Nualart, Peccati and Tudor.     

The asymptotic validity of theF-test in a two-stage least squares model

The usualF-test has been used to test a general linear hypothesis for a two-stage least squares method in a system of economic equations. However, we find that thisF-test is actually asymptotically invalid. Some suggestions are given for testing a general linear hypothesis in this situation.     

Improved Estimation of Parameter Matrices in a One-Sample and Two-Sample Problems

In this paper, the problems of estimating the covariance matrix in a Wishart distribution (refer as one-sample problem) and the scale matrix in a multi-variate F distribution (which arise naturally from a two-sample setting) are considered. A new class of estimators which shrink the eigenvalues towards their harmonic mean is proposed. It is shown that the new estimator dominates the best linear estimator under two scale invariant loss functions.     

Estimation of multivariate normal covariance and precision matrices in a star-shape model with missing data

In this paper, we study the problem of estimating the covariance matrix Σ and the precision matrix Ω (the inverse of the covariance matrix) in a star-shape model with missing data. By considering a type of Cholesky decomposition of the precision matrix Ω=Ψ^′Ψ, where Ψ is a lower triangular matrix with positive diagonal elements, we get the MLEs of the covariance matrix and precision matrix and prove that both of them are biased. Based on the MLEs, unbiased estimators of the covariance matrix and precision matrix are obtained. A special group G, which is a subgroup of the group consisting all lower triangular matrices, is introduced. By choosing the left invariant Haar measure on G as a prior, we obtain the closed forms of the best equivariant estimates of Ω under any of the Stein loss, the entropy loss, and the symmetric loss. Consequently, the MLE of the precision matrix (covariance matrix) is inadmissible under any of the above three loss functions. Some simulation results are given for illustration.     

Testing some covariance structures under a growth curve model

This paper considers three types of problems: (i) the problem of independence of two sets, (ii) the problem of sphericity of the covariance matrix Σ, and (iii) the problem of intraclass model for the covariance matrix Σ, when the column vectors of X are independently distributed as multivariate normal with covariance matrix Σ and E(X) = BξA,A and B being given matrices and ξ and Σ being unknown. These problems are solved by the likelihood ratio test procedures under some restrictions on the models, and the null distributions of the test statistics are established.     

On the power of Wilks' U-test for MANOVA

Let V₁,…, V_m, W₁,…, W_n be independent p × 1 random vectors having multivariate normal distributions with common nonsingular covariance matrix Σ and with EW_α = 0, α = 1,…, n. In this canonical form of the multivariate linear model, the problem is to test H: EV_αazμ_α = 0, α = 1,…, m vs K: not H. It is shown that when the rank of the noncentrality matrix (μ₁…μ_m) Σ^?1 (μ₁…μ_m) is one, the power of Wilks' U-test (the likelihood ratio test) strictly decreases with the dimension p and the hypothesis degrees of freedom m. This generalizes results known for the noncentral F-test in the univariate case.     

Existence of the uniformly minimum risk equivariant estimators of parameters in a class of normal linear models

In this paper, we study the existence of the uniformly minimum risk equivariant (UMRE) estimators of parameters in a class of normal linear models, which include the normal variance components model, the growth curve model, the extended growth curve model, and the seemingly unrelated regression equations model, and so on. The necessary and sufficient conditions are given for the existence of UMRE estimators of the estimable linear functions of regression coefficients, the covariance matrixV and (trV)^α, where α > 0 is known, in the models under an affine group of transformations for quadratic losses and matrix losses, respectively. Under the (extended) growth curve model and the seemingly unrelated regression equations model, the conclusions given in literature for estimating regression coefficients can be derived by applying the general results in this paper, and the sufficient conditions for non-existence of UMRE estimators ofV and tr(V) are expanded to be necessary and sufficient conditions. In addition, the necessary and sufficient conditions that there exist UMRE estimators of parameters in the variance components model are obtained for the first time.     

Nonnegative quadratic estimation and quadratic sufficiency in general linear models

Notions of linear sufficiency and quadratic sufficiency are of interest to some authors. In this paper, the problem of nonnegative quadratic estimation for β^′Hβ+hσ² is discussed in a general linear model and its transformed model. The notion of quadratic sufficiency is considered in the sense of generality, and the corresponding necessary and sufficient conditions for the transformation to be quadratically sufficient are investigated. As a direct consequence, the result on (ordinary) quadratic sufficiency is obtained. In addition, we pose a practical problem and extend a special situation to the multivariate case. Moreover, a simulated example is conducted, and applications to a model with compound symmetric covariance matrix are given. Finally, we derive a remark which indicates that our main results could be extended further to the quasi-normal case.     

A Generalization of Rao's Covariance Structure with Applications to Several Linear Models

This paper presents a generalization of Rao's covariance structure. In a general linear regression model, we classify the error covariance structure into several categories and investigate the efficiency of the ordinary least squares estimator (OLSE) relative to the Gauss–Markov estimator (GME). The classification criterion considered here is the rank of the covariance matrix of the difference between the OLSE and the GME. Hence our classification includes Rao's covariance structure. The results are applied to models with special structures: a general multivariate analysis of variance model, a seemingly unrelated regression model, and a serial correlation model.     

Admissibilities of linear estimator in a class of linear models with a multivariate t error variable

This paper discusses admissibilities of estimators in a class of linear models,which include the following common models:the univariate and multivariate linear models,the growth curve model,the extended growth curve model,the seemingly unrelated regression equations,the variance components model,and so on.It is proved that admissible estimators of functions of the regression coefficient β in the class of linear models with multivariate t error terms,called as Model II,are also ones in the case that error terms have multivariate normal distribution under a strictly convex loss function or a matrix loss function.It is also proved under Model II that the usual estimators of β are admissible for p 2 with a quadratic loss function,and are admissible for any p with a matrix loss function,where p is the dimension of β.     

A Factorization problem and the problem of predicting non-stationary vector-valued stochastic processes

Summary In this paper a prediction theory is developed under the general idea that the infinite dimensional covariance matrix is a self-adjoint element in a symmetric Banach algebra. The usual Wiener's spectral factorization method for solving stationary Wiener-Hopf equations has been extended to this algebra. Finally, a theorem for factoring a positive definite covariance matrix into upper and lower triangular factors with similar inverses has been proved.     

A Note on Random Effects Growth Curve Models

The robustness of regression coefficient estimator is a hot topic in regression analysis all the while. Since the response observations are not independent, it is extraordinarily difficult to study this problem for random effects growth curve models, especially when the design matrix is non-full of rank. The paper not only gives the necessary and sufficient conditions under which the generalized least square estimate is identical to the the best linear unbiased estimate when error covariance matrix is an arbitrary positive definite matrix, but also obtains a concise condition under which the generalized least square estimate is identical to the maximum likelihood estimate when the design matrix is full or non-full of rank respectively. In addition, by using of the obtained results, we get some corollaries for the the generalized least square estimate be equal to the maximum likelihood estimate under several common error covariance matrix assumptions. Illustrative examples for the case that the design matrix is full or non-full of rank are also given.     

Thompson metric method for solving a class of nonlinear matrix equation

Based on the elegant properties of the Thompson metric, we prove that the general nonlinear matrix equation X^q-A^∗F(X)A=Q(q>1) always has a unique positive definite solution. An iterative method is proposed to compute the unique positive definite solution. We show that the iterative method is more effective as q increases. A perturbation bound for the unique positive definite solution is derived in the end.     

A method for approximations to the pdf'S and cdf'S of GLSE's and its application to the seemingly unrelated regression model

This paper first develops a valid method for approximations to the pdf's and cdf's of GLSE in linear models and, applying this method to the Zellner estimator with an unrestricted sample covariance in the seemingly unrelated regression model, obtains an approximate pdf with a bound of ordern ⁻² and an approximate covariance matrix with a bound of ordern ⁻³ This research was done at the London School of Economics while the authors were British Council scholars. Kariya is grateful to Professor J. Durbin for a general discussion on asymptotic expansions. Further the authors deeply appreciate Professor Y. Kataoka and anonymous referee for their invaluable comments and suggestions.     

On the robustness of least squares procedures in regression models

The criterion robustness of the standard likelihood ratio test (LRT) under the multivariate normal regression model and also the inference robustness of the same test under the univariate set up are established for certain nonnormal distributions of errors. Restricting attention to the normal distribution of errors in the context of univariate regression models, conditions on the design matrix are established under which the usual LRT of a linear hypothesis (under homoscedasticity of errors) remains valid if the errors have an intraclass covariance structure. The conditions hold in the case of some standard designs. The relevance of C. R. Rao's (1967 In Proceedings Fifth Berkeley Symposium on Math. Stat. and Prob., Vol. 1, pp. 355–372) and G. Zyskind's (1967, Ann. Math. Statist.38 1092–1110) conditions in this context is discussed.     

A general model for consecutive-k-out-of-n: F repairable system with exponential distribution and (k−1)-step Markov dependence

In this paper, a general model for consecutive-k-out-of-n: F repairable system with exponential distribution and (k−1)-step Markov dependence is introduced. The lifetime of a component is an exponential random variable, its parameter depends on the number of consecutive failed components that precede the component. The repair time is also an exponential random variable. A priority repair rule on the basis of the system failure risk is adopted. Then the transition density matrix of the system is determined. Some reliability indices, including the system availability, rate of occurrence of failures and reliability are evaluated accordingly. For the demonstration of the model and methodology, a linear system example and a circular system example are investigated.