A new method for generalizing Burr and related distributions

Authors

  • Tanujit Chakraborty Sorbonne University Abu Dhabi
  • Suchismita Das S P Jain School of Global Management
  • Swarup Chattopadhyay Indian Statistical Institute

DOI:

https://doi.org/10.1515/ms-2022-0016

Keywords:

Burr distribution, Exponentiated distributions, Stochastic ordering, Reliability properties, Maximum likelihood

Abstract

A new method has been proposed to generalize Burr-XII distribution, also called Burr distribution, by adding an extra parameter to an existing Burr distribution for more
exibility. In this method, the exponent of the Burr distribution is modeled using a nonlinear function of the data and one additional parameter. The models of this newly introduced generalized Burr family can signifficantly increase the exibility of the former Burr distribution with respect to the density and hazard rate shapes. Families expanded using the method proposed here is heavy-tailed and belongs to the maximum domain of attractions of the Frechet distribution. The method is further applied to yield three-parameter classical Pareto and generalized exponentiated distributions which shows the broader application of the proposed idea of generalization. A relevant model of the new generalized Burr family has been considered in detail, with particular emphasis on the hazard functions, stochastic orders, estimation procedures, and testing methods are derived. Finally, as empirical evidence, the new distribution is applied to the analysis of large-scale heavy-tailed network data and compared with other commonly used distributions available for fitting degree distributions of networks. Experimental results suggest that the proposed Burr distribution with nonlinear exponent better fits the large-scale heavytailed networks better than the popularly used Marhsall-Olkin generalization of Burr and exponentiated Burr distributions.

Author Biographies

Tanujit Chakraborty, Sorbonne University Abu Dhabi

Department of Science and Engineering
Sorbonne University Abu Dhabi
Abu Dhabi
UAE

Suchismita Das, S P Jain School of Global Management

Department of Data Science
S P Jain School of Global Management
Mumbai-400070
INDIA

Swarup Chattopadhyay, Indian Statistical Institute

Machine Intelligence Unit
Indian Statistical Institute
203, B.T. Road
Kolkata-700108
INDIA

Published

2022-02-16

Issue

Section

Articles - Other topics