Showing 1–13 of 13 results for author: Afendras, G

An AUK-based index for measuring and testing the joint dependence of a random vector

Authors: Georgios Afendras, Marianthi Markatou, Albert Vexler

Abstract: We present an index of dependence that allows one to measure the joint or mutual dependence of a $d$-dimensional random vector with $d>2$. The index is based on a $d$-dimensional Kendall process. We further propose a standardized version of our index of dependence that is easy to interpret, and provide an algorithm for its computation. We discuss tests of total independence based on consistent est… ▽ More We present an index of dependence that allows one to measure the joint or mutual dependence of a $d$-dimensional random vector with $d>2$. The index is based on a $d$-dimensional Kendall process. We further propose a standardized version of our index of dependence that is easy to interpret, and provide an algorithm for its computation. We discuss tests of total independence based on consistent estimates of the area under the Kendall curve. We evaluate the performance of our procedures via simulation, and apply our methods to a real data set. △ Less

Submitted 23 December, 2020; v1 submitted 24 November, 2020; originally announced November 2020.

Comments: 33 pages (plus 8 pages supplementary material), 7 figures, 9 tables

MSC Class: Primary 62H20; 62H05; 62E10; secondary 62-09; 62G99

Multi-Panel Kendall Plot in Light of an ROC Curve Analysis Applied to Measuring Dependence

Authors: Albert Vexler, Georgios Afendras, Marianthi Markatou

Abstract: The Kendall plot ($\K$-plot) is a plot measuring dependence between the components of a bivariate random variable. The $\K$-plot graphs the Kendall distribution function against the distribution function of $VU$, where $V$ and $U$ are independent uniform $[0,1]$ random variables. We associate $\K$-plots with the receiver operating characteristic ($\ROC$) curve, a well-accepted graphical tool in bi… ▽ More The Kendall plot ($\K$-plot) is a plot measuring dependence between the components of a bivariate random variable. The $\K$-plot graphs the Kendall distribution function against the distribution function of $VU$, where $V$ and $U$ are independent uniform $[0,1]$ random variables. We associate $\K$-plots with the receiver operating characteristic ($\ROC$) curve, a well-accepted graphical tool in biostatistics for evaluating the ability of a biomarker to discriminate between two populations. The most commonly used global index of diagnostic accuracy of biomarkers is the area under the $\ROC$ curve ($\AUC$). In parallel with the $\AUC$, we propose a novel strategy to measure the association between random variables from a continuous bivariate distribution. First, we discuss why the area under the conventional Kendall curve ($\AUK$) cannot be used as an index of dependence. We then suggest a simple and meaningful extension of the definition of the $\K$-plots and define an index of dependence that is based on $\AUK$. This measure characterizes a wide range of two-variable relationships, thereby completely detecting the underlying dependence structure. Properties of the proposed index satisfy the mathematical definition of a measure. Finally, simulations and real data examples illustrate the applicability of the proposed method. △ Less

Submitted 21 November, 2018; originally announced November 2018.

Comments: Statistics: A Journal of Theoretical and Applied Statistics. Accepted

On the limiting distribution of sample central moments

Authors: Georgios Afendras, Nickos Papadatos, Violetta Piperigou

Abstract: We investigate the limiting behavior of sample central moments, examining the special cases where the limiting (as the sample size tends to infinity) distribution is degenerate. Parent (non-degenerate) distributions with this property are called \emph{singular}, and we show in this article that the singular distributions contain at most three supporting points. Moreover, using the \emph{delta}-met… ▽ More We investigate the limiting behavior of sample central moments, examining the special cases where the limiting (as the sample size tends to infinity) distribution is degenerate. Parent (non-degenerate) distributions with this property are called \emph{singular}, and we show in this article that the singular distributions contain at most three supporting points. Moreover, using the \emph{delta}-method, we show that the (second order) limiting distribution of sample central moments from a singular distribution is either a multiple, or a difference of two multiples of independent chi-square random variables with one degree of freedom. Finally, we present a new characterization of normality through the asymptotic independence of the sample mean and all sample central moments. △ Less

Submitted 6 June, 2018; originally announced June 2018.

Comments: 26 pages, 1 figure

The out-of-source error in multi-source cross validation-type procedures

Authors: Georgios Afendras, Marianthi Markatou

Abstract: A scientific phenomenon under study may often be manifested by data arising from processes, i.e. sources, that may describe this phenomenon. In this contex of multi-source data, we define the "out-of-source" error, that is the error committed when a new observation of unknown source origin is allocated to one of the sources using a rule that is trained on the known labeled data. We present an unbi… ▽ More A scientific phenomenon under study may often be manifested by data arising from processes, i.e. sources, that may describe this phenomenon. In this contex of multi-source data, we define the "out-of-source" error, that is the error committed when a new observation of unknown source origin is allocated to one of the sources using a rule that is trained on the known labeled data. We present an unbiased estimator of this error, and discuss its variance. We derive natural and easily verifiable assumptions under which the consistency of our estimator is guaranteed for a broad class of loss functions and data distributions. Finally, we evaluate our theoretical results via a simulation study. △ Less

Submitted 22 December, 2016; originally announced December 2016.

Comments: 16 pages, 4 tables

Journal ref: New Advances in Statistics and Data Science 2017, 27-44

Statistical Distances and Their Role in Robustness

Authors: Marianthi Markatou, Yang Chen, Georgios Afendras, Bruce G. Lindsay

Abstract: Statistical distances, divergences, and similar quantities have a large history and play a fundamental role in statistics, machine learning and associated scientific disciplines. However, within the statistical literature, this extensive role has too often been played out behind the scenes, with other aspects of the statistical problems being viewed as more central, more interesting, or more impor… ▽ More Statistical distances, divergences, and similar quantities have a large history and play a fundamental role in statistics, machine learning and associated scientific disciplines. However, within the statistical literature, this extensive role has too often been played out behind the scenes, with other aspects of the statistical problems being viewed as more central, more interesting, or more important. The behind the scenes role of statistical distances shows up in estimation, where we often use estimators based on minimizing a distance, explicitly or implicitly, but rarely studying how the properties of a distance determine the properties of the estimators. Distances are also prominent in goodness-of-fit, but the usual question we ask is "how powerful is this method against a set of interesting alternatives" not "what aspect of the distance between the hypothetical model and the alternative are we measuring?" Our focus is on describing the statistical properties of some of the distance measures we have found to be most important and most visible. We illustrate the robust nature of Neyman's chi-squared and the non-robust nature of Pearson's chi-squared statistics and discuss the concept of discretization robustness. △ Less

Submitted 21 December, 2016; originally announced December 2016.

Comments: 23 pages

Journal ref: New Advances in Statistics and Data Science 2017, 3-26

Optimality of Training/Test Size and Resampling Effectiveness of Cross-Validation Estimators of the Generalization Error

Authors: Georgios Afendras, Marianthi Markatou

Abstract: An important question in constructing Cross Validation (CV) estimators of the generalization error is whether rules can be established that allow "optimal" selection of the size of the training set, for fixed sample size $n$. We define the {\it resampling effectiveness} of random CV estimators of the generalization error as the ratio of the limiting value of the variance of the CV estimator over t… ▽ More An important question in constructing Cross Validation (CV) estimators of the generalization error is whether rules can be established that allow "optimal" selection of the size of the training set, for fixed sample size $n$. We define the {\it resampling effectiveness} of random CV estimators of the generalization error as the ratio of the limiting value of the variance of the CV estimator over the estimated from the data variance. The variance and the covariance of different average test set errors are independent of their indices, thus, the resampling effectiveness depends on the correlation and the number of repetitions used in the random CV estimator. We discuss statistical rules to define optimality and obtain the "optimal" training sample size as the solution of an appropriately formulated optimization problem. We show that in a broad class of loss functions the optimal training size equals half of the total sample size, independently of the data distribution. We optimally select the number of folds in $k$-fold cross validation and offer a computational procedure for obtaining the optimal splitting in the case of classification (via logistic regression). We substantiate our claims both, theoretically and empirically. △ Less

Submitted 9 November, 2015; originally announced November 2015.

Comments: 53 pages, 6 figures, 16 tables

Uniform Integrability of the OLS Estimators, and the Convergence of their Moments

Authors: Georgios Afendras, Marianthi Markatou

Abstract: The problem of convergence of moments of a sequence of random variables to the moments of its asymptotic distribution is important in many applications. These include the determination of the optimal training sample size in the cross validation estimation of the generalization error of computer algorithms, and in the construction of graphical methods for studying dependence patterns between two bi… ▽ More The problem of convergence of moments of a sequence of random variables to the moments of its asymptotic distribution is important in many applications. These include the determination of the optimal training sample size in the cross validation estimation of the generalization error of computer algorithms, and in the construction of graphical methods for studying dependence patterns between two biomarkers. In this paper we prove the uniform integrability of the ordinary least squares estimators of a linear regression model, under suitable assumptions on the design matrix and the moments of the errors. Further, we prove the convergence of the moments of the estimators to the corresponding moments of their asymptotic distribution, and study the rate of the moment convergence. The canonical central limit theorem corresponds to the simplest linear regression model. We investigate the rate of the moment convergence in canonical central limit theorem proving a sharp improvement of von Bahr's (1965) theorem. △ Less

Submitted 13 June, 2018; v1 submitted 9 November, 2015; originally announced November 2015.

Comments: 10 pages

MSC Class: 62J05; 62E20; 60E15; 60F05; 05A10

Journal ref: TEST 2016, Vol. 25, No 4, 775-784

A factorial moment distance and an application to the matching problem

Authors: G. Afendras, N. Papadatos

Abstract: In this note we introduce the notion of factorial moment distance for non-negative integer-valued random variables and we compare it with the total variation distance. Furthermore, we study the rate of convergence in the classical matching problem and in a generalized matching distribution. In this note we introduce the notion of factorial moment distance for non-negative integer-valued random variables and we compare it with the total variation distance. Furthermore, we study the rate of convergence in the classical matching problem and in a generalized matching distribution. △ Less

Submitted 5 November, 2014; originally announced November 2014.

Comments: 10 pages

MSC Class: Primary 60E05; 60E15; Secondary 44A10; 41A25

Journal ref: Theory of Probability and Its Applications 2017, Vol. 62, No 3, 617-628

Orthogonal polynomials in the Cumulative Ord family and its application to variance bounds

Authors: Giorgos Afendras, Narayanaswamy Balakrishnan, Nickos Papadatos

Abstract: This article presents and reviews several basic properties of the Cumulative Ord family of distributions; this family contains all the commonly used discrete distributions. A complete classification of the Ord family of probability mass functions is related to the orthogonality of the corresponding Rodrigues polynomials. Also, for any random variable $X$ of this family and for any suitable functio… ▽ More This article presents and reviews several basic properties of the Cumulative Ord family of distributions; this family contains all the commonly used discrete distributions. A complete classification of the Ord family of probability mass functions is related to the orthogonality of the corresponding Rodrigues polynomials. Also, for any random variable $X$ of this family and for any suitable function $g$ in $L^2(\mathbb{R},X)$, the article provides useful relationships between the Fourier coefficients of $g$ (with respect to the orthonormal polynomial system associated to $X$) and the Fourier coefficients of the forward difference of $g$ (with respect to another system of polynomials, orthonormal with respect to another distribution of the system). Finally, using these properties, a class of bounds for the variance of $g(X)$ is obtained, in terms of the forward differences of $g$. These bounds unify and improve several existing results. △ Less

Submitted 13 June, 2018; v1 submitted 8 August, 2014; originally announced August 2014.

Comments: 31 pages, 1 Table

MSC Class: Primary 60E05; 62E99; 05E35; 42A61; Secondary 60E15

Journal ref: Statistics 2018, Vol. 52, No 2, 364-392

A note on a variance bound for the multinomial and the negative multinomial distribution

Authors: G. Afendras, V. Papathanasiou

Abstract: We prove a Chernoff-type upper variance bound for the multinomial and the negative multinomial distribution. An application is also given. We prove a Chernoff-type upper variance bound for the multinomial and the negative multinomial distribution. An application is also given. △ Less

Submitted 12 June, 2018; v1 submitted 14 October, 2011; originally announced October 2011.

Comments: 8 pages

MSC Class: Primary 60E15

Journal ref: Naval Research Logistics 2014, Vol. 61, No 3, 179-183

Unified extension of variance bounds for integrated Pearson family

Authors: G. Afendras

Abstract: We use some properties of orthogonal polynomials to provide a class of upper/lower variance bounds for a function $g(X)$ of an absolutely continuous random variable $X$, in terms of the derivatives of $g$ up to some order. The new bounds are better than the existing ones. We use some properties of orthogonal polynomials to provide a class of upper/lower variance bounds for a function $g(X)$ of an absolutely continuous random variable $X$, in terms of the derivatives of $g$ up to some order. The new bounds are better than the existing ones. △ Less

Submitted 8 June, 2018; v1 submitted 1 October, 2011; originally announced October 2011.

Comments: 14 pages

MSC Class: 60E15

Journal ref: Annals of the Institute of Statistical Mathematics 2013, Vol. 65, No 4, 687-702

Moment-based inference for Pearson's quadratic q subfamily of distributions

Authors: Giorgos Afendras

Abstract: The author uses a Stein-type covariance identity to obtain moment estimators for the parameters of the quadratic polynomial subfamily of Pearson distributions. The asymptotic distribution of the estimators is obtained, and normality and symmetry tests based on it are provided. Simulation is used to compare the performance of the proposed tests with that of other existing tests for symmetry and nor… ▽ More The author uses a Stein-type covariance identity to obtain moment estimators for the parameters of the quadratic polynomial subfamily of Pearson distributions. The asymptotic distribution of the estimators is obtained, and normality and symmetry tests based on it are provided. Simulation is used to compare the performance of the proposed tests with that of other existing tests for symmetry and normality. △ Less

Submitted 5 April, 2011; v1 submitted 31 March, 2011; originally announced April 2011.

Comments: 21 pages, 14 figures, 10 tables, submitted for publication

MSC Class: 62E01

Journal ref: Communications in Statistics Theory and Methods 2013, Vol. 42, No 12, 1-10

An extended Stein-type covariance identity for the Pearson family with applications to lower variance bounds

Authors: G. Afendras, N. Papadatos, V. Papathanasiou

Abstract: For an absolutely continuous (integer-valued) r.v. $X$ of the Pearson (Ord) family, we show that, under natural moment conditions, a Stein-type covariance identity of order $k$ holds (cf. [Goldstein and Reinert, J. Theoret. Probab. 18 (2005) 237--260]). This identity is closely related to the corresponding sequence of orthogonal polynomials, obtained by a Rodrigues-type formula, and provides conve… ▽ More For an absolutely continuous (integer-valued) r.v. $X$ of the Pearson (Ord) family, we show that, under natural moment conditions, a Stein-type covariance identity of order $k$ holds (cf. [Goldstein and Reinert, J. Theoret. Probab. 18 (2005) 237--260]). This identity is closely related to the corresponding sequence of orthogonal polynomials, obtained by a Rodrigues-type formula, and provides convenient expressions for the Fourier coefficients of an arbitrary function. Application of the covariance identity yields some novel expressions for the corresponding lower variance bounds for a function of the r.v. $X$, expressions that seem to be known only in particular cases (for the Normal, see [Houdré and Kagan, J. Theoret. Probab. 8 (1995) 23--30]; see also [Houdré and Pérez-Abreu, Ann. Probab. 23 (1995) 400--419] for corresponding results related to the Wiener and Poisson processes). Some applications are also given. △ Less

Submitted 3 May, 2011; v1 submitted 21 July, 2010; originally announced July 2010.

Comments: Published in at http://dx.doi.org/10.3150/10-BEJ282 the Bernoulli (http://isi.cbs.nl/bernoulli/) by the International Statistical Institute/Bernoulli Society (http://isi.cbs.nl/BS/bshome.htm)

Report number: IMS-BEJ-BEJ282

Journal ref: Bernoulli 2011, Vol. 17, No. 2, 507-529