Assessing Local Influence in Canonical Correlation Analysis

The first order local influence approach is adopted in this paper to assess the local influence of observations to canonical correlation coefficients, canonical vectors and several relevant test statistics in canonical correlation analysis. This approach can detect different aspects of influence due to different perturbation schemes. In this paper, we consider two different kinds, namely, the additive perturbation scheme and the case-weights perturbation scheme. It is found that, under the additive perturbation scheme, the influence analysis of any canonical correlation coefficient can be simplified to just observing two predicted residuals. To do the influence analysis for canonical vectors, a scale invariant norm is proposed. Furthermore, by choosing proper perturbation scales on different variables, we can compare the different influential effects of perturbations on different variables under the additive perturbation scheme. An example is presented to illustrate the effectiveness of the first order local influence approach.     

Influence Diagnostics in Common Principal Components Analysis

In principal components analysis, the influence function and local influence approaches have been well established as important diagnostic tools. In this article, we first review the generalized local influence approach in the restricted likelihood framework. We then apply the restricted likelihood local influence diagnostic in the common principal components analysis. One special part of this local influence result is an elliptical norm of the empirical influence function, which is comparable to the deletion diagnostic scaled by the same matrix which requires iterative solutions for parameter estimates with every case deleted. Local influence diagnostics are constructed by some basic building blocks that are obtained directly from the maximum likelihood estimates of the parameters, and which are based on the original data and thus require less computation. A numerical example illustrates the technique and some joint influence effects are identified by the proposed method.     

Katz Correspondence for Quasi-Unipotent Overconvergent Isocrystals

Let k be a field and X = Spec (k[t,t ^-1 ]). Katz proved that a differential equations with coefficients in k((t ^-1 )) is uniquely extended to a special algebraic differential equation on X when k is of characteristic 0. He also proved that a finite extension of k((t ^-1 )) is uniquely extended to a special covering of X when k is of any characteristic. These theorems are called canonical extension or Katz correspondence. We shall prove a p-adic analogue of canonical extension for quasi-unipotent overconvergent isocrystals. As a consequence, we can show that the local index of a quasi-unipotent overconvergent is equal to its Swan conductor.     

Robust statistical modeling using the Birnbaum‐Saunders‐t distribution applied to insurance

In this paper, we carry out robust modeling and influence diagnostics in Birnbaum‐Saunders (BS) regression models. Specifically, we present some aspects related to BS and log‐BS distributions and their generalizations from the Student‐t distribution, and develop BS‐t regression models, including maximum likelihood estimation based on the EM algorithm and diagnostic tools. In addition, we apply the obtained results to real data from insurance, which shows the uses of the proposed model. Copyright © 2011 John Wiley & Sons, Ltd.     

Influence Analysis for Linear Measurement Error Models

In this paper, we present a unified diagnostic method for linear measurement error models based upon the corrected likelihood of Nakamura (1990, Biometrika, 77, 127–137). Both global influence and local influence are discussed. The case-deletion model and mean-shift outlier model are considered, and they are shown to be approximately equivalent. Several diagnostic measures are derived and discussed. It is found that they can be written in terms of the residual and leverage measure. Some existing results are improved. Numerical example illustrates that our method is useful for diagnosing influential observations.     

Diagnostics on LALR(k) conflicts based on a method for LR(k) Testing

A user of an LALR(k) parser generator system may have difficulties in understanding how a given LALR(k) conflict is generated. This is especially difficult if the conflict does not correspond to an LR(k) conflict.A practical method for giving informative diagnostics on LALR(k) conflicts is presented. The diagnostics distinguish between those LALR(k) conflicts that correspond to LR(k) conflicts and those that do not. As a side effect the user is thus informed whether or not his grammar is in fact LR(k) despite not being LALR(k).The method is based on an algorithm for testing LR(k)-ness using the LR(0) machine supplied with LALR(k) lookahead sets. This algorithm is presented and its correctness is proved.     

Generalized symmetric spaces,Yu. P. Solovyov’s formula,and the generalized Hermitian geometry

We collect some basic results on canonical affinor structures of classical type on generalized symmetric spaces. Yu. P. Solovyov’s stimulating influence on this topic during its initial stages is illustrated. Using special canonical f-structures on homogeneous k-symmetric spaces, we also present a new collection of homogeneous Hermitian f-manifolds. Dedicated to the memory of Yu. P. Solovyov Translated from Fundamentalnaya i Prikladnaya Matematika, Vol. 13, No. 8, pp. 43–60, 2007.     

Goodness of fit of an assigned set of scores for the analysis of association in a contingency table

The problem of association between two attributes in ap×q contingency table can be looked upon as the problem of relationship between two vector variablesx andy. If there is only one true non-zero canonical correlation betweenx andy, the association between the two attributes is of rank 1 and in this case, one set of scores is adequate to describe the association completely; these scores are nothing but the coefficients in the canonical variates corresponding to the true non-zero canonical correlation. Given a set of hypothetical scoresα ₁,α₁,⋯,α_p for the rows, one is interested in testing their goodness of fit. Tests for this are suggested in this paper. For obtaining these tests, a preliminary result about direction and collinearity factors in discriminant analysis, whenS irrelevant variables are eliminated, is needed. This is derived in part one of this paper. This research was sponsored by the Office of Naval Research, Contract No. N00014-68-A-0515, Project No. NR042-260. Reproduction in whole or in part is permitted for any purpose of the United States Government.     

On the continuum limit of a discrete inverse spectral problem on optimal finite difference grids

We consider finite difference approximations of solutions of inverse Sturm‐Liouville problems in bounded intervals. Using three‐point finite difference schemes, we discretize the equations on so‐called optimal grids constructed as follows: For a staggered grid with 2 k points, we ask that the finite difference operator (a k × k Jacobi matrix) and the Sturm‐Liouville differential operator share the k lowest eigenvalues and the values of the orthonormal eigenfunctions at one end of the interval. This requirement determines uniquely the entries in the Jacobi matrix, which are grid cell averages of the coefficients in the continuum problem. If these coefficients are known, we can find the grid, which we call optimal because it gives, by design, a finite difference operator with a prescribed spectral measure. We focus attention on the inverse problem, where neither the coefficients nor the grid are known. A key question in inversion is how to parametrize the coefficients, i.e., how to choose the grid. It is clear that, to be successful, this grid must be close to the optimal one, which is unknown. Fortunately, as we show here, the grid dependence on the unknown coefficients is weak, so the inversion can be done on a precomputed grid for an a priori guess of the unknown coefficients. This observation leads to a simple yet efficient inversion algorithm, which gives coefficients that converge pointwise to the true solution as the number k of data points tends to infinity. The cornerstone of our convergence proof is showing that optimal grids provide an implicit, natural regularization of the inverse problem, by giving reconstructions with uniformly bounded total variation. The analysis is based on a novel, explicit perturbation analysis of Lanczos recursions and on a discrete Gel'fand‐Levitan formulation. © 2005 Wiley Periodicals, Inc.     

The Cauchy Problem for an Axially Symmetric Equation and the Schwarz Potential Conjecture for the Torus

We present our results in this paper in two parts. In the first part, we consider the Cauchy problem for the axially symmetric equation with entire Cauchy data given on an initial plane (see Eq. (2.1)). We solve the Cauchy problem and obtain its solutions in two cases, depending on whether k is a positive even integer or k is a positive odd integer. For k odd, we demonstrate that the solution has more singularities due to the propagation of the singularities of the coefficients. In the second part, the Cauchy problem for the same equation is considered, but instead, its entire Cauchy data are given on an initial sphere (see Eq. (3.1)). Whenever k is a positive even integer, we obtain the global existence of the solution and determine all possible singularities. Whenever k is a positive odd integer, we discuss both local and global solutions. As a consequence of our results in this paper, we show that the Schwarz Potential Conjecture (see the Introduction) for the even dimensional torus is true.     

Isomorphism Classes of Algebras with Radical Cube Zero II

We study to classify, up to isomorphism, algebras Λ over a field k such that the radical cubed is zero and Λ modulo the radical is a product of copies of k. The number of local quasi-Frobenius k-algebras with the condition is shown to be not less than the cardinality of k. In particular, the canonical forms of those algebras of dimension 5 are presented and their isomorphism classes are completely determined under some conditions on k.      

Nonparametric Bayesian Modeling for Multivariate Ordinal Data

This article proposes a probability model for k-dimensional ordinal outcomes, that is, it considers inference for data recorded in k-dimensional contingency tables with ordinal factors. The proposed approach is based on full posterior inference, assuming a flexible underlying prior probability model for the contingency table cell probabilities. We use a variation of the traditional multivariate probit model, with latent scores that determine the observed data. In our model, a mixture of normals prior replaces the usual single multivariate normal model for the latent variables. By augmenting the prior model to a mixture of normals we generalize inference in two important ways. First, we allow for varying local dependence structure across the contingency table. Second, inference in ordinal multivariate probit models is plagued by problems related to the choice and resampling of cutoffs defined for these latent variables. We show how the proposed mixture model approach entirely removes these problems. We illustrate the methodology with two examples, one simulated dataset and one dataset of interrater agreement.     

Influence diagnostics in the varying coefficient model with longitudinal data

One or few observations can be highly influential on estimates of regression coefficients in the linear regression model. In this paper we derive influence diagnostics for the varying coefficients model with longitudinal data. We note that diagnostics in this context is quite different from the classical regression model in the sense that regression coefficients vary as time varies. A version of Cook’s distance is suggested to reflect this specific aspect of varying coefficient model. An algorithm to present some guidelines to determine influential observations deserving special attention is developed. An illustrative example based on the AIDS data is also given.     

Canonical correlation analysis based on information theory

In this article, we propose a new canonical correlation method based on information theory. This method examines potential nonlinear relationships between p×1 vector Y-set and q×1 vector X-set. It finds canonical coefficient vectors a and b by maximizing a more general measure, the mutual information, between a^TX and b^TY. We use a permutation test to determine the pairs of the new canonical correlation variates, which requires no specific distributions for X and Y as long as one can estimate the densities of a^TX and b^TY nonparametrically. Examples illustrating the new method are presented.     

The likelihood ratio tests for the dimensionality of regression coefficients

In this paper we give a unified derivation of the likelihood ratio (LR) statistics for testing the hypothesis on the dimensionality of regression coefficients under a usual MANOVA model. We also derive the LR statistics under a general MANOVA model and study their asymptotic null and nonnull distributions. Further it is shown that the test statistic used by Bartlett [4] for testing the hypothesis that the last p?k canonical correlations are all zero is the LR statistic.     

非线性测量误差模型的影响分析

本文对非线性测量误差模型给出了统一的诊断方法,并证明了数据删除模型与均值漂移模型的等价性,由此出发得到了Cook距离、残差、杠杆值等诊断统计量.本文还讨论了非线性测量误差模型的局部影响分析,并给出了一个具体应用实例.推广了Zhao & Lee(1995)的结果.     

Ultraconvergence of the derivative of high-order finite element method for elliptic problems with constant coefficients

In this article, for second order elliptic problems with constant coefficients, the local ultraconvergence of the derivative of finite element method using piecewise polynomials of degrees k (k ≥ 2) is studied by the interpolation postprocessing technique. Under suitable regularity and mesh conditions, we prove that at an interior vertex, which is away from the boundary with a fixed distance, the gradient of the postprecessed finite element solution using piecewise polynomials of degrees k (k ≥ 2) converges to the gradient of the exact solution with order . Numerical experiments are used to illustrate our theoretical findings.     

A geometric characterisation of linear k-blocking sets

In this paper we prove that any linear k-blocking set is either a canonical subgeometry or a projection of some canonical subgeometry. Received 6 February 2001.     

The local ultraconvergence of high‐order finite element method for second‐order elliptic problems with constant coefficients over a rectangular partition

In this article, we will discuss the local ultraconvergence of high‐degree finite element method based on a rectangular partition for the second‐degree elliptic problem with constant coefficients in Ω ? ?² , u( y ) = 0 on ?Ω . Based on suitable regularity, ultraconvergence of the displacement of the extrapolated kth (k ≥ 3) degree finite element solution has been obtained by an extrapolation technique. Finally, numerical experiments are applied to demonstrate our theoretical findings.