Hypothesis tests for high-dimensional covariance structures

We consider hypothesis testing for high-dimensional covariance structures in which the covariance matrix is a (i) scaled identity matrix, (ii) diagonal matrix, or (iii) intraclass covariance matrix. Our purpose is to systematically establish a nonparametric approach for testing the high-dimensional covariance structures (i)–(iii). We produce a new common test statistic for each covariance structure and show that the test statistic is an unbiased estimator of its corresponding test parameter. We prove that the test statistic establishes the asymptotic normality. We propose a new test procedure for (i)–(iii) and evaluate its asymptotic size and power theoretically when both the dimension and sample size increase. We investigate the performance of the proposed test procedure in simulations. As an application of testing the covariance structures, we give a test procedure to identify an eigenvector. Finally, we demonstrate the proposed test procedure by using a microarray data set.

     

Interactive High-Dimensional Data Visualization

Abstract

We propose a rudimentary taxonomy of interactive data visualization based on a triad of data analytic tasks: finding Gestalt, posing queries, and making comparisons. These tasks are supported by three classes of interactive view manipulations: focusing, linking, and arranging views. This discussion extends earlier work on the principles of focusing and linking and sets them on a firmer base. Next, we give a high-level introduction to a particular system for multivariate data visualization—XGobi. This introduction is not comprehensive but emphasizes XGobi tools that are examples of focusing, linking, and arranging views; namely, high-dimensional projections, linked scatterplot brushing, and matrices of conditional plots. Finally, in a series of case studies in data visualization, we show the powers and limitations of particular focusing, linking, and arranging tools. The discussion is dominated by high-dimensional projections that form an extremely well-developed part of XGobi. Of particular interest are the illustration of asymptotic normality of high-dimensional projections (a theorem of Diaconis and Freedman), the use of high-dimensional cubes for visualizing factorial experiments, and a method for interactively generating matrices of conditional plots with high-dimensional projections. Although there is a unifying theme to this article, each section—in particular the case studies—can be read separately.     

A graphical method for assessing multivariate normality

Summary  In this paper we suggest a simple graphical device for assessing multivariate normality. The method is based on the characteristic that linear combinations of the sample mean and sample covariance matrix are independent if and only if the random variable is normally distributed. We demonstrate the usage of the suggested method and compare it to the classical Q-Q plot by using some multivariate data sets.     

Interpolation Scheme for Planar Cubic <Emphasis Type="Italic">G</Emphasis><Superscript>2</Superscript> Spline Curves

In this paper a method for interpolating planar data points by cubic G ² splines is presented. A spline is composed of polynomial segments that interpolate two data points, tangent directions and curvatures at these points. Necessary and sufficient, purely geometric conditions for the existence of such a polynomial interpolant are derived. The obtained results are extended to the case when the derivative directions and curvatures are not prescribed as data, but are obtained by some local approximation or implied by shape requirements. As a result, the G ² spline is constructed entirely locally.     

Pattern Search Method for Discrete <Emphasis Type="Italic">L</Emphasis><Subscript>1</Subscript>–Approximation

We propose a pattern search method to solve a classical nonsmooth optimization problem. In a deep analogy with pattern search methods for linear constrained optimization, the set of search directions at each iteration is defined in such a way that it conforms to the local geometry of the set of points of nondifferentiability near the current iterate. This is crucial to ensure convergence. The approach presented here can be extended to wider classes of nonsmooth optimization problems. Numerical experiments seem to be encouraging. This work was supported by M.U.R.S.T., Rome, Italy.     

Rejoinder

Royston proposed a normal probability plot to detect nonnormality of univariate data. The normal probability plot was provided with normalized acceptance regions to enhance its interpretability. By using the theory of spherical distributions and the idea of principal component analysis, we propose an approach to extending Royston’s normal plot to detecting nonmultivariate normality in analyzing high-dimensional data. The performance of the proposed multivariate normal plot is demonstrated by Monte Carlo studies and illustrated by two real datasets.

Datasets, computer code and documentation of the code are available in the online supplements.     

Symmetry properties of Cayley graphs of small valencies on the alternating group A<Subscript>5</Subscript>

The normality of symmetry property of Cayley graphs of valencies 3 and 4 on the alternating group A₅ is studied. We prove that all but four such graphs are normal; that A₅ is not 5-CI. A complete classification of all arc-transitive Cayley graphs on A₅ of valencies 3 and 4 as well as some examples of trivalent and tetravalent GRRs of A₅ is given.     

Dimensionality Reductions in ℓ<Subscript>2</Subscript> that Preserve Volumes and Distance to Affine Spaces

Let X be a subset of n points of the Euclidean space, and let 0 < ε < 1. A classical result of Johnson and Lindenstrauss [JL] states that there is a projection of X onto a subspace of dimension O(ε^-2 log n) with distortion ≤ 1+ ε. We show a natural extension of the above result to a stronger preservation of the geometry of finite spaces. By a k-fold increase of the number of dimensions used compared with [JL], a good preservation of volumes and of distances between points and affine spaces is achieved. Specifically, we show how to embed a subset of size n of the Euclidean space into a O(ε^-2 log n)-dimensional Euclidean space, so that no set of size s ≤ k changes its volume by more than (1 + ε^s-1. Moreover, distances of points from affine hulls of sets of at most k - 1 points in the space do not change by more than a factor of 1 + ε. A consequence of the above with k = 3 is that angles can be preserved using asymptotically the same number of dimensions as the one used in [JL]. Our method can be applied to many problems with high-dimensional nature such as Projective Clustering and Approximated Nearest Affine Neighbour Search. In particular, it shows a first polylogarithmic query time approximation algorithm to the latter. We also show a structural application that for volume respecting embedding in the sense introduced by Feige [F], the host space need not generally be of dimensionality greater than polylogarithmic in the size of the graph.     

Displaying Variation in Large Datasets: Plotting a Visual Summary of Effect Sizes

Displaying the component-wise between-group differences high-dimensional datasets is problematic because widely used plots such as Bland–Altman and Volcano plots do not show what they are colloquially believed to show. Thus, it is difficult for the experimentalist to grasp why the between-group difference of one component is “significant” while that of another component is not. Here, we propose a type of “Effect Plot” that displays between-group differences in relation to respective underlying variability for every component of a high-dimensional dataset. We use synthetic data to show that such a plot captures the essence of what determines “significance” for between-group differences in each component, and provide guidance in the interpretation of the plot. Supplementary online materials contain the code and data for this article and include simple R functions to produce an effect plot from suitable datasets.     

Bijections preserving invertibility of differences of matrices on <Emphasis Type="Italic">H</Emphasis><Subscript><Emphasis Type="Italic">n</Emphasis></Subscript>

We characterise bijections on the space of hermitian matrices preserving the invertibility of differences of matrix pairs in both directions.     

A fuzzy bootstrap test for the mean with <Emphasis Type="Italic">D</Emphasis><Subscript><Emphasis Type="Italic">p,q</Emphasis></Subscript>-distance

In this paper, we consider the problem of testing a simple hypothesis about the mean of a fuzzy random variable. For this purpose, we take a distance between the sample mean and the mean in the null hypothesis as a test statistic. An asymptotic test about the fuzzy mean is obtained by using a central limit theorem. The asymptotical distribution is ω ₂-distribution. The ω ₂-distribution is only known for special cases, thus we have considered random LR-fuzzy numbers. In the fuzzy concept, in addition to the existence of several versions of the central limit theorem, there is another practical disadvantage: The limit law is, in most cases, difficult to handle. Therefore, the central limit theorem for fuzzy random variable does not seem to be a very useful tool to make inferences on the mean of fuzzy random variable. Thus we use the bootstrap technique. Finally, by means of a simulation study, we show that the bootstrap method is a powerful tool in the statistical hypothesis testing about the mean of fuzzy random variables.     

An elementary proof of the <Emphasis Type="Italic">A</Emphasis><Subscript>2</Subscript> bound

A martingale transform T, applied to an integrable locally supported function f, is pointwise dominated by a positive sparse operator applied to |f|, the choice of sparse operator being a function of T and f. As a corollary, one derives the sharp A _p bounds for martingale transforms, recently proved by Thiele-Treil-Volberg, as well as a number of new sharp weighted inequalities for martingale transforms. The (very easy) method of proof (a) only depends upon the weak-L ¹ norm of maximal truncations of martingale transforms, (b) applies in the vector valued setting, and (c) has an extension to the continuous case, giving a new elementary proof of the A ² bounds in that setting.     

A new functor from <Emphasis Type="Italic">D</Emphasis><Subscript>5</Subscript>-Mod to <Emphasis Type="Italic">E</Emphasis><Subscript>6</Subscript>-Mod

We find a new representation of the simple Lie algebra of type E ₆ on the polynomial algebra in 16 variables, which gives a fractional representation of the corresponding Lie group on 16-dimensional space. Using this representation and Shen’s idea of mixed product, we construct a new functor from D ₅-Mod to E ₆-Mod. A condition for the functor to map a finite-dimensional irreducible D ₅-module to an infinite-dimensional irreducible E ₆-module is obtained. Our results yield explicit constructions of certain infinite-dimensional irreducible weight E6-modules with finite-dimensional weight subspaces. In our approach, the idea of Kostant’s characteristic identities plays a key role.     

The adaptive lasso in high-dimensional sparse heteroscedastic models

In this paper we study the asymptotic properties of the adaptive Lasso estimate in high-dimensional sparse linear regression models with heteroscedastic errors. It is demonstrated that model selection properties and asymptotic normality of the selected parameters remain valid but with a suboptimal asymptotic variance. A weighted adaptive Lasso estimate is introduced and investigated. In particular, it is shown that the new estimate performs consistent model selection and that linear combinations of the estimates corresponding to the non-vanishing components are asymptotically normally distributed with a smaller variance than those obtained by the “classical” adaptive Lasso. The results are illustrated in a data example and by means of a small simulation study.     

<Emphasis Type="Italic">l</Emphasis><Subscript>1</Subscript>-<Emphasis Type="Italic">l</Emphasis><Subscript>2</Subscript> regularization of split feasibility problems

Numerous problems in signal processing and imaging, statistical learning and data mining, or computer vision can be formulated as optimization problems which consist in minimizing a sum of convex functions, not necessarily differentiable, possibly composed with linear operators and that in turn can be transformed to split feasibility problems (SFP); see for example Censor and Elfving (Numer. Algorithms 8, 221–239 1994). Each function is typically either a data fidelity term or a regularization term enforcing some properties on the solution; see for example Chaux et al. (SIAM J. Imag. Sci. 2, 730–762 2009) and references therein. In this paper, we are interested in split feasibility problems which can be seen as a general form of Q-Lasso introduced in Alghamdi et al. (2013) that extended the well-known Lasso of Tibshirani (J. R. Stat. Soc. Ser. B 58, 267–288 1996). Q is a closed convex subset of a Euclidean m-space, for some integer m ≥ 1, that can be interpreted as the set of errors within given tolerance level when linear measurements are taken to recover a signal/image via the Lasso. Inspired by recent works by Lou and Yan (2016), Xu (IEEE Trans. Neural Netw. Learn. Syst. 23, 1013–1027 2012), we are interested in a nonconvex regularization of SFP and propose three split algorithms for solving this general case. The first one is based on the DC (difference of convex) algorithm (DCA) introduced by Pham Dinh Tao, the second one is nothing else than the celebrate forward-backward algorithm, and the third one uses a method introduced by Mine and Fukushima. It is worth mentioning that the SFP model a number of applied problems arising from signal/image processing and specially optimization problems for intensity-modulated radiation therapy (IMRT) treatment planning; see for example Censor et al. (Phys. Med. Biol. 51, 2353–2365, 2006).     

Path-following interior point algorithms for the Cartesian <Emphasis Type="Italic">P</Emphasis><Subscript>*</Subscript>(κ)-LCP over symmetric cones

In this paper, we establish a theoretical framework of path-following interior point algorithms for the linear complementarity problems over symmetric cones (SCLCP) with the Cartesian P _*(κ)-property, a weaker condition than the monotonicity. Based on the Nesterov-Todd, xy and yx directions employed as commutative search directions for semidefinite programming, we extend the variants of the short-, semilong-, and long-step path-following algorithms for symmetric conic linear programming proposed by Schmieta and Alizadeh to the Cartesian P _*(κ)-SCLCP, and particularly show the global convergence and the iteration complexities of the proposed algorithms. This work was supported by National Natural Science Foundation of China (Grant Nos. 10671010, 70841008)     

A New Inequality for Weakly （K1, K2）-Quasiregular Mappings

We obtain a new inequality for weakly （K1,K2）-quasiregular mappings by using the McShane extension method. This inequality can be used to derive the self-improving regularity of （K1, K2）-Quasiregular Mappings.     

A new non-interior continuation method for <Emphasis Type="Italic">P</Emphasis><Subscript>0</Subscript>-NCP based on a SSPM-function

In this paper, we consider a new non-interior continuation method for the solution of nonlinear complementarity problem with P ₀-function (P ₀-NCP). The proposed algorithm is based on a smoothing symmetric perturbed minimum function (SSPM-function), and one only needs to solve one system of linear equations and to perform only one Armijo-type line search at each iteration. The method is proved to possess global and local convergence under weaker conditions. Preliminary numerical results indicate that the algorithm is effective.     

Minimum <Emphasis Type="Italic">Φ</Emphasis>-Divergence Estimator and <Emphasis Type="Italic">Φ</Emphasis>-Divergence Statistics in Generalized Linear Models with Binary Data

In this paper, we assume that the data are distributed according to a binomial distribution whose probabilities follow a generalized linear model. To fit the data the minimum φ-divergence estimator is studied as a generalization of the maximum likelihood estimator. We use the minimum φ-divergence estimator, which is the basis of some new statistics, for solving the problems of testing in a generalized linear model with binary data. A wide simulation study is carried out for studying the behavior of the new family of estimators as well as of the new family of test statistics. This work was partially supported by Grant MTM2006-06872 and UCM2006-910707.