The modified odd Burr XII-G family of distributions: Properties and applications

Alexsandro A. Ferreira · 2025

Read The modified odd Burr XII-G family of distributions: Properties and applications by Alexsandro A. Ferreira (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-2510-17567.txt.

Introduction

Statisticians build new families of probability distributions to fit data whose shapes older curves cannot match, such as data with two humps or a heavy bathtub shaped risk curve. Alexsandro A Ferreira and Gauss M Cordeiro develop one such family, the modified odd Burr XII-G family, which grows out of an earlier exponentiated class of distributions. The construction is meant to bend a wide range of baseline distributions into more flexible shapes.

Ferreira and Cordeiro derive the mathematical properties of the new family and build a regression model inside it, so that outcomes can be tied to explanatory variables. They estimate the parameters by maximum likelihood and run simulations to check that the estimates settle near the true values as samples grow. Three real data sets show the practical value of the proposals.

Catalog record

CREATORAlexsandro A. Ferreira
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-17567

Browse the full library · About the Tower of Babel

Full text

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

The modified odd Burr XII-G family of distributions: Properties and applications Alexsandro A. Ferreira Abstract The modified odd Burr XII-G family is developed, capable of incorporating bimodal and bathtub shapes in its baseline distributions, with properties derived from the exponentiated-G class. A regression model is developed within this family. The parameters are estimated by maximum likelihood, and simulations are performed to verify their consistency. The usefulness of the proposals is demonstrated by means of three real data sets. Keywords: COVID-19, Dengue, Maximum likelihood estimation, Modified odd log-logistic-G, Regression model, Weibull distribution 1 Introduction The flexibility of classical distributions, such as Weibull, gamma, and exponential, has been a significant focus of research in recent decades. Several studies have introduced new parameters into these distributions to improve their modeling capabilities and adapt them to various types of data. A notable example is the approach proposed by Mudholkar and Srivastava, 1993 , which adds an extra parameter to the Weibull distribution, allowing it to handle a bathtub-shaped failure rate function (hrf). In this context, several other studies have made substantial contributions, including those by Marshall and Olkin, 1997 , Gupta et al., 1998 , Eugene et al., 2002 , Zografos and Balakrishnan, 2009 , Cordeiro and de Castro, 2011 , Alexander et al., 2012 , Cordeiro et al., 2013 , Alzaatreh et al., 2013 , Bourguignon et al., 2014 , Alizadeh et al., 2015 , Chipepa et al., 2019 , Baharith and Alamoudi, 2021 , and Tlhaloganyang et al., 2022 , among many others. Consider G(x)=G(x;\bm{\xi}) as the cumulative distribution function (cdf) of any given baseline distribution, where \bm{\xi} is the parameter vector for G(x) . Let r(v)=r(v;\bm{\psi}) denote the probability density function (pdf) of a random variable V\in[\,c,d\,] (for -\infty\leq c<d\leq\infty ), having parameter vector \bm{\psi} . Next, take a function W[G(x)]\in[\,c,d\,] of G(x) . This function is assumed to be differentiable, monotonically non-decreasing, and W[G(x)]\rightarrow c as x\rightarrow-\infty and W[G(x)]\rightarrow d as x\rightarrow\infty . Under these conditions, the transformed-transformer (T-X) family has a cdf in the form ( Alzaatreh et al., 2013 ) . \displaystyle F(x)=F(x;\bm{\xi},\bm{\psi})=\int^{W[G(x;\bm{\xi})]}_{0}r(v;\bm{\psi})\,\text{d}v\,. (1) Based on ( 1 ), the modified odd Burr XII-G (MOBXII-G) family is proposed, which is obtained using W[G(x)]=\frac{G(x)}{1-G(x)[1+G(x)]/2} proposed by Chesneau and El Achi, 2020 , and r(v)=\tau\,\lambda\,v^{\tau-1}(1+v^{\tau})^{-(\lambda+1)} having the pdf of the Burr XII distribution. Thus, the cdf of the new family takes the form (for x\in\text{I\!R} , and \tau,\lambda>0 ) \displaystyle F(x)=1-\left(1+\left[\frac{2\,G(x)}{2-G(x)[1+G(x)]}\right]^{\tau}\right)^{-\lambda}\,, (2) and its pdf can be expressed as \displaystyle f(x)= \displaystyle\frac{2^{\tau}\,\tau\,\lambda\,g(x)\,[2+G(x)^{2}]\,G(x)^{\tau-1}}{\{2-G(x)[1+G(x)]\}^{\tau+1}}\left\{1+\left[\frac{2\,G(x)}{2-G(x)[1+G(x)]}\right]^{\tau}\right\}^{-(\lambda+1)}\,,\,\,x\in\text{I\!R}\,. (3) The hrf associated with ( 3 ) is \displaystyle h(x)=\frac{2^{\tau}\,\tau\,\lambda\,g(x)\,[2+G(x)^{2}]\,G(x)^{\tau-1}}{\{2-G(x)[1+G(x)]\}^{\tau+1}}\left\{1+\left[\frac{2\,G(x)}{2-G(x)[1+G(x)]}\right]^{\tau}\right\}^{-1}\,. (4) Henceforth, let X\sim\text{MOBXII-G}(\tau,\lambda,\bm{\xi}) be a random variable with pdf ( 3 ). Note that for \lambda=1 , the MOBXII-G reduces to modified odd log-logistic-G (MOLL-G), with only one extra parameter. The introduction of the MOBXII-G family is primarily driven by its enhanced flexibility compared to other established families, particularly the Kumaraswamy-G (K-G) ( Cordeiro and de Castro, 2011 ) and beta-G (B-G) ( Eugene et al., 2002 ) . These two families have been extensively studied in the literature, giving rise to over 100 distinct distributions each, as observed by Selim, 2020 . Their widespread application and acceptance underscore their robustness and adaptability in fitting diverse data sets. However, the MOBXII-G family stands out as a particularly attractive alternative in this area due to its superior flexibility. This is evident in its ability to handle bimodal and bathtub shapes in its baseline models, as illustrated in Figures 1 , 2 , and 4 , which show several examples of these shapes. As a result, the MOBXII-G family can more efficiently model real-world data with these characteristics, as proven by the empirical analysis in Section 6 . The rest of the article unfolds as follows. Section 2 discusses three special models of the new family and Section 3 describes its main properties. Section 4 provides a regression model for a special case of the new family, and simulations are reported in Section 5 . Section 6 presents three applications to real data, and some conclusions are addressed in Section 7 . 2 Some MOBXII-G models This section presents the pdfs for three members of the new family. Their cdfs and hrfs are readily obtained using Equations ( 2 ) and ( 4 ), respectively. The figures presented here are generated in R ( R Core Team, 2023 ) . 2.1 The modified odd Burr XII Weibull (MOBXIIW) Consider the Weibull distribution ( \alpha,\beta>0 ) in ( 3 ) with cdf (for x>0 ) \displaystyle G(x)=1-\text{e}^{-(x/\beta)^{\alpha}}. Thus, the density of the MOBXIIW becomes \displaystyle f(x)=\,\, \displaystyle\frac{2^{\tau}\,\tau\,\lambda\,\alpha\,\beta^{-\alpha}x^{\alpha-1}\text{e}^{-(x/\beta)^{\alpha}}\left[2+\left(1-\text{e}^{-(x/\beta)^{\alpha}}\right)^{2}\right]\left(1-\text{e}^{-(x/\beta)^{\alpha}}\right)^{\tau-1}}{\left[2-\left(1-\text{e}^{-(x/\beta)^{\alpha}}\right)\left(2-\text{e}^{-(x/\beta)^{\alpha}}\right)\right]^{\tau+1}} \displaystyle\times\left\{1+\left[\frac{2\left(1-\text{e}^{-(x/\beta)^{\alpha}}\right)}{2-\left(1-\text{e}^{-(x/\beta)^{\alpha}}\right)\left(2-\text{e}^{-(x/\beta)^{\alpha}}\right)}\right]^{\tau}\right\}^{-(\lambda+1)} (5) Figure 1: Density and hrf of \text{MOBXIIW}(\tau,\lambda,\alpha,\beta) . 2.2 The modified odd Burr XII Kumaraswamy (MOBXIIK) The Kumaraswamy cdf with parameters a,b>0 is expressed by (for 0<x<1 ) \displaystyle G(x)=1-(1-x^{a})^{b}\,. (6) From ( 3 ) and ( 6 ), the density of the MOBXIIK can be expressed as \displaystyle f(x)= \displaystyle\,\,\frac{2^{\tau}\,\tau\,\lambda\,a\,b\,x^{a-1}(1-x^{a})^{b-1}\left\{2+\left[1-\left(1-x^{a}\right)^{b}\right]^{2}\right\}\left[1-\left(1-x^{a}\right)^{b}\right]^{\tau-1}}{\left\{2-\left[1-\left(1-x^{a}\right)^{b}\right]\left[2-\left(1-x^{a}\right)^{b}\right]\right\}^{\tau+1}} \displaystyle\times\left\{1+\left(\frac{2\left[1-\left(1-x^{a}\right)^{b}\right]}{2-\left[1-\left(1-x^{a}\right)^{b}\right]\left[2-\left(1-x^{a}\right)^{b}\right]}\right)^{\tau}\right\}^{-(\lambda+1)}\,. Figure 2: Density and hrf \text{MOBXIIK}(\tau,\lambda,a,b) 2.3 The modified odd Burr XII normal (MOBXIIN) The MOBXIIN density follows from ( 3 ) and the normal baseline N(\mu,\sigma^{2}) (for \mu\in\text{I\!R} , and \sigma>0) as \displaystyle f(x)= \displaystyle\frac{2^{\tau}\,\tau\,\lambda\,\phi(z)\,[2+\Phi(z)^{2}]\,\Phi(z)^{\tau-1}}{\{2-\Phi(z)[1+\Phi(z)]\}^{\tau+1}}\left\{1+\left[\frac{2\,\Phi(z)}{2-\Phi(z)[1+\Phi(z)]}\right]^{\tau}\right\}^{-(\lambda+1)}\,, where z=(x-\mu)/\sigma , and \phi(\cdot) and \Phi(\cdot) are the pdf and cdf of the standard normal, respectively. Figure 3: Density and hrf of \text{MOBXIIN}(\tau,\lambda,\mu,\sigma) . Figure 4: Density and hrf of \text{MOBXIIN}(\tau,\lambda,\mu,\sigma) . The pdfs and hrfs of three new members of the family are illustrated in Figures 1 , 2 , and 4 , respectively. These models can handle a wide range of data, including right-skewed, symmetric, left-skewed, and even bimodal data. Additionally, their hrfs can have various shapes, including unimodal, increasing-decreasing-increasing, or bathtub. 3 Properties 3.1 Useful expansions The well-studied exponentiated-G (exp-G) class has a pdf of \pi_{\delta}(x)=\delta\,g(x)\,G(x)^{\delta-1} (for \delta>0 ). Several notable distributions fall under this class, including the exp-Weibull ( Mudholkar and Srivastava, 1993 ) , exp-exponential ( Gupta and Kundu, 2001 ) , exp-Fréchet ( Nadarajah and Kotz, 2003 ) , and exp-gamma ( Nadarajah and Gupta, 2007 ) , as documented in Table 1 of Tahir and Nadarajah, 2015 . Thus, the density of the MOBXII-G family admits an expression based on the density of the exp-G class. Initially, the cdf given in ( 2 ) can be expressed as \displaystyle F(x)=1-\left(\frac{\left\{1-G(x)[1+G(x)]/2\right\}^{\tau}}{\left\{1-G(x)[1+G(x)]/2\right\}^{\tau}+G(x)^{\tau}}\right)^{\lambda}\,. (7) Applying the binomial theorem twice, one has \displaystyle\left\{1-G(x)[1+G(x)]/2\right\}^{\tau}=\sum^{\infty}_{m=0}a_{m}G(x)^{m}\,, (8) where \displaystyle a_{m}=a_{m}(\tau)=\sum_{(i,j)\in I_{m}}(-1)^{i}2^{-i}\binom{\tau}{i}\binom{i}{j}\,, and I_{m}=\{(i,j)\in\mathbb{N}^{2}_{0}\mid i+j=m,j\leq i\} , \mathbb{N}_{0}=\{0,1,2,\ldots\} . Again, by the binomial theorem, a power series for G(x)^{\tau} can be found as \displaystyle G(x)^{\tau}=\{1-[1-G(x)]\}^{\tau}=\sum^{\infty}_{m=0}b_{m}G(x)^{m}\,, (9) where \displaystyle b_{m}=b_{m}(\tau)=\sum^{\infty}_{\ell=m}(-1)^{\ell+m}\binom{\tau}{\ell}\binom{\ell}{m}\,. By inserting ( 8 ), and ( 9 ) into ( 7 ), \displaystyle F(x)=1-\left(\frac{\sum^{\infty}_{m=0}a_{m}G(x)^{m}}{\sum^{\infty}_{m=0}d_{m}G(x)^{m}}\right)^{\lambda}\,, where d_{m}=a_{m}+b_{m} . Then, the quotient of two power series can be represented as \displaystyle F(x)=1-\left(\sum^{\infty}_{m=0}\omega_{m}\,G(x)^{m}\right)^{\lambda}\,, (10) where \omega_{0}=a_{0}/d_{0} , and for m>0 , \displaystyle\omega_{m}=\frac{1}{d_{0}}\left(a_{m}-\sum^{m}_{n=1}d_{n}\,\omega_{m-n}\right)\,. Following the findings of Munir, 2013 for a power series raised to a non-zero real number, the cdf of the MOBXII family can be expressed in the form \displaystyle F(x)=1-\sum^{\infty}_{m=0}\vartheta_{m}\,G(x)^{m}\,, (11) where \vartheta_{0}=\omega_{0}^{\lambda} , and for m>0 , \displaystyle\vartheta_{m}=\frac{1}{m\,\omega_{0}}\left(\sum^{m-1}_{q=0}\left[\lambda\,m-(\lambda+1)q\right]\vartheta_{q}\,\omega_{m-q}\right)\,. The pdf of the MOBXII-G family follows by differentiating ( 11 ) as \displaystyle f(x)=\sum^{\infty}_{m=0}\,\varphi_{m+1}\,\pi_{m+1}(x)\,, (12) where \varphi_{m+1}=-\vartheta_{m+1} and \pi_{m+1}(x) is the density of the exp-G class with power (m+1) . Thus, Equation ( 12 ) shows that the density of the new family

Read the complete text