See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/327130546

The odd Lomax generator of distributions: Properties, estimation and

applications

Article  in  Journal of Computational and Applied Mathematics · September 2019

DOI: 10.1016/j.cam.2018.08.008

CITATIONS READS

0 164

5 authors, including:

Gauss M. Cordeiro Ahmed Z. Afify

Federal University of Pernambuco Benha University
414 PUBLICATIONS   6,778 CITATIONS    57 PUBLICATIONS   454 CITATIONS   

SEE PROFILE SEE PROFILE

Edwin Ortega
University of São Paulo
178 PUBLICATIONS   3,730 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

new distributions View project

The Odd Log-Logistic Dagum Distribution: Properties and Applications View project

All content following this page was uploaded by Ahmed Z. Afify on 15 September 2018.

The user has requested enhancement of the downloaded file.

 Journal of Computational and Applied Mathematics 347 (2019) 222–237

Contents lists available at ScienceDirect

Journal of Computational and Applied

Mathematics
journal homepage: www.elsevier.com/locate/cam

The odd Lomax generator of distributions: Properties,

estimation and applications
Gauss M. Cordeiro a , Ahmed Z. Afify b, *, Edwin M.M. Ortega c ,
Adriano K. Suzuki d , Mohamed E. Mead e,f
a
Departamento de Estatística, Universidade Federal de Pernambuco, Brazil
b
Department of Statistics, Mathematics and Insurance, Benha University, Egypt
c
Departamento de Ciências Exatas, Universidade de São Paulo, Brazil
d
Departamento de Matemática Aplicada e Estatística, Universidade de São Paulo, Brazil
e
Department of Statistics, Mathematics and Insurance, Zagazig University, Egypt
f
Department of Mathematics, Faculty of Science and Arts, Khulais, University of Jeddah, Saudi Arabia

article info a b s t r a c t
Article history: We introduce a new family of continuous distributions called the odd Lomax-G class and
Received 6 October 2017 provide four special models. We derive explicit expressions for the ordinary and incomplete
Received in revised form 19 July 2018 moments, generating function, Rényi entropy, order statistics and probability weighted
moments. The maximum likelihood and least squares methods are used to estimate the
MSC:
model parameters. The flexibility of the proposed family is illustrated by means of two
62N01
applications to real data sets.
62N02
© 2018 Elsevier B.V. All rights reserved.
Keywords:
Generating function
Lomax model
Maximum likelihood
Order statistic
T-X family

1. Introduction

In many applied sciences such as medicine, engineering and finance, among others, modeling and analyzing lifetime
data are crucial. Several lifetime models have been adopted to model different types of survival data. The quality of the
procedures used in a statistical analysis depends heavily on the generated family of distributions and considerable effort has
been directed to define new statistical models. However, there still remain many important problems involving real data,
which do not follow any of the popular statistical models. So, the procedure of expanding a family of distributions by adding
new shape parameters is well recognized in the statistical literature.
The role of the extra shape parameters is to introduce skewness and to vary tail weights. Further, various models
have been constructed by extending some useful lifetime distributions and investigated them with respect to different
characteristics. Some generalized distributions provide more flexibility to model both monotonic and non-monotonic failure
rates even though the baseline failure rate may be monotonic.
Several distributions have been proposed to model lifetime data. The Lomax (Lx) distribution is one of the most commonly
used distributions for this purpose and it has many applications in a variety of fields such as medical and biological sciences,

*Corresponding author.
E-mail addresses: [email protected] (G.M. Cordeiro), [email protected] (A.Z. Afify), [email protected] (E.M.M. Ortega), [email protected]
(A.K. Suzuki), [email protected] (M.E. Mead).

https://doi.org/10.1016/j.cam.2018.08.008
0377-0427/© 2018 Elsevier B.V. All rights reserved.
 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237 223

income and wealth inequality, engineering, actuarial science, lifetime and reliability modeling. It also has been suggested
as heavy tailed alternative to the exponential, Weibull and gamma distributions. Properties of the Lx distribution have been
studied by several authors, for example, moments of order statistics [1], inference under generalized order statistics [2],
estimation by generalized probability weighted moments [3] and comparison of Bayesian and non-Bayesian estimation
based on record values [4] and [5].
Many authors developed generalizations of the Lx distribution. For example, [6] proposed the exponentiated Lx, [7]
and [8] studied the Marshall–Olkin extended Lx, [9] introduced the beta Lx, Kumaraswamy Lx and McDonald Lx, [10]
pioneered the gamma Lx, [11] defined the Poisson-Lx, [12] studied the Weibull Lx and [13] proposed the transmuted Weibull
Lx distributions.
In this paper, we define and study a new family of distributions based on the Lx distribution. In fact, based on the T-X
family pioneered by [14], we construct a new generator so-called the odd Lomax-G (OLxG for short) family and provide some
of its mathematical properties.
The cumulative distribution function (cdf) and probability density function (pdf) (for x > 0) of the OLxG generator with
two additional shape parameters α > 0 and β > 0 are defined by
G(x;ϕ) ]−α
G(x; ϕ)
∫ [
1−G(x;ϕ)
F (x; α, β, ϕ) = αβ α (β + t )−α−1 dt = 1 − β α β + (1)
0 1 − G(x; ϕ)
and
αβ α g(x; ϕ)
]−α−1
G(x; ϕ)
[
f (x; α, β, ϕ) = β+ , (2)
[1 − G(x; ϕ)] 2
1 − G(x; ϕ)
respectively, where g(x; ϕ) = dG(x; ϕ)/dx. Henceforth, a random variable with density function (2) is denoted by
X ∼OLxG(α, β, ϕ). For α = 1, the OLx family becomes the Marshall–Olkin class [15].
We can omit the dependence on the parameters and write G(x) = G(x; ϕ), f (x) = f (x; α, β, ϕ), etc. Let G(x) = 1 − G(x).
The hazard rate function (hrf) of X is given by
ατ (x)
h(x) = ,
β + (1 − β) G(x)
By inverting (1), the quantile function (qf) of X follows as
{ 1
}
β − β (1 − u) α
Q (u) = F −1
(u) = QG 1
, (3)
β + (1 − β) (1 − u) α
where QG (u) = G−1 (u) is the qf of the baseline G distribution and u ∈ (0, 1).
The rest of the paper is organized as follows. In Section 2 , we present four special distributions in the new family. In
Section 3, we derive a linear representation for the OLxG pdf. In Section 4, we obtain some mathematical properties of
the OLx class. Estimation of the model parameters by maximum likelihood is presented in Section 5. We provide simulation
results for the maximum likelihood estimates (MLEs) in the OLx-Weibull and OLx-Lomax models in Section 6. In Section 7, we
estimate the parameters of two special models of the OLxG family using least-squares method and perform some simulations
to verify the adequacy of these estimates. Two applications to real data prove empirically the usefulness of the new family
in Section 8. Finally, some conclusions are offered in Section 9.

2. Some special OLxG models

In this section, we provide four special models of the OLx family. The pdf (2) will be most tractable when the functions
G(x) and g(x) have simple analytic expressions.

2.1. The OLx-Weibull (OLxW) distribution

b b
Let G(x) = 1 − exp[−(a x) ] and g(x) = b ab xb−1 exp[−(a x) ] (for x > 0) be the cdf and pdf of the Weibull distribution
with positive parameters a and b, respectively. Then, the pdf of the OLxW model can be easily determined from (2). For
b = 1, we obtain the OLx-exponential (OLxE) distribution. For β = 2, we have the OLx-Rayleigh (OLxR) distribution. Fig. 1
displays some possible shapes of the OLxW density function.

2.2. The OLx-Lomax (OLxLx) distribution

The cdf and pdf (for x > 0) of the Lx distribution with positive parameters a and b are G(x) = 1 − [1 + (x/b)]−a and
g(x) = (a/b)[1 + (x/b)]−(a+1) , respectively. Then, the OLxLx density function follows simply from (2). Plots of this density
are displayed in Fig. 2.
224 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237

Fig. 1. Plots of the OLxW density function for some parameter values.

Fig. 2. Plots of the OLxLx density function for some parameter values.

2.3. The OLx-log-logistic (OLxLL) distribution

The cdf and pdf (for x > 0) of the log-logistic (LL) distribution with parameters a > 0 and b > 0 are
[ ( x )b ]−1 [ ( x )b ]−2
G(x) = 1 − 1 + and g(x) = ba −b b−1
x 1+ ,
a a
respectively. The OLxLL density function can be determined from (2). Fig. 3 displays plots of the OLxLL density function for
selected parameter values.

2.4. The OLx-Lindley (OLxLi) distribution

2
The Lindley (Li) distribution with one parameter a > 0 has pdf and cdf (for x > 0) given by g(x) = 1a+a (1 + x) e−ax
and G(x) = 1 − 1+1a++aax e−ax , respectively. Then, the OLxLi density function follows from (2). The plots in Fig. 4 reveal great
flexibility of the OLxLi density.

3. Linear representation

We provide a useful linear representation for the OLxG pdf. It can be expressed as
αβ α g(x)
[ ]−α−1
G(x)
f (x) = β+ . (4)
[1 − G(x)]2 1 − G(x)
  
A
 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237 225

Fig. 3. Plots of the OLxLL density function for some parameter values.

Fig. 4. Plots of the OLxLi density function for some parameter values.

Consider the power series

∞ ( )
∑ −q
(1 + dz ) −q
= dn z n . (5)
n
n=0

Applying (5) to the quantity A and inserting the resulting expansion in Eq. (4) gives
∞ (
α ∑ −α − 1 −k
)
f (x) = β g(x)G(x)k [1 − G(x)]−k−2 . (6)
β k
k=0

For |z | < 1, the power series holds

∞ ( )
∑ −q
(1 − z ) −q
= (−1) n
zn. (7)
n
n=0

Applying the power series (7) to [1 − G(x)]−k−2 and substituting the result in (6) yields
∞
∑
f (x) = υk,j hk+j+1 (x), (8)
k,j=0
226 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237

where
(−1)j α
( )( )
−α − 1 −k − 2
υk,j = ,
(k + j + 1) β k+1 k j
and hγ (x) = γ g (x) G(x)γ −1 denotes the exponentiated-G (ExG) class density with power parameter γ > 0. Then, the OLxG
density function can be expressed as a double linear combination of ExG densities. So, several mathematical quantities of
the OLxG family can be determined from those quantities of the ExG class. Eq. (8) is the main result of this section.
By integrating (8), we obtain
∞
∑
F (x) = υk,j Hk+j+1 (x),
k,j=0

where Hk+j+1 (x) is the cdf of the ExG family with power parameter (k + j + 1).

4. Properties of the OLx family

In this section, we obtain some structural properties of the OLx family including the ordinary and incomplete moments,
moment generating function (mgf), Rényi entropy, order statistics and probability weighted moments (pwms). Henceforth,
let Zk+j+1 denote a random variable having the ExG distribution with power parameter k+j+1. Some mathematical quantities
of X can be expressed from those of Zk+j+1 .

4.1. Ordinary and incomplete moments

The rth moment of X , say µ′r , follows directly from (8) as

∞
∑
µ′r = E(X r ) = υk,j E(Zkr+j+1 ). (9)
k,j=0

The sth incomplete moment of X can be expressed from (8) as

∫ z ∞
∑ ∫ z
ms (z) = s
x f (x)dx = υk,j xs hk+j+1 (x)dx. (10)
−∞ k,j=0 −∞

The first incomplete moment of X can be obtained from (10) as

∞
∑
m1 (z) = υk,j Jk+j+1 (z),
k,j=0
∫z ∫ G(z)
where Jk+j+1 (z) = −∞ x hk+j+1 (x)dx = (k + j + 1) 0 QG (u) uk+j du is the first incomplete moment of the ExG class. It can
be computed analytically or numerically from the baseline qf QG (u) = G−1 (u; ϕ). In Table 1, we provide numerical values for
the mean, variance, skewness and kurtosis of the OLxW distribution for some parameters to illustrate the effects on these
measures.

4.2. Generating function

The mgf M(t) = E(et X ) of X can be determined from Eq. (8) as

∞
∑ ∞
∑
M(t) = υk,j Mk+j+1 (t) = (k + j + 1) υk,j τ (t , k + j),
k,j=0 k,j=0
∫1
where Mk+j+1 (t) is the mgf of Zk+j+1 and τ (t , k + j) = 0
exp [t QG (u)] uk+j du. Hence, M(t) can follow from the ExG generating
function.

4.3. Rényi entropy

The Rényi entropy of a random variable X represents a measure of variation of the uncertainty. It is given by
(∫ ∞ )
Iθ (X ) = (1 − θ)−1 log f (x)θ dx , θ > 0 and θ ̸ = 1.
−∞
Using the pdf (2), we can write
]−θ (α+1)
(αβ α )θ g(x)θ
[
G(x)
f (x)θ = β + .
[1 − G(x)]2θ 1 − G(x)
 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237 227

Applying the power series (5) to the last term, we have

( )θ ∑
∞ (
α −θ (α + 1) −k
)
f (x)θ = β g(x)θ G(x)k [1 − G(x)]−2θ −k .
β k
k=0

After applying the power series (7) to [1 − G(x)]−2θ −k , we obtain the Rényi entropy of the OLx family as
[ ∞
]
∫ ∞
θ
∑
Iθ (X ) = (1 − θ ) −1
log mk,n g(x) G(x) n+k
dx ,
k,n=0 −∞

where
( )θ (
α −2θ − k −θ (α + 1)
)( )
mk,n = (−1)n β −k .
β n k

4.4. Order statistics

Order statistics make their appearance in many areas of statistical theory and practice. Let X1 , . . . , Xn be a random sample
from the OLx family. The pdf of Xi:n can be expressed (for i = 1, . . . , n) as
n−i
f (x)
( )
∑ n−i
fi:n (x) = (−1)j F (x)j+i−1 , (11)
B (i, n − i + 1) j
j=0

where B(·, ·) is the beta function.

By applying the binomial theorem in Eq. (1), we can write
j+i−1 ( ) [ ]−kα
j+i−1 G(x)
β kα β +
∑
F (x)j+i−1 = (−1)k . (12)
k 1 − G(x)
k=0

Combining (2) and (12), we obtain

j+i−1 ]−(k+1)α−1
αβ α g(x)
( ) [
j+i−1 G(x)
β kα β +
∑
f (x) F (x)j+i−1 = (−1)k .
[1 − G(x)]2 k=0
k 1 − G(x)

Applying the power series (5) to the last term, we have

j+i−1 ∞
α ∑∑ −(k + 1)α − 1 g(x)G(x)r
( )( )
j+i−1
f (x) F (x)j+i−1 = (−1)k β −r .
β k r [1 − G(x)]r +2
k=0 r =0

Applying again the power series (7) to [1 − G(x)]−r −2 , we obtain

j+i−1 ∞
α ∑ ∑
( )( )
−r − 2 j + i − 1
f (x) F (x)j+i−1 = (−1)k+s β −r
β s k
k=0 r ,s=0

−(k + 1)α − 1
( )
× g(x)G(x)s+r . (13)
r
Inserting the previous equation in (11), the pdf of Xi:n reduces to
∞
∑
fi:n (x) = dr ,s hs+r +1 (x), (14)
r ,s=0

where hs+r +1 (x) is the ExG density with power parameter s + r + 1 and
n−i j+i−1
∑ ∑ (−1)k+s αβ −r −1 (n − i)(−r − 2)
d r ,s =
B (i, n − i + 1) j s
j=0 k=0

−(k + 1)α − 1
( )( )
j+i−1
× .
k r
Then, the density function of the OLxG order statistics is a linear combination of ExG densities. Based upon Eq. (14), some
mathematical properties of Xi:n can follow from those properties of Zs+r +1 .
228 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237

4.5. Probability weighted moments

For j, i = 1, 2, . . ., the (j, i)th pwm of X , say ρj,i , is formally defined by

∫ ∞
ρj,i = E X F (X ) =
j i
xj F (x)i f (x)dx.
{ }
−∞

From Eq. (13), we can write

i ∞ ( )( )
∑ ∑ i −r − 2
f (x) F (x) = α
i
(−1) k+s
β −r −1
k s
k=0 r ,s=0

−(k + 1)α − 1
( )
× g(x) G(x)s+r .
r
The last equation can be expressed as
∞
∑
f (x) F (x)i = br(i),s hs+r +1 (x),
r ,s=0

where
i
(−1)k+s β −r −1 i −(k + 1)α − 1
( )( )( )
∑ −r − 2
b(i)
r ,s =α .
(s + r + 1) k s r
k=0

Hence, the quantity ρj,i is given by

∞
∑ ∫ ∞ ∞
∑ j
ρj,i = b(i)
r ,s xj hs+r +1 (x)dx = br(i),s E(Zs+r +1 ).
r ,s=0 −∞ r ,s=0

5. Maximum likelihood estimation

Several approaches for parameter estimation were proposed in the literature but the maximum likelihood method is
the most commonly employed. The MLEs enjoy desirable properties and can be used to obtain confidence intervals for the
model parameters. The normal approximation for these estimators in large samples can be easily handled either analytically
or numerically. We consider the estimation of the unknown parameters of the OLxG family from complete samples only by
maximum likelihood.
Let x1 , . . . , xn be a random sample from this family with parameters α, β and ϕ. Let θ =(α, β, ϕ⊺ )⊺ be the p × 1 parameter
vector. The log-likelihood function for θ is given by
n
∑
ℓ = ℓ( θ ) = n log (α) + nα log (β) + log g (xi ; ϕ)
i=1
n n
∑ ∑
+ (α − 1) log G (xi ; ϕ) − (α + 1) log [β + (1 − β) G(xi ; ϕ)] .
i=1 i=1

Explicit expressions for the score vector components,

∂ℓ )⊺
U (θ) = = Uα , Uβ , Uϕk ,
(
∂θ
can be obtained from the corresponding author upon request.
The above log-likelihood can be maximized numerically using the R ( optim function), SAS (PROC NLMIXED) or Ox
program (sub-routine MaxBFGS), among others. For interval estimation of the parameters, we determine the p × p observed
information matrix J(θ ) = {− ∂∂r ∂ℓs } (for r , s = α, β, ϕ ), whose elements can be obtained under request.
2

The standard likelihood regularity conditions are verified for the OLxG family since its support is assumed not depending
on unknown parameters and the functions G(x; ϕ) and g(x; ϕ) in (2) are considered twice differentiable with respect to
the variable x. Under these conditions, the distribution of θ̂ can be approximated by a multivariate normal Np (0, J(θ̂ )−1 )
distribution when n → ∞ to obtain confidence intervals for the parameters. Here, J(θ̂ ) is the total observed information
matrix evaluated at θ̂ . The method of the re-sampling bootstrap can be used for computing the biases of the MLEs of the
model parameters. Good interval estimates may also be obtained using the bootstrap percentile method.
 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237 229

Table 1
Mean, variance, skewness and kurtosis of the OLxW distribution for selected
parameter values.
α β b Mean Variance Skewness Kurtosis
0.5 0.75 0.5 7.0317 283.635 7.0303 98.2992
1.5 1.3276 0.9191 1.1705 4.6351
2 1.1776 0.4271 0.7189 3.3461
5 1.0244 0.0615 −0.1737 2.7621
0.5 1.5 0.5 9.5563 377.5911 6.0948 75.1646
1.5 1.5867 0.9767 0.9453 4.1256
2 1.3614 0.4277 0.5169 3.1586
5 1.0960 0.0537 −0.3627 3.0908
0.5 2 0.5 10.8069 423.1887 5.7597 67.6683
1.5 1.6986 0.9916 0.8638 3.9872
2 1.4385 0.4237 0.4427 3.1323
5 1.1243 0.0503 −0.4357 3.2696
0.5 5 0.5 15.6835 596.3156 4.8668 49.7403
1.5 2.0649 1.0048 0.6489 3.7623
2 1.6824 0.3975 0.2454 3.2049
5 1.2075 0.0397 −0.6389 3.9785
0.75 0.5 0.5 2.2947 40.4543 8.2417 133.6999
1.5 0.8621 0.4659 1.4119 5.5017
2 0.8457 0.2599 0.9194 3.7928
5 0.8922 0.0532 −0.0223 2.7137
1.5 0.5 0.5 0.4417 1.6781 9.6124 186.3075
1.5 0.5015 0.1514 1.4924 6.1007
2 0.5648 0.1102 0.9611 4.0935
5 0.7613 0.0364 −0.0277 2.8377
2 0.5 0.5 0.2218 0.4244 10.1498 213.313
1.5 0.4023 0.0943 1.4809 6.1545
2 0.4796 0.0769 0.9435 4.1072
5 0.7139 0.0311 −0.0518 2.8757
5 0.5 0.5 0.0264 0.0051 10.1066 233.6047
1.5 0.2049 0.0219 1.3288 5.5796
2 0.2903 0.0255 0.8181 3.8059
5 0.5859 0.0194 −0.1419 2.8924

6. Simulation study

In this section, we perform a simulation study in order to evaluate some frequentist properties of the MLE θ̂ . The
observations are simulated from the OLxG family by taking the Weibull and Lomax for the baseline distributions.
We consider the following values for the parameters: α = 1.50, β = 10.0, a = 5.0 and b = 0.50 for the OLxW model
and α = 0.50, β = 3.0, a = 10.0 and b = 5.0 for the OLxLx model. All simulations are performed using the R software with
the Optim function. The results are obtained from 1000 Monte Carlo simulations. In each replication, a random sample of
size n is drawn from the two distributions and the parameters are estimated by maximum likelihood. The sample sizes are
taken as n = 50, 100, 250, 500 and 750.
In Tables 2 and 3, we provide the averages of the MLEs (Mean), their biases (Bias) and mean square errors (MSE) for the
OLxW and OLxLx distributions, respectively.
We conclude from the figures in Tables 2 and 3 that the average estimates tend to be closer to the true parameters when
n increases in agreement with first-order asymptotic theory.

7. Least squares estimation

We also perform the estimation of the model parameters by least squares which is an alternative estimation method to
maximum likelihood. The least square estimates (LSEs), α̃ , β̃ , ã and b̃ of α , β , a and b, are defined as those arguments that
minimize the objective function
n
( ]−α )2
G(x(i) ; ϕ)
[
∑
α i
Q (θ ) = 1−β β + − ,
1 − G(x(i) ; ϕ) n+1
i=1

where x(i) is a possible outcome of the ith order statistic based on n observations obtained from X .
The minimum point α̃, β̃, ã and b̃ can also be given as a solution of the following system of non-linear equations
∂ Q (θ ) ∂ Q (θ ) ∂ Q (θ ) ∂ Q (θ )
= = = = 0.
∂α ∂β ∂a ∂b
230 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237

Table 2
Summaries of the MLEs for the OLxW distribution.
Sample size Parameter (True value) Summaries of the parameters
Mean Bias MSE
α (1.5) 1.1482 −0.3518 0.5459
n = 50 β (10.0) 9.3849 −0.6151 11.2325
a (5.0) 6.8545 1.8545 12.9516
b (0.5) 0.5191 0.0191 0.0046
α (1.5) 1.2693 −0.2307 0.3428
n = 100 β (10.0) 9.4219 −0.5781 2.1144
a (5.0) 6.1857 1.1857 6.6402
b (0.5) 0.5061 0.0061 0.0017
α (1.5) 1.436 −0.064 0.2783
n = 250 β (10.0) 9.683 −0.317 1.2252
a (5.0) 5.637 0.637 3.802
b (0.5) 0.4991 −0.0009 0.0006
α (1.5) 1.5137 0.0137 0.1946
n = 500 β (10.0) 9.9005 −0.0995 1.1666
a (5.0) 5.2304 0.2304 2.1305
b (0.5) 0.4983 −0.0017 0.0003
α (1.5) 1.5244 0.0244 0.1478
n = 750 β (10.0) 9.9379 −0.0621 0.8122
a (5.0) 5.1078 0.1078 1.5099
b (0.5) 0.4987 −0.0013 0.0003

Table 3
Summaries of the MLEs for the OLxLx distribution.
Sample size Parameter (True value) Summaries of the parameters
Mean Bias MSE
α (0.5) 0.9280 0.4280 0.9503
n = 50 β (3.0) 4.7330 1.7330 13.5051
a (10.0) 9.5779 −0.4221 3.3120
b (5.0) 6.5445 1.5445 17.3905
α (0.5) 0.8466 0.3466 0.5993
n = 100 β (3.0) 3.6189 0.6189 3.8423
a (10.0) 9.4323 −0.5677 2.5229
b (5.0) 6.5023 1.5023 12.258
α (0.5) 0.6186 0.1186 0.1027
n = 250 β (3.0) 3.1664 0.1664 0.9306
a (10.0) 9.7011 −0.2989 1.1883
b (5.0) 5.6671 0.6671 4.0882
α (0.5) 0.5501 0.0501 0.0309
n = 500 β (3.0) 3.0643 0.0643 0.4327
a (10.0) 9.8614 −0.1386 0.6001
b (5.0) 5.3210 0.3210 1.7653
α (0.5) 0.5305 0.0305 0.0133
n = 750 β (3.0) 3.0765 0.0765 0.3079
a (10.0) 9.9232 −0.0768 0.4663
b (5.0) 5.2014 0.2014 0.9163

The properties of the LSEs are not easy to obtain as those of the MLEs for more complex distributions. Tables 4 and 5
give the averages of the LSEs (Mean), biases (Bias) and MSEs for the OLxW and OLxLx distributions, respectively. Clearly, the
figures of these tables indicate that the LSEs become better for very large samples.

8. Applications

In this section, we provide two applications to real data to illustrate the flexibility of the OLxW, OLxLx and OLxLL
distributions introduced in Section 2. The MLEs of the model parameters are determined and some goodness-of-fit statistics
for these distributions are compared with other competitive models.
The model selection is carried out based upon the Anderson–Darling (A∗ ) and Cramer–von Mises (W ∗ ) statistics. The
smaller these statistics are, the better the fit is.
 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237 231

Table 4
Summaries of the LSEs for the OLxW distribution.
Sample size Parameter (True value) Summaries of the parameters
Mean Bias MSE
α (1.5) 1.2315 −0.2685 0.2704
n = 50 β (10.0) 9.5688 −0.4312 0.9492
a (5.0) 5.8466 0.8466 2.5795
b (0.5) 0.5197 0.0197 0.0050
α (1.5) 1.3026 −0.1974 0.1471
n = 100 β (10.0) 9.6948 −0.3052 0.5319
a (5.0) 5.5555 0.5555 1.2539
b (0.5) 0.5123 0.0123 0.0021
α (1.5) 1.3546 −0.1454 0.0959
n = 250 β (10.0) 9.6967 −0.3033 0.4997
a (5.0) 5.4908 0.4908 1.1284
b (0.5) 0.5054 0.0054 0.0007
α (1.5) 1.3877 −0.1123 0.0735
n = 500 β (10.0) 9.7397 −0.2603 0.3616
a (5.0) 5.4314 0.4314 0.9053
b (0.5) 0.5020 0.0020 0.0004
α (1.5) 1.4019 −0.0981 0.0664
n = 750 β (10.0) 9.7537 −0.2463 0.3434
a (5.0) 5.4051 0.4051 0.8753
b (0.5) 0.5011 0.0011 0.0002

Table 5
Summaries of the LSEs for the OLxLx distribution.
Sample size Parameter (True value) Summaries of the parameters
Mean Bias MSE
α (0.5) 0.5744 0.0744 0.0768
n = 50 β (3.0) 3.9065 0.9065 4.9080
a (10.0) 10.0674 0.0674 0.6779
b (5.0) 5.1273 0.1273 2.3522
α (0.5) 0.5677 0.0677 0.0586
n = 100 β (3.0) 3.5638 0.5638 2.2730
a (10.0) 9.9927 −0.0073 0.5362
b (5.0) 5.1785 0.1785 1.9514
α (0.5) 0.5511 0.0511 0.0353
n = 250 β (3.0) 3.2197 0.2197 0.7989
a (10.0) 9.9149 −0.0851 0.4158
b (5.0) 5.2243 0.2243 1.5613
α (0.5) 0.5419 0.0419 0.0251
n = 500 β (3.0) 3.1140 0.1140 0.4116
a (10.0) 9.8991 −0.1009 0.3374
b (5.0) 5.2273 0.2273 1.2494
α (0.5) 0.5350 0.0350 0.0181
n = 750 β (3.0) 3.0668 0.0668 0.2732
a (10.0) 9.9002 −0.0998 0.2430
b (5.0) 5.1969 0.1969 0.9110

8.1. Application 1: Cancer patients data

The first data set refers to the remission times (in months) of a random sample of 128 bladder cancer patients [16].
For these data, we compare the fits of the OLxW and OLxLx distributions with those of the Kumaraswamy complemen-
tary Weibull geometric (KCWG) [17], odd log-logistic exponentiated Weibull (OLLEW) [18], Lomax Weibull (LxW) [19],
Kumaraswamy Weibull (KwW) [20], transmuted complementary Weibull geometric (TCWG) [21], McDonald Weibull
(McW) [22], alpha logarithmic transformed Weibull (ALTW) [23], transmuted Lomax (TLx) [24], beta exponentiated Lomax
(BELx) [25], McDonald Lomax (McLx) [9], Kumaraswamy Lomax (KwLx) [9], Weibull Lomax (WLx) [12], transmuted Weibull
Lomax (TWLx) [13] and Burr X Lomax (BXLx) [26] distributions. Their densities (for x > 0) are provided in Appendix A.

8.2. Application 2: Time-to-failure of turbocharger data

The second data set refers to the time-to-failure (103 h) of turbocharger of one type of engine given in [27]. The data
consist of 40 observations. For these data, we compare the fits of the OLxLL distribution and those of other competitive
232 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237

Table 6
Goodness-of-fit statistics for bladder cancer data.
Model W∗ A∗ Model W∗ A∗
OLxW 0.0119 0.0748 OLxLx 0.0170 0.1015
KCWG 0.0125 0.0800 TLx 0.0195 0.1254
OLLEW 0.0156 0.1002 BELx 0.0253 0.1685
LxW 0.0217 0.1435 McLx 0.0253 0.1685
KwW 0.0450 0.2920 KwLx 0.0283 0.1858
TCWG 0.0427 0.3059 WLx 0.0314 0.2060
McW 0.0503 0.3298 TWLx 0.0337 0.2216
ALTW 0.0984 0.6111 BXLx 0.0543 0.3549

Table 7
MLEs and their standard errors for bladder cancer data.
Model Estimates
ALTW 0.1807 1.2795 0.0312
(α, β, λ) (0.2259) (0.2021) (0.0281)
OLxW 0.2990 30.0881 3.2867 0.5728
(α, β, a, b) (0.0025) (8.8590) (0.0590) (0.0015)
OLLEW 61.5820 0.6024 0.4629 3.5199
(α, β, γ , θ ) (156.93) (0.4109) (0.5823) (2.2381)
LxW 2.0716 2.7007 0.1666 1.4270
(α, β, a, b) (6.0×10−5 ) (0.3690) (0.0002) (0.1020)
KwW 0.2161 0.4588 4.1184 2.9406
(α, β, a, b) (0.2520) (0.5540) (6.2350) (8.7900)
TCWG 106.069 1.7115 0.2168 0.0095
(α, β, λ, γ ) (124.8) (0.099) (0.61) (0.0065)
KCWG 0.0600 0.5470 2.6804 1.31667 0.3969
(α, β, γ , a, b) (0.1187) (0.2817) (3.9610) (0.9567) (0.4623)
McW 0.1192 0.5582 4.0633 2.6036 0.0393
(α, β, a, b, c) (0.1090) (0.1780) (2.1110) (2.4520) (0.2020)
BELx 2.6937 1.4468 0.0773 0.5653 2.7032
(β, θ, λ, a, b) (4.8833) (4.1809) (0.1129) (0.9866) (8.1174)
McLx 0.8084 11.292 1.5060 4.1886 2.1045
(α, β, a, η, c) (3.5451) (18.295) (0.2837) (26.365) (3.1111)
TWLx 0.2010 5.4950 10.5705 1.5186 −0.0006
(α, β, a, b, λ) (0.1800) (5.4010) (21.344) (0.2970) (0.5050)
OLxLx 0.8092 13.0451 2.8727 3.5641
(α, β, a, b) (0.0280) (7.0490) (0.1140) (0.5850)
WLx 0.1102 6.9692 36.1153 1.5152
(α, β, a, b) (0.1810) (6.9280) (109.938) (0.2870)
KwLx 0.3903 12.295 1.5157 12.0549
(α, β, a, b) (1.228) (11.754) (0.266) (45.016)
TLx 4.0161 0.0524 −0.8442
(α, β, λ) (1.4869) (0.0258) (0.1394)
BXLx 0.2982 1.0200 0.9338
(α, β, θ ) (0.0511) (0.6640) (0.2499)

models, namely: the Lomax log-logistic (LxLL) [19], Kumaraswamy transmuted log-logistic (KwTLL) [28], Kumaraswamy
log- logistic (KwLL) [29], transmuted log-logistic (TLL) [30], McDonald log-logistic (McLL) [31], beta log-logistic (BLL) [32]
and generalized transmuted log-logistic (GTLL) [33] distributions. Their densities (for x > 0) are given in Appendix B.
Tables 6 and 8 list the values of W ∗ and A∗ for both data sets, whereas the MLEs of the model parameters and their
corresponding standard errors (in parentheses) are listed in Tables 7 and 9.
The plots of the fitted OLxW and OLxLx densities and other best eight fitted densities, for the cancer data, are displayed
in Fig. 5. The PP-plots of the fitted distributions are given in Fig. 7 . They indicate that the OLxW and OLxLx models provide
the best fits to these data and can be considered very competitive models to other distributions with positive supports.
The fitted OLxLL pdf and other fitted densities for the time-to-failure data are displayed in Fig. 6, whereas Fig. 8 displays
the PP-plots for these fitted models. These plots reveal that the OLxLL distribution gives the best fit to these data and that it
can be considered a very competitive model to other non-nested distributions.

9. Conclusions

In this paper, we propose a new family of distributions with two extra generator parameters, which can include as special
cases all classical continuous distributions. For any parent continuous distribution G, we define the so-called odd Lomax-G
family with two extra positive parameters. The new family can extend several widely known distributions and some of its
special models are discussed. We demonstrate that its density function is a linear combination of exponentiated-G densities.
 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237 233

Table 8
Goodness-of-fit statistics for time-to-failure data.
Model W∗ A∗
OLxLL 0.0799 0.5932
LxLL 0.1047 0.7461
KwTLL 0.2029 1.3255
KwLL 0.2072 1.3504
TLL 0.2160 1.3994
McLL 0.2894 1.8180
BLL 0.3068 1.9211
GTLL 0.5590 3.3694

Table 9
MLEs and their standard errors for time-to-failure data.
Model Estimates
TLL 6.2257 4.8416 0.0007
(α, β, λ) (0.9951) (0.6540) (0.8200)
OLxLL 71.1860 4.9519 13.7071 3.8888
(α, β, a, b) (1.3×10−3 ) (1.707) (1.8×10−3 ) (0.5200)
LxLL 104.8721 19.0401 10.2460 4.0872
(α, β, a, b) (0.0120) (4.1430) (5.1×10−4 ) (0.517)
KwLL 0.5882 0.7449 28.9123 77.0219
(α, β, a, b) (0.8820) (0.1010) (25.7920) (61.3480)
BLL 4.9462 0.7316 25.3193 22.3871
(α, β, a, b) (1.2120) (0.1210) (18.7250) (14.3040)
KwTLL 0.1739 0.6619 56.2427 118.6118 0.0006
(α, β, a, b, λ) (0.1300) (0.1060) (44.7200) (100.255) (1.402)
GTLL 0.8271 2.0222 34.47210 0.7329 4.2×10−6
(α, β, a, b, λ) (0.4090) (0.2030) (31.1840) (18.7070) (0.046)
McLL 17.3407 0.61280 32.4696 49.3263 0.6020
(α, β, a, b, c) (0.2880) (19.9640) (28.4240) (10.0180) (1.4040)

Fig. 5. The estimated OLxW and OLxLx pdfs and other estimated pdfs for bladder cancer data. (left panel) The estimated OLxW, KCWG, OLLEW, LxW and
KwW densities. (right panel) The estimated OLxLx, TLx, BELx, KwLx and McLx densities.

We obtain some mathematical properties of the new family, which include ordinary and incomplete moments, quantile
and generating functions, probability weighted moments, order statistics and other measures. The model parameters are
estimated by the methods of maximum likelihood and least squares. Monte Carlo simulation results are reported for both
methods. The usefulness of the new family is proved empirically by means of two real data sets.

Acknowledgment

The authors would like to thank the Editor in Chief and the anonymous referees for very careful reading and valuable
comments which greatly improved the paper.
234 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237

Fig. 6. The estimated OLxLL pdf and other estimated pdfs for time-to-failure data. (left panel) The estimated OLxLL, LxLL, KwTLL, KwLL and TLL densities.
(right panel) The estimated OLxLL, McLL, BLL and GTLL densities.

Fig. 7. PP-plots of the OLxW and OLxLx distributions and other competitive distributions.

Appendix A

In this appendix, we give the densities used in the first

{ application:
β β a−1
] }b−1
β a
e−(γ x) 1−e−(γ x) α a 1−e−(γ x)
[ ] [
β−1
KCWG: f (x) = α βγ ab(γ x)
a
β a+1
1 −[ β a
] ;
α+(1−α)e−(γ x)
[ ]
α+(1−α)e−(γ x)
]γ }θ −1
x β x β γ θ−1 x β
[ ] { [
θ βγ xβ−1 e−( α ) 1−e−( α ) 1− 1−e−( α )
OLLEW: f (x) = } }2 ;
x β ]γ θ
{
x β ]γ θ
[ { [
α β 1−e−( α ) + 1− 1−e−( α )
 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237 235

Fig. 8. PP-plots of the OLxLL distribution and other competitive distributions.

[ ]−α−1
LxW: f (x) = α bβ α ab xb−1 β + (ax)
b
;
β
( β
)a−1 { [ β
]a }b−1
KwW: f (x) = abβα β xβ−1 e−(α x) 1 − e−(α x) 1 − 1 − e−(α x) ;
β β −3 β
[ ] [ ]
TCWG: f (x) = αβγ (γ x)β−1 e−(γ x) α + (1 − α) e−(γ x) × α (1 − λ) − (α − αλ − λ − 1) e−(γ x) ;
β a−1
] }b−1
β c
β c αβ β
[ ] { [
McW: f (x) = B(a/c ,b) xβ−1 e−(α x) 1 − e−(α x) 1 − 1 − e−(α x) ;
β−1 −λx β
(α−
[ 1)λβ x ( e )] , α
ALTW: f (x) = β ̸= 1;
log(α) α−(α−1) 1−e−λx

TLx: f (x) = αβ(1 + β x)−α−1 1 − λ + 2λ(1 + β x)−α ;

[ ]

βθ λ
]aβ−1 { ]β }b−1
+ λx)−θ −1 1 − (1 + λx)−θ
(1 1 − 1 − (1 + λx)−θ
[ [
BELx: f (x) = B(a,b)
;
}η
α α
a− 1
{ c
c αβ α
[ ( ) ] [ ( ) ]
β β
McLx: f (x) = B(a/c ,η+1) (β + x)−α−1 1 − β+x 1 − 1 − β+x ;
[ a−1 ] { [ a b−1
]}
abα x −α−1 x −α x −α
( ) ( ) ( )
KwLx: f (x) = β
1+ β 1− 1+ β 1− 1− 1+ β ;
)bα−1 [ )−α ]b−1 [( )α ]b
1+ βx
( (
abα x x −a −1
WLx: f (x) = β
1+ β
1− 1+ β
e ;
]b )α ]b
)α
{ }
)bα−1 [( [ )−α ]b−1 [(
1+ βx −a 1+ βx −1
( (
abα −a −1
TWLx: f (x) = β
1+ x
β
e x
× 1− 1+ β 1 − λ + 2λe ;

)α−1 [( )α ] [( x )α ]2 { [( )α ]2 }θ −1
− 1+ βx −1
(
2θ α − 1+ β −1
BXLx: f (x) = β
1+ x
β
1+ x
β
−1 e 1−e .
The parameters of the above densities are positive real numbers, except α ∈ (0, 1) for the KCWG model, and |λ| ≤ 1 for

the TCWG, TLx and TWL models.

236 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237

Appendix B

In this appendix, we give the

[ densities
] used
{ in the[ second application:
( x )b −1
( )b −α−1]}
LxLL: f (x) = α bβ α a−b xb−1 1 + β + ln 1 + ax
a
;
( )β ]−1 a−1
{ }
[ ( )β ]−2 [
KwTLL: f (x) = a b β α −β xβ−1 1 + αx 1 − 1 + αx

( x )β ]−1 a−1 ( x )β ]−1 a ( x )β ]−1 a b−1

{ }{ } { [ ]{ }}
[ ( x )β ]−1 [ [ [
× 1 − λ + 2λ 1 + α 1+λ 1+ α × 1− 1+λ 1+ α 1− 1+ α ;
{ [ ]a }b−1
abβ
[ ( x )β ]−a−1
KwLL: f (x) = α aβ
xaβ−1 1 + α
1− 1− 1
;β
( ) 1+ αx
{ }
[ ( x )β ]−2 [ ( )β ]−1
TLL: f (x) = β α −β xβ−1 1 + α
1 − λ + 2λ 1 + αx ;

( x )β ]−1 c b−1
{ [ ]}
βc
( x )β a−1 [ ( x )β ]−a−1 [
McLL: f (x) = α B(a/c ,b) α
1+ α
× 1− 1− 1+ α ;

βα −aβ
[ ( x )β ]−a−b
BLL: f (x) = B(a,b)
xaβ−1 1 + α
;
{ }a−1 { { }b }
−β β−1
[ ( x )β ]−2 [ ( x )β ]−1 [ ( x )β ]−1
GTLL: f (x) = βα x 1+ α
1− 1+ α
× a (1 + λ) − λ (a + b) 1 − 1 + α
.
The parameters of the above densities are all positive real numbers except |λ| ≤ 1.

References

[1] J. Saran, N. Pushkarna, Moments of order statistics from doubly truncated Lomax distribution, J. Statist. Res. 33 (1999) 57–66.
[2] M.S. Moghadam, F. Yahmaei, M. Babanezhad, Inference for Lomax distribution under generalized order statistics, Appl. Math. Sci. 6 (2012) 5241–5525.
[3] A.M. Abd-Elfattah, A.H. Alharby, Estimation of Lomax distribution based on generalized probability weighted moments, J. King Abdulaziz Univ. 22
(2010) 171–184.
[4] A.H. Abd-Ellah, Comparison of estimates using record statistics from Lomax model: Bayesian and non-Bayesian approaches, J. Statist. Res. 3 (2006)
139–158.
[5] P. Nasiri, S. Hosseini, Statistical inferences for Lomax distribution based on record values (Bayesian and classical), J. Mod. Appl. Statist. Methods 11
(2012) 179–189.
[6] I.B. Abdul-Moniem, H.F. Abdel-Hameed, On exponentiated Lomax distribution, Int. J. Math. Arch. 3 (2012) 2144–2150.
[7] M.E. Ghitany, F.A. AL-Awadhi, L.A. Alkhalfan, Marshall–Olkin extended Lomax distribution and its applications to censored data, Comm. Statist. Theory
Methods 36 (2007) 1855–1866.
[8] R.C. Gupta, M.E. Ghitany, D.K. Al-Mutairi, Estimation of reliability from Marshall–Olkin extended Lomax distributions, J. Stat. Comput. Simul. 80 (2010)
937–947.
[9] A.J. Lemonte, G.M. Cordeiro, An extended Lomax distribution, Statistics 47 (2013) 800–816.
[10] G.M. Cordeiro, E.M.M. Ortega, B.V. Popovic, The gamma-Lomax distribution, J. Stat. Comput. Simul. 85 (2015) 305–319.
[11] B. Al-Zahrani, H. Sagor, The Poisson-Lomax distribution, Rev. Colombiana Estadst. 37 (2014) 223–243.
[12] M.H. Tahir, G.M. Cordeiro, M. Mansoor, M. Zubair, The Weibull–Lomax distribution: properties and applications, Hacet. J. Math. Stat. 44 (2015) 461–480.
[13] A.Z. Afify, Z.M. Nofal, H.M. Yousof, Y.M. El Gebaly, N.S. Butt, The transmuted Weibull Lomax distribution: properties and application, Pak. J. Statist.
Oper. Res. 11 (2015) 135–152.
[14] A. Alzaatreh, C. Lee, F. Famoye, A new method for generating families of distributions, Metron 71 (2013) 63–79.
[15] A.W. Marshall, I. Olkin, A new method for adding a parameter to a family of distributions with application to the exponential and Weibull families,
Biometrika 84 (1997) 641–652.
[16] E.T. Lee, J.W. Wang, Statistical Methods for Survival Data Analysis, third ed., Wiley, New York, 2003.
[17] A.Z. Afify, G.M. Cordeiro, N.S. Butt, E.M.M. Ortega, E.M.M.A.K. Suzuki, A new lifetime model with variable shapes for the hazard rate, Braz. J. Probab.
Stat. 31 (2017) 516–541.
[18] A.Z. Afify, M. Alizadeh, M. Zayed, T.G. Ramires, F. Louzada, The odd log–logistic exponentiated Weibull distribution: regression modeling, properties,
and applications, Iran. J. Sci. Technol. Trans. A: Sci. (2018) 1–16. http://dx.doi.org/10.1007/s40995-018-0524-x.
[19] G.M. Cordeiro, E.M.M. Ortega, B.V. Popovic, R. Pescim, The Lomax generator of distributions: properties, minification process and regression model,
Appl. Math. Comput. 247 (2014) 465–486.
[20] G.M. Cordeiro, E.M.M. Ortega, S. Nadarajah, The Kumaraswamy Weibull distribution with application to failure data, J. Franklin Inst. B 349 (2010)
11174–11197.
[21] A.Z. Afify, Z.M. Nofal, N.S. Butt, Transmuted complementary Weibull geometric distribution, Pak. J. Statist. Oper. Res. 10 (2014) 435–454.
[22] G.M. Cordeiro, E.M. Hashimoto, E.M.M. Ortega, The McDonald Weibull model, Statist. A J. Theoret. Appl. Statist. 48 (2014) 256–278.
[23] M. Nassar, A.Z. Afify, S. Dey, D. Kumar, A new extension of Weibull distribution: properties and different methods of estimation, J. Comput. Appl. Math.
336 (2018) 439–457.
[24] S.K. Ashour, M.A. Eltehiwy, Transmuted Lomax distribution, Amer. J. Appl. Math. Statist. 1 (2013) 21–127.
[25] M.E. Mead, On five-parameter Lomax distribution: properties and applications, Pak. J. Statist. Oper. Res. 12 (2016) 185–199.
[26] H.M. Yousof, A.Z. Afify, G.G. Hamedani, G. Aryal, The Burr X generator of distributions for lifetime data, J. Stat. Theory Appl. 16 (2017) 288–305.
[27] K. Xu, M. Xie, L.C. Tang, S.L. Ho, Application of neural networks in forecasting engine systems reliability, Appl. Soft Comput. 2 (2003) 255–268.
[28] A.Z. Afify, Z.M. Nofal, G.M. Cordeiro, H.M. Yousof, The Kumaraswamy transmuted-G family of distributions: properties and applications, J. Data Sci. 14
(2016) 245–270.
 G.M. Cordeiro et al. / Journal of Computational and Applied Mathematics 347 (2019) 222–237 237

[29] T.V.F. de Santana, E.M.M. Ortega, G.M. Cordeiro, G.O. Silva, The Kumaraswamy log–logistic distribution, J. Stat. Theory Appl. 3 (2012) 265–291.
[30] D.C.T. Granzotto, F. Louzada, The transmuted log–logistic distribution: modeling, inference and an application to a polled tabapua race time up to first
calving data, Comm. Statist. Theory Methods 44 (2015) 3387–3402.
[31] M.H. Tahir, M. Mansoor, M. Zubair, G. Hamedani, McDonald log–logistic distribution with an application to breast cancer data, J. Stat. Theory Appl. 13
(2014) 65–82.
[32] A.J. Lemonte, The beta log–logistic distribution, Braz. J. Probab. Stat. 28 (2014) 313–332.
[33] Z.M. Nofal, A.Z. Afify, H.M. Yousof, G.M. Cordeiro, The generalized transmuted-G family of distributions, Comm. Statist. Theory Methods 46 (2017)
4119–4136.

View publication stats