arXiv subjects

Andrew B. Nobel

Publications and source records attributed to Andrew B. Nobel.

28 records · Page 2Linked to original sources

arXiv2010

Uniform Approximation of Vapnik-Chervonenkis Classes

For any family of measurable sets in a probability space, we show that either (i) the family has infinite Vapnik-Chervonenkis (VC) dimension or (ii) for every epsilon > 0 there is a finite partition pi such the pi-boundary of each set has measure at most epsilon. Immediate corollaries include the fact that a family with finite VC dimension has finite bracketing numbers, and satisfies uniform laws of large numbers for every ergodic process. From these corollaries, we derive analogous results for VC major and VC graph families of functions.

math.PR↗

arXiv2010

Uniform convergence of Vapnik--Chervonenkis classes under ergodic sampling

We show that if $\mathcal{X}$ is a complete separable metric space and $\mathcal{C}$ is a countable family of Borel subsets of $\mathcal{X}$ with finite VC dimension, then, for every stationary ergodic process with values in $\mathcal{X}$, the relative frequencies of sets $C\in\mathcal{C}$ converge uniformly to their limiting probabilities. Beyond ergodicity, no assumptions are imposed on the sampling process, and no regularity conditions are imposed on the elements of $\mathcal{C}$. The result extends existing work of Vapnik and Chervonenkis, among others, who have studied uniform convergence for i.i.d. and strongly mixing processes. Our method of proof is new and direct: it does not rely on symmetrization techniques, probability inequalities or mixing conditions. The uniform convergence of relative frequencies for VC-major and VC-graph classes of functions under ergodic sampling is established as a corollary of the basic result for sets.

math.PR↗

arXiv2010

On the maximal size of Large-Average and ANOVA-fit Submatrices in a Gaussian Random Matrix

We investigate the maximal size of distinguished submatrices of a Gaussian random matrix. Of interest are submatrices whose entries have average greater than or equal to a positive constant, and submatrices whose entries are well-fit by a two-way ANOVA model. We identify size thresholds and associated (asymptotic) probability bounds for both large-average and ANOVA-fit submatrices. Results are obtained when the matrix and submatrices of interest are square, and in rectangular cases when the matrix submatrices of interest have fixed aspect ratios. In addition, we obtain a strong, interval concentration result for the size of large average submatrices in the square case. A simulation study shows good agreement between the observed and predicted sizes of large average submatrices in matrices of moderate size.

math.ST↗

arXiv2010

Uniform Approximation and Bracketing Properties of VC classes

We show that the sets in a family with finite VC dimension can be uniformly approximated within a given error by a finite partition. Immediate corollaries include the fact that VC classes have finite bracketing numbers, satisfy uniform laws of averages under strong dependence, and exhibit uniform mixing. Our results are based on recent work concerning uniform laws of averages for VC classes under ergodic sampling.

math.PR↗

arXiv2010

The Gap Dimension and Uniform Laws of Large Numbers for Ergodic Processes

Let F be a family of Borel measurable functions on a complete separable metric space. The gap (or fat-shattering) dimension of F is a combinatorial quantity that measures the extent to which functions f in F can separate finite sets of points at a predefined resolution gamma > 0. We establish a connection between the gap dimension of F and the uniform convergence of its sample averages under ergodic sampling. In particular, we show that if the gap dimension of F at resolution gamma > 0 is finite, then for every ergodic process the sample averages of functions in F are eventually within 10 gamma of their limiting expectations uniformly over the class F. If the gap dimension of F is finite for every resolution gamma > 0 then the sample averages of functions in F converge uniformly to their limiting expectations. We assume only that F is uniformly bounded and countable (or countably approximable). No smoothness conditions are placed on F, and no assumptions beyond ergodicity are placed on the sampling processes. Our results extend existing work for i.i.d. processes.

math.PR↗

arXiv2009

Finding large average submatrices in high dimensional data

The search for sample-variable associations is an important problem in the exploratory analysis of high dimensional data. Biclustering methods search for sample-variable associations in the form of distinguished submatrices of the data matrix. (The rows and columns of a submatrix need not be contiguous.) In this paper we propose and evaluate a statistically motivated biclustering procedure (LAS) that finds large average submatrices within a given real-valued data matrix. The procedure operates in an iterative-residual fashion, and is driven by a Bonferroni-based significance score that effectively trades off between submatrix size and average value. We examine the performance and potential utility of LAS, and compare it with a number of existing methods, through an extensive three-part validation study using two gene expression datasets. The validation study examines quantitative properties of biclusters, biological and clinical assessments using auxiliary information, and classification of disease subtypes using bicluster membership. In addition, we carry out a simulation study to assess the effectiveness and noise sensitivity of the LAS search procedure. These results suggest that LAS is an effective exploratory tool for the discovery of biologically relevant structures in high dimensional data. Software is available at https://genome.unc.edu/las/.

q-bio.GN↗

arXiv2008

A statistical framework for testing functional categories in microarray data

Ready access to emerging databases of gene annotation and functional pathways has shifted assessments of differential expression in DNA microarray studies from single genes to groups of genes with shared biological function. This paper takes a critical look at existing methods for assessing the differential expression of a group of genes (functional category), and provides some suggestions for improved performance. We begin by presenting a general framework, in which the set of genes in a functional category is compared to the complementary set of genes on the array. The framework includes tests for overrepresentation of a category within a list of significant genes, and methods that consider continuous measures of differential expression. Existing tests are divided into two classes. Class 1 tests assume gene-specific measures of differential expression are independent, despite overwhelming evidence of positive correlation. Analytic and simulated results are presented that demonstrate Class 1 tests are strongly anti-conservative in practice. Class 2 tests account for gene correlation, typically through array permutation that by construction has proper Type I error control for the induced null. However, both Class 1 and Class 2 tests use a null hypothesis that all genes have the same degree of differential expression. We introduce a more sensible and general (Class 3) null under which the profile of differential expression is the same within the category and complement. Under this broader null, Class 2 tests are shown to be conservative. We propose standard bootstrap methods for testing against the Class 3 null and demonstrate they provide valid Type I error control and more power than array permutation in simulated datasets and real microarray experiments.

stat.AP↗

arXiv2007

Regression estimation from an individual stable sequence

We consider univariate regression estimation from an individual (non-random) sequence $(x_1,y_1),(x_2,y_2), ... \in \real \times \real$, which is stable in the sense that for each interval $A \subseteq \real$, (i) the limiting relative frequency of $A$ under $x_1, x_2, ...$ is governed by an unknown probability distribution $μ$, and (ii) the limiting average of those $y_i$ with $x_i \in A$ is governed by an unknown regression function $m(\cdot)$. A computationally simple scheme for estimating $m(\cdot)$ is exhibited, and is shown to be $L_2$ consistent for stable sequences $\{(x_i,y_i)\}$ such that $\{y_i\}$ is bounded and there is a known upper bound for the variation of $m(\cdot)$ on intervals of the form $(-i,i]$, $i \geq 1$. Complementing this positive result, it is shown that there is no consistent estimation scheme for the family of stable sequences whose regression functions have finite variation, even under the restriction that $x_i \in [0,1]$ and $y_i$ is binary-valued.

math.PR↗

arXiv2007

Density estimation from an individual numerical sequence

This paper considers estimation of a univariate density from an individual numerical sequence. It is assumed that (i) the limiting relative frequencies of the numerical sequence are governed by an unknown density, and (ii) there is a known upper bound for the variation of the density on an increasing sequence of intervals. A simple estimation scheme is proposed, and is shown to be $L_1$ consistent when (i) and (ii) apply. In addition it is shown that there is no consistent estimation scheme for the set of individual sequences satisfying only condition (i).

math.PR↗

arXiv2006

Denoising Deterministic Time Series

This paper is concerned with the problem of recovering a finite, deterministic time series from observations that are corrupted by additive, independent noise. A distinctive feature of this problem is that the available data exhibit long-range dependence and, as a consequence, existing statistical theory and methods are not readily applicable. This paper gives an analysis of the denoising problem that extends recent work of Lalley, but begins from first principles. Both positive and negative results are established. The positive results show that denoising is possible under somewhat restrictive conditions on the additive noise. The negative results show that, under more general conditions on the noise, no procedure can recover the underlying deterministic series.

nlin.CD↗