Multivariate Rényi inaccuracy measures based on copulas: properties and application

Shital Saha and Suchandan Kayal · 2025

Read Multivariate Rényi inaccuracy measures based on copulas: properties and application by Shital Saha and Suchandan Kayal (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-17215.txt.

Introduction

Shital Saha and Suchandan Kayal propose new measures of inaccuracy for joint data in this 2025 paper. Inaccuracy measures how far an assumed model sits from the true spread of data. Their versions build on a classic entropy family and on tools that capture how several variables move together.

Bounds for the new measures are derived from a known limit on joint spreads. The authors compare the measures under orderings that sort joint behavior from below and from above. Two sister measures are also proposed from related joint tools. A method to estimate one measure from data is given, and its error is studied in simulation. A real data set shows the measure can help pick a joint model.

Catalog record

CREATORShital Saha and Suchandan Kayal
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE76 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-17215

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Multivariate Rényi inaccuracy measures based on copulas: properties and application Shital Saha and Suchandan Kayal Abstract Department of Mathematics, National Institute of Technology Rourkela, Rourkela-769008, Odisha, India Abstract We propose Rényi inaccuracy measure based on multivariate copula and multivariate survival copula, respectively dubbed as multivariate cumulative copula Rényi inaccuracy measure and multivariate survival copula Rényi inaccuracy measure. Bounds of multivariate cumulative copula Rényi inaccuracy and multivariate survival copula Rényi inaccuracy measures have been obtained using Fréchet-Hoeffding bound. We discuss the comparison studies of the multivariate cumulative copula Rényi inaccuracy and multivariate survival copula Rényi inaccuracy measures based on lower orthant and upper orthant orders. We have also proposed multivariate co-copula Rényi inaccuracy and multivariate dual copula Rényi inaccuracy measures based on multivariate co-copula and dual copula. Similar properties have been explored. Further, we propose semiparametric estimator of multivariate cumulative copula Rényi inaccuracy measure. A simulation study is performed to compute standard deviation, absolute bias and mean squared error of the proposed estimator. Finally, a data set is considered to show that the multivariate cumulative copula Rényi inaccuracy measure can be applied as a model (copula) selection criteria. Keywords: Multivariate cumulative copula Rényi inaccuracy measure; multivariate survival copula Rényi inaccuracy measure; multivariate co-copula Rényi inaccuracy measure; multivariate dual copula Rényi inaccuracy measure; semiparametric estimation. MSCs: 94A17; 60E15; 62B10. 1 Introduction A copula is a class of functions that can be used to describe a stochastic concept of dependency of the random variables. Using the concept of copula, a joint distribution can be connected with the marginals arising from different families of probability distributions. In several applications of finance and economy, such as pricing, banking and risk management, the very popular measure “correlation coefficient” has been widely used as a measure of dependence. It mainly measures the linear dependence of the normal random variables. However, when one is interested to measure nonlinear dependence in random variables, the idea of correlation coefficient fails. It is further possible to have linearly uncorrelated random variables, although they have non-linear correlation between them. The copula functions are introduced in order to capture nonlinear dependence between the random variables. In particular, the copula functions are able to capture the dependence in the tail region for non-normal variables. In other words, we say that the copula completely describes (asymmetric and tail) dependence between the random variables. Due to these properties, the concept of copula has been used in many applied fields. For example, Zhang and Jiang (2019) found some applications of the copula function in financial risk analysis. The authors have done some progress in the field of internet finance employing the copula function. Due to flexibility of copula function, it has been used in the modelling of several degradation processes. Fang et al. (2020) developed a copula-based framework for analyzing the reliability of a degrading system. Xiang et al. (2023) developed a copula-based probabilistic method to investigate the yield loss probability to various drought conditions in south-eastern Australia. It is of recent interest to study copula-based information and divergence measures. For example, Ma and Sun (2011) combined the concepts of copula and entropy, and then introduced copula entropy. They established that there is no difference between the negative copula entropy and the mutual information. They have also proposed a method for mutual information estimation. Hosseini and Nooghabi (2021) proposed two inaccuracy measures using co-copula and dual of a copula. Further, the authors have investigated its various properties. Preda et al. (2023) studied some generalized copula-based inaccuracy measures with some properties. Saha and Kayal (2023) addressed copula-based extropy measures. The authors have obtained relations among cumulative copula extropy, Spearman’s rho, Kendall’s tau and Blest’s measure of rank correlation. Additionally, some estimators of the copula-based measures have been proposed. Sunoj and Nair (2023) proposed survival copula entropy and then explored various properties. They discussed an application of the copula-based proposed measure to an aortic regurgitation study. Multivariate cumulative copula entropy (CCE) has been proposed by Arshad et al. (2024) and some properties have been addressed. Using empirical beta copula, the authors presented a non-parametric estimator of the CCE. Ghosh and Sunoj (2024) studied various mutual information measures and mutual entropy based on copula theory. We note that the inaccuracy measure between two distributions was proposed by Kerridge (1961) . Let X and Y be nonnegative random variables with absolutely continuous cumulative distribution functions (CDFs) F_{X}(\cdot) and F_{Y}(\cdot) , probability density functions (PDFs) f_{X}(\cdot) and f_{Y}(\cdot) , respectively. Further, assume that f_{X}(\cdot) be the probaility density function (PDF) for a set of data points and f_{Y}(\cdot) be the assigned PDF. Then, the inaccuracy between X and Y is measured by \displaystyle I(X,Y)=-\int_{0}^{\infty}f_{X}(x)\log f_{Y}(x)dx=E_{f_{X}}\left(-\log f_{Y}(X)\right), (1.1) which is the average information content of the assigned distribution with respect to the true density. It quantifies the deflexion of a probabilistic model of interest from a reference model. I(X,Y) is useful in model selection since minimizing the Kullback-Leibler distance (see Kullback and Leibler (1951) ) is equivalent to minimizing the Kerridge inaccuracy measure. For some development of inaccuracy measure in comparing various statistical distributions, we refer to Choe (2017) . Besides this, the inaccuracy measure has applications in many applied fields. Few applications of the inaccuracy measure in coding theory were discussed by Nath (1968) . We recall that the inaccuracy measure in ( 1.1 ) is a generalization of the Shannon entropy, given by (see Shannon (1948) ) \displaystyle H(X)=-\int_{0}^{\infty}f_{X}(x)\log f_{X}(x)dx. (1.2) We get ( 1.2 ) from ( 1.1 ), after substituting f_{X}(\cdot) in place of f_{Y}(\cdot) . Due to the importance of the inaccuracy measure, it has been considered by several researchers, see for instance Kozesnik and Fischer (1978) , Taneja et al. (2009) , Kumar et al. (2011) , Balakrishnan et al. (2024) , Mohammadi et al. (2024) , and the references cited therein. We notice various attempts by researchers in developing the generalizations of the information/divergence measures. Due to the presence of an extra parameter or more, the generalized measures have been very useful in many applied fields. Among the parametric generalizations, the Renyi entropy is the most eminent measure of uncertainty (see Rényi (1961) ), given by \displaystyle H_{\gamma}(X)=\psi(\gamma)\log\int_{0}^{\infty}f^{\gamma}_{X}(x)dx,~~0<\gamma(\neq 1), (1.3) where \psi(\gamma)=\frac{1}{1-\gamma}. Note that H_{\gamma}(X) is a parametric generalization of H(X) , which can be deduced taking \gamma tending to 1. For some applications of the Renyi entropy, the interested readers may refer to Baraniuk et al. (2001) , Lake (2005) , Bashkirov (2006) , Ghosh and Basu (2021) and Liu et al. (2024) . Motivated by ( 1.3 ), Nath (1968) proposed a parametric generalization of the Kerridge inaccuracy measure, given by I_{\gamma}(X,Y)=\psi(\gamma)\log\int_{0}^{\infty}f_{X}(x)f^{\gamma-1}_{Y}(x)dx,~~0<\gamma(\neq 1). (1.4) Note that as \gamma tends to 1, ( 1.4 ) becomes the inaccuracy measure in ( 1.1 ). Further, when f_{X}(\cdot)=f_{Y}(\cdot) , ( 1.4 ) reduces to the Renyi entropy given in ( 1.3 ). Along this line of researcher, other generalized inaccuracy measure has been proposed and studied by Kayal and Sunoj (2017) and Kayal et al. (2017) . Motivated by the cumulative residual entropy by Rao et al. (2004) , Ghosh and Kundu (2020) proposed cumulative residual inaccuracy of order \gamma as CRI_{\gamma}(X,Y)=\psi(\gamma)\log\int_{0}^{\infty}\overline{F}_{X}(x)\overline{F}^{\gamma-1}_{Y}(x)dx,~~0<\gamma(\neq 1). (1.5) Analogous to ( 1.5 ), Ghosh and Kundu (2018) provided another measure, known

Read the complete text