TEXT SIZE

CrossRef (0)
Generalized half-logistic Poisson distributions

aDepartment of Mathematical Science, Bayero University Kano, Nigeria
Correspondence to: Department of Mathematical Science, Bayero University Kano, Nigeria. E-mail: mmmahmoud12@sci.just.edu.jo, mmuhammad.mth@buk.edu.ng
Received December 27, 2016; Revised April 7, 2017; Accepted June 3, 2017.
Abstract

In this article, we proposed a new three-parameter distribution called generalized half-logistic Poisson distribution with a failure rate function that can be increasing, decreasing or upside-down bathtub-shaped depending on its parameters. The new model extends the half-logistic Poisson distribution and has exponentiated half-logistic as its limiting distribution. A comprehensive mathematical and statistical treatment of the new distribution is provided. We provide an explicit expression for the rth moment, moment generating function, Shannon entropy and Rényi entropy. The model parameter estimation was conducted via a maximum likelihood method; in addition, the existence and uniqueness of maximum likelihood estimations are analyzed under potential conditions. Finally, an application of the new distribution to a real dataset shows the flexibility and potentiality of the proposed distribution.

Keywords : half-logistic Poisson, moments, entropy, maximum likelihood estimates
1. Introduction

Lifetime data may exhibit a decreasing, increasing, or a bathtub failure rate function when modeling and analyzing random phenomena. This arises in several areas of studies, such as biomedical studies, reliability, actuarial science, computer science, demography, and engineering. There are several lifetime models that have been used successfully for to analyze lifetime data in practical applications such as exponential, Weibull, Gompertz, generalized exponential, and half-logistic. For example, exponential distribution can accommodate lifetime data with a decreasing density despite having only a constant failure rate function. Gompertz distribution has a decreasing and unimodal density but has an increasing failure rate function. However, these distributions are unable to accommodate lifetime data with a non-monotone failure rate such as the bathtub or upside-down bathtub. Many researchers have attempted to provide several methods to generate new lifetime models with the ability to fit data with monotone or non-monotone failure rate to overcome these problems. These methods include the generalization of a distribution by exponentiation procedure.

Method of exponentiation is an of the important and commonly used techniques to add a parameter to a lifetime model, the new model becomes more flexible and can accommodate both monotones as well as non-monotone failure rate functions. For example, Kuş (2007) proposed en exponential Poisson (EP) distribution that possesses a decreasing failure rate function; however, Barreto-Souza and Cribari-Neto (2009) proposed generalized exponential Poisson (GEP) by exponentiation of the EP. The GEP distribution can accommodate decreasing, increasing and upside down bathtub failure rates. Similarly, Adamidis and Loukas (1998) introduced exponential geometric distribution, while Silva et al. (2010) come up with the generalized exponential geometric distribution. Tahmasbi and Rezaei (2008) introduced exponential-logarithmic distribution similarly, Pappas et al. (2015) proposed a generalized exponential-logarithmic. Silva and Cordeiro (2015) introduced a new class of distribution called the BurrXII Poisson distribution and recently Muhammad (2016a) proposed generalized BurrXII Poisson distribution. To read more about exponentiated distributions see Ali et al. (2007) and Raja and Mir (2011). A new two-parameter distribution known as the half-logistic Poisson (HLP) distribution was also introduced by Muhammad and Yahaya (2017) using the procedure followed by Kuş (2007) and Chung and Kang (2014).

The cumulative distribution function (cdf) of the HLP distribution is given by

$G(x)=(1-e-λ(1-e-αx1+e-αx)1-e-λ),$

with x > 0, α > 0, and λ > 0. The HLP distribution can accommodate data with increasing or decreasing hazard functions.

The first motivation of this study is to propose a new three-parameter probability distribution with increasing, decreasing and upside down bathtub-shaped failure functions. The second motivation of this study is to find the association of the density functions and hazard rate functions in practical applications. Last, we are motivated because the exponentiation method provides additional flexibility to a model especially in both the density and failure rate functions. The HLP distribution cannot fit data set with non-monotone failure rate functions; therefore, we hope that the new distribution that is the generalized half-logistic Poisson (GHLP) will provide solutions to many problems in various fields in practical applications.

The rest of the paper is arranged as follows, in Section 2 we provide the density of the GHLP and consider some important mathematical and statistical properties. Section 3 discusses the maximum likelihood estimate (MLE) and a simulation study. Section 4 provides applications of the new model to a real data set Section 5 provides the conclusions.

2. The proposed model

The cumulative distribution function of the GHLP distribution with parameters α, β, λ > 0 is given by

$F(x)=(1-e-λ(1-e-αx1+e-αx))β(1-e-λ)-β,$

where the corresponding probability density and hazard rate functions are given by

$f(x)=2 αβλe-αx(1-e-λ)β(1+e-αx)2(1-e-λ(1-e-αx1+e-αx))β-1e-λ(1-e-αx1+e-αx),$$h(x)=2 αβλe-αx (1-e-λ(1-e-αx1+e-αx))β-1e-λ(1-e-αx1+e-αx)(1+e-αx)2 ((1-e-λ)β-(1-e-λ(1-e-αx1+e-αx))β),$

respectively. The limiting distribution given by (2.1) when λ → 0+ is limλ→0+F(x) = ((1 − eαx)/(1 + eαx))β for β > 0, which is the cdf of the exponentiated half-logistic distribution.

### Theorem 1

The probability density function (pdf) given by (2.2) is decreasing function for β ≤ 1.

Proof

We obtained the first derivative of log f (x) as

$∂ log f(x)∂x=-α-2α e-αx1-e-αx+2αλ(β-1) e-αxe-λ(1-e-αx1+e-αx)(1+e-αx)2 (1-e-λ(1-e-αx1+e-αx))-2αλe-αx(1+e-αx)2,$

thus, for β ≤ 1, log f (x) < 0.

Figure 1 below provide the plot of the pdf ( f (x)) of the GHLP and illustrated that the pdf can be unimodal function for β > 1.

The limiting behavior of the pdf given by (2.2) are: for β < 1, limx→0f (x) = ∞; for β = 1, limx→0f (x) = αλ/{2(1 − eλ)}; for β > 1, limx→0f (x) = 0 and limx→∞f (x) = 0 for all β > 0. The limiting behavior of the hazard rate function given by (2.3) are: for β < 1, limx→0h(x) = ∞; for β = 1, limx→0h(x) = αλ/{2(1 − eλ)} and for β > 1, limx→0h(x) = 0. Figure 1 below shows that the hazard rate function (h(x)) of the GHLP given by (2.3) can be decreasing, increasing or unimodal functions.

The pth-quantile function of the GHLP can easily be derived by inverting (2.1) as

$ξ(p)=-α-1 log (λ+log (1-p1β(1-e-λ))λ-log (1-p1β(1-e-λ))),$

therefore, we can find the numerical values of the median and other percentiles of X with (2.5), also the median of X can be obtained as ξ(0.5). Moreover, equation (2.5) can be used to generate random data distributed GHLP(α, β, λ) by setting p ~ U(0, 1), where U(0, 1) is the uniform distribution.

### 2.1. Moments

Here, we provide the following lemma which is very useful in computations of several important properties of the GHLP. first we recall that, Bkd(a, b) = (∂k+dB(a, b))/(∂akbd) for k+a > 0, b+d > 0, where $B(a,b)=∫01ua-1(1-u)b-1du$ is a beta function.

Lemma 1

For ȷn, n = 1, 2, …, 6, and ȷn ∈ ℝ, let,

$J(ȷ1,ȷ2,ȷ3,ȷ4,ȷ5,ȷ6)=∫0∞xȷ1 e-ȷ2x(1-e-αx)ȷ3(1+e-αx)ȷ4(1-e-λ(1-e-αx1+e-αx))ȷ5 e-ȷ6(1-e-αx1+e-αx)dx,$

then,

$J(ȷ1,ȷ2,ȷ3,ȷ4,ȷ5,ȷ6)=∑i=0∞∑k,l=0∞ξi,k,l B0ȷ1(ȷ3+k+1,ȷ2α+l),$

where$ξi,k,l=(ȷ5i)(-(ȷ4+k)l){(-1)i+k+ȷ1(λi+ȷ6)k}/{αȷ1+1k!}$.

Proof

See Appendix A.

Table 1 provide some numerical values of J( ȷ1, ȷ2, ȷ3, ȷ4, ȷ5, ȷ6) for some various parameter values computed using R software.

Various characteristics and features of a distribution can be analyzed through its moments such as mean, variance, moment generating function, etc. If X follows the GHLP distribution, then, the rth moment of X can be obtained by considering Lemma 1 as follows:

$E(Xr)=2αβλ(1-e-λ)β∫0∞xr e-αx(1+e-αx)2(1-e-λ(1-e-αx1+e-αx))β-1e-λ(1-e-αx1+e-αx)dx,$

thus,

$E(Xr)=2αβλ(1-e-λ)βJ(r,α,0,2,β-1,λ).$

Therefore, the moments of GHLP can easily be computed using mathematical software such as MATLAB, MATHEMATICA, and R. Figure 2 provide the plots of the mean μ and variance σ2 of the GHLP for α = 1.5, it is clear that both the mean and variance are increasing as β increases and decreasing as λ increases.

The moment generating function of the GHLP distribution can be computed directly using MX(t) = E(etX) which can be expanded to

$MX(t)=∑r=0∞trr!E(Xr),$

thus, by putting (2.9) in (2.11) we have

$MX(t)=∑r=0∞2αβλtr(1-e-λ)βr!J(r,α,0,2,β-1,λ).$

One of the alternative measures for the skewness and kurtosis of a distribution are the Bowley skewness (B) and Moor’s kurtosis (M) defined by B = [ξ(3/4) + ξ(1/4) − 2 ξ(2/4)]/[ξ(3/4) − ξ(1/4)] and M = [ξ(3/8) − ξ(1/8) + ξ(7/8) − ξ(5/8)]/[ξ(6/8) − ξ(2/8)], respectively, where ξ(·) is given by (2.5). Figure 3 is the plots of the Bowley skewness and Moor’s kurtosis of the GHLP for α > 0. It is clear that both the skewness and kurtosis are decreasing in β and unimodal function in λ.

2.2. Entropy

Entropy is defined as a measure of uncertainty of a random variable. Here, we consider the two most important entropies known as the Shannon and Rényi entropies. The Shannon entropy is defined by E[− log f (x)]. For a random variable X with GHLP, the Shannon entropy can be computed by considering the Proposition 1 as follows.

### Proposition 1

Let X be a random variable with pdf given by (2.2), then,

$E(log(1+e-αX))=2αβλ(1-e-λ)β∂∂tJ(0,α,0,2-t,β-1,λ)∣t=0,$$E(log (1-e-λ(1-e-αX1+e-αX)))=2αβλ(1-e-λ)β∂∂tJ(0,α,0,2,β+t-1,λ)∣t=0,$$E(1-e-αX1+e-αX)=2αβλ(1-e-λ)βJ(0,α,1,3,β-1,λ).$

where J(·, ·, ·, ·, ·, ·) is given by (2.7).

Hence,

$E[-log f(X)]=-log (2αβλ(1-e-λ)β)+αE(X)+4αβλ(1-e-λ)β∂∂tJ(0,α,0,2-t,β-1,λ)∣t=0-2αβλ(β-1)(1-e-λ)β∂∂tJ(0,α,0,2,β+t-1,λ)∣t=0+2αβλ2(1-e-λ)βJ(0,α,1,3,β-1,λ).$

The Rényi entropy of a random variable X is defined by $IR(ρ)=1/(1-ρ)log[∫0∞f(x)ρdx]$, where ρ > 0 and ρ ≠ 1. The Rényi entropy of X that has GHLP is obtain as follows.

$∫0∞fρ(x) dx=∫0∞2ραρβρλρe-αρxe-λρ(1-e-αx1+e-αx)(1-e-λ)βρ(1+e-αx)2ρ(1-e-λ(1-e-αx1+e-αx))ρ(β-1)dx,$

by applying Lemma 1 the integral become

$∫0∞fρ(x) dx=2ραρβρλρ(1-e-λ)βρJ(0,αρ,0,2ρ,ρ(β-1),λρ).$

Hence, the Rényi entropy of X is

$IR(ρ)=11-ρlog [2ραρβρλρ(1-e-λ)βρJ(0,αρ,0,2ρ,ρ(β-1),λρ)].$
3. Estimation

In this section, we discuss MLEs, one of the most popular and common methods used in inference. In this method the approximate MLEs are obtained either analytically or numerically using some mathematical packages. Let Xi (i = 1, 2, …, n) be a random sample of size n from the GHLP distribution with observed values x1, x2, x3, …, xn. The log likelihood function, i.e. log (θ) for complete data set of the GHLP distribution is given by

$log ℓ(θ)=n log 2+n log α+n log β+n log λ-nβ log(1-e-λ)-α∑i=1nxi-2∑i=1nlog(1+e-αxi)+(β-1)∑i=1nlog (1-e-λ(1-e-αxi1+e-αxi))-λ∑i=1n(1-e-αxi1+e-αxi).$

Hence, the MLEs of θ = (α, β, λ)T, say θ̂ = (α̂, β̂, λ̂)T is the solution of the nonlinear equations (3.2)(3.4).

$∂ℓ∂α=nα-∑i=1nxi-∑i=1n2xi e-αxi1+e-αxi+∑i=1n2(β-1)λ xie-αxie-λ(1-e-αxi1+e-αxi)(1+e-αxi)2(1-e-λ(1-e-αxi1+e-αxi))-∑i=1n2λxie-αxi(1+e-αxi)2=0,$$∂ℓ∂β=nβ-n log (1-e-λ)+∑i=1nlog (1-e-λ(1-e-αxi1+e-αxi))=0,$$∂ℓ∂λ=nλ-nβe-λ(1-e-λ)+(β-1)λ∑i=1n(1-e-αxi1+e-αxi) e-λ(1-e-αxi1+e-αxi)(1-e-λ(1-e-αxi1+e-αxi))-∑i=1n(1-e-αxi1+e-αxi)=0.$

For the interval estimate and hypothesis tests of the parameters we required J(θ) the 3 × 3 Fisher information matrix defined by J(θ) = −{∂2(log (θ))/∂θθT }. The approximate of the MLEs of θ, the θ̂, can be approximated as N3(0, J(θ̂)−1) under usual conditions for parameters in the interior of the parameter space, but not on the boundary. The asymptotic distribution of $n(θ^-θ)$ is N3(0, J(θ̂)−1), where J(θ̂) is the unit information matrix evaluated at θ̂, which can be used to construct the approximate confidence interval for each parameter. A 100(1 − ε)% asymptotic confidence interval for each parameter θr is given by $ACIr=(θ^r-Zɛ/2I^rr,θ^r+Zɛ/2I^rr)$, where Irr is the (r, r) diagonal element of In(θ)−1 for r = 1, 2, 3 and Zε/2 is the quantile (1 − ε/2) of the standard normal distribution. The elements of J(θ) can be obtained from the author under request. The existence and uniqueness of MLEs of a probability models based on some certain sufficient conditions have been considered in various literature by many researchers, the existence and uniqueness of maximum likelihood estimators of the EP was analyzed by Kuş (2007), for the exponential geometric by Adamidis and Loukas (1998), Generalized exponential-power series by Mahmoudi and Jafari (2012), extended exponentialgeometric by Adamidis at el. (2005), recently generalized BurrXII Poisson by Muhammad (2016a) and the complementary exponentiated BurrXII-Poisson by Muhammad (2017) among others. The following theorems provide the existence and uniqueness of the MLEs of the GHLP under some possible conditions with the proofs provided in Appendix B.

### Theorem 2

Let g1(α; β, λ, xi) denote the function on the right hand side of the equation (3.2) where β and λ are the true values of the parameters, then, the equation g1(α; β, λ, xi) = 0 has at least one root for β ≠ 1 and for β = 1 the root lies in the interval$(n/{(2+λ/2)∑i=1nxi},n/∑i=1nxi)$.

### Theorem 3

Let g2(β; α, λ, xi) denote the function on the right hand side of the equation (3.3) where α and λ are the true values of the parameters, then, the equation g2(β; α, λ, xi) = 0 has at most one root for$log(1-e-λ)>n-1∑i=1nlog (1-e-λ{(1-e-αxi)/(1+e-αxi))}$and is unique.

### Theorem 4

Let g3(λ; α, β, xi) denote the function on the right hand side of the equation (3.4) where α and β are the true values of the parameters, then, the equation g3(β; α, λ, xi) = 0 has at least one root for$β>1/2+(1/n)∑i=1n{(1-e-αxi)/(1+e-αxi)}$.

### 3.1. Simulation study

Here, we evaluate the performance of the MLEs given by equations (3.2)(3.4) depending on sample size n for the GHLP distribution. In this process we generated 10,000 samples of size n = 20, 30, 40, 50, 100, and 150 from the GHLP distribution for some various values of α, β, and λ. MLEs are obtained by solving the nonlinear equations (3.2)(3.4) using mlninb in R. The MLEs of α̂, β̂, and λ̂ and their standard deviations sd( α̂), sd(β̂), and sd(λ̂) of the parameters are given in Table 2 below. The results show that each MLE converges to its true value in all cases when the sample size increases and the standard deviations of the MLEs decrease as the sample size increases.

4. Real data illustration

In this section, we provide an application of the GHLP distribution to a real data set. We used the Akaike information criterion (AIC), consistent Akaike information criterion (CAIC), and Kolmogorov-Smirnov (K-S) test to compare the GHLP and some other existing distributions. The model with the smallest values of these measures fit the data better than the other distributions. The competing distributions are:

• The HLP distribution with cdf given by (1.1).

• Olapade-generalized half-logistic by Olapade (2014) with F(x) = 1 − (2β(1 + ex/α)β).

• Power-half-logistic by Krishnarani (2016) with F(x) = 1 − 2(1 + eαxβ )−1.

• The generalized half-logistic with F(x) = (1 − eαx)β(1 + eαx)β. The distribution appears in the study of its estimation procedures by Arora et al. (2010), Kang and Seo (2011), Seo et al. (2012, 2013), and Kantam et al. (2013), where some of its important properties such as the rth-moments, probability weighted moments, and Shannon and Rényi entropies can be obtained from Cordeiro et al. (2014).

• Poisson-half-logistic by Abdel-Hamid (2016) with F(x) = (eλ{(1−eαx)/(1+eαx)} − 1)(eλ − 1)−1.

• Generalized exponential by Mudholkar and Srivastava (1993) with F(x) = (1 − exp(−αx))β.

• The EP by Kuş (2007) with F(x) = (eλeαxeλ)(1 − eλ)−1.

• Poisson-odd exponential uniform by Muhammad (2016b) with F(x) = [1−eλ(1−eα(x/(βx)))]/(1−eλ).

• Nadarajah and Haghighi (2011) exponential type (NH) with F(x) = 1 − e1−(1+λx)α.

• The half-logistic distribution with cdf as F(x) = (1 − eαx)(1 + eαx)−1.

The data set is the remission times (in months) of a random sample of 128 bladder cancer patients provided by Lee and Wang (2003).

In Figure 4 the TTT (total time on test)-plot show that the data has upside-down bathtub failure rate function and GHLP distribution has the ability to accommodate upside-down bathtub failure rate curve.

We estimate the unknown parameters of each model by the method of maximum likelihood. The numerical values of the log-likelihood ( (θ)), AIC, CAIC, K-S, and its p-value obtained are presented in Table 3 below. We also used the muhaz package in R software to obtain the empirical hazard function of the given data set and then fitted with the estimated hazard function of the GHLP obtained using the MLEs in

Table 3 indicates that our proposed model the GHLP distribution has the smallest values of the AIC, CAIC, and K-S; thus, GHLP provide better fit than the other competing models. Figure 5 provides the plots of the (i) histogram and estimated density (ii) empirical cdf and estimated cdf of the

GHLP distribution. Figure 6 shows (i) empirical and estimated hazard functions (ii) quantile-quantile of the GHLP distribution for the given data set.

5. Conclusions

We introduced a new three-parameter lifetime distribution with increasing, decreasing and upside-down bathtub-shaped hazard rate functions. We also provide explicit expressions for the rth ordinary moment, moment generating function, Shannon and Rényi entropies. The estimation of the model parameters was conducted by the maximum likelihood method. The practical significance and applicability of the new distribution are demonstrated in an application to real data, which shows that the GHLP performs better than other existing distributions in terms of fit.

Appendix A

### Proof of Lemma 1

For (2.6), by applying generalized binomial expanding on

$(1-e-λ(1-e-αx1+e-αx))ȷ5=∑i=0∞(ȷ5i)(-1)ie-λi(1-e-αx1+e-αx),$

then expanding

$e-(ȷ6+λi)(1-e-αx1+e-αx)=∑k=0∞(-1)k (ȷ6+λi)kk!(1-e-αx1+e-αx)k$

we get

$J(ȷ1,ȷ2,ȷ3,ȷ4,ȷ5,ȷ6)=∑i=0∞∑k=0∞(ȷ5i)(-1)i+k(λi+ȷ6)kk!∫0∞xȷ1e-ȷ2x(1-e-αx)ȷ3+k(1+e-αx)ȷ4+kdx.$

Let u = 1 − eαx. By the generalized binomial expansion in the denominator we obtain

$J(ȷ1,ȷ2,ȷ3,ȷ4,ȷ5,ȷ6)=∑i=0∞∑k,l=0∞(ȷ5i) (-(ȷ4+k)l)(-1)i+k+ȷ1(λi+ȷ6)kαȷ1+1k!×∫01lnȷ1(1-u) (1-u)ȷ2α+l-1uȷ3+kdu$

thus, we have (2.7).

Appendix B

### Proof of Theorem 2

Let g1(α; β, λ, xi) be the right hand side of (3.2).

For β = 1, let

$w1=-2∑i=1nxie-αxi1+e-αxi-2λ∑i=1nxie-αxi(1+e-αxi)2$

then,

$limα→0w1=-∑i=1nxi-λ2∑i=1nxi and limα→∞w1=0,g1(α;β,λ,xi)=nα-∑i=1nxi+w1>nα-∑i=1nxi+limα→0w1g1(α;β,λ,xi)>0, if α

On the other side

$g1(α;β,λ,xi)=nα-∑i=1nxi+w1

hence,

$g1(α;β,λ,xi)<0, if α>n∑i=1nxi$

thus, the root of g1(α; β, λ, xi) = 0 lies in the interval $(n/{(2+λ/2)∑i=1nxi},n/∑i=1nxi)$.

For β ≠ 1, limα→0g1 = ∞ and $limα→∞ g1=-∑i=1nxi<0$, hence, g1(α; β, λ, xi) is a monotone decreasing function from positive to negative; thus, g1(α; β, λ, xi) = 0 has at least one root.

### Proof of Theorem 3

Consider that limβ→0g2 = ∞, then we show that limβ→∞g2 < 0. But $limβ→∞ g2=-n log(1-e-λ)+∑i=0nlog (1-e-λ((1-e-αxi)/(1+e-αxi)))$; thus, limβ→∞g2 < 0 only if $log(1-e-λ)>n-1∑i=0nlog (1-e-λ((1-e-αxi)/(1+e-αxi)))$. To prove the uniqueness we show that g2 is a decreasing function, that is $g2′<0$ and $g2′=-n/β2<0$.

### Proof of Theorem 4

We start with $limλ→∞ g3=-∑i=1n((1-e-αxi)/(1+e-αxi))<0$; therefore, we show that limλ→0g3 > 0. And $limλ→0 g3=nβ-n/2-∑i=1n((1-e-αxi)/(1+e-αxi))$, thus, limλ→0g3 > 0 only if $β>1/2+(1/n)∑i=1n((1-e-αxi)/(1+e-αxi))$.

Figures
Fig. 1. Plots of the probability density function and hazard function of the generalized half-logistic Poisson for different values of parameters.
Fig. 2. Plots of the mean and variance of generalized half-logistic Poisson for α = 1.5.
Fig. 3. Plots of the skewness and kurtosis of generalized half-logistic Poisson for α = 0.5.
Fig. 4. TTT-plot for the remission times (in months) data. TTT = total time on test.
Fig. 5. Plots of the (i) histogram and estimated density (ii) empirical and estimated cdfs of generalized half-logistic Poisson (GHLP) distribution for the remission times (in months) of bladder cancer patients. cdf = cumulative distribution function.
Fig. 6. Plots of the (i) empirical hazard and estimated hazard (ii) quantile-quantile function of generalized half-logistic Poisson (GHLP) distribution for the remission times (in months) of bladder cancer patients.
TABLES

### Table 1

Numerical values of J(ȷ1, ȷ2, ȷ3, ȷ4, ȷ5, ȷ6) for some values of ȷ1, ȷ2, ȷ3, ȷ4, ȷ5, and ȷ6

ȷ1 ȷ2 ȷ3 ȷ4 ȷ5 ȷ6 J( ȷ1, ȷ2, ȷ3, ȷ4, ȷ5, ȷ6) ȷ1 ȷ2 ȷ3 ȷ4 ȷ5 ȷ6 J( ȷ1, ȷ2, ȷ3, ȷ4, ȷ5, ȷ6)
0.5 1.2 2.1 2.3 0.6 3.0 0.01595 0.5 0.2 0.1 0.3 0.6 0.3 7.24441
0.6 2.2 1.1 4.3 3.6 1.3 0.00392 0.1 2.2 4.1 3.3 3.6 1.5 0.00065
3.0 2.0 4.1 3.3 6.0 5.0 0.00150 1.0 6.0 1.0 3.0 2.0 1.0 0.00024
1.0 1.0 1.0 1.0 1.0 1.0 0.23417 1.5 1.0 1.0 2.5 1.0 0.5 0.34307
1.5 7.0 1.0 2.5 1.0 1.5 0.00018 10 2.0 2.0 5.0 1.0 5.0 14.2023
10 2.0 12 5.0 11 15 0.00123 8.0 5.0 2.0 1.0 9.0 1.0 0.00240

### Table 2

MLEs and standard deviations for some various values of parameters

Sample size n Actual values Estimated values Standard deviations

α β λ α̂ β̂ λ̂ sd(α̂) sd(β̂) sd(λ̂)
20 0.6 1.0 0.2 0.3484 0.6445 0.4752 0.2779 0.7451 1.0270
1.0 1.2 1.0 1.0751 1.4988 1.3013 0.3698 0.8655 1.4253
0.1 0.2 0.5 0.3826 0.1842 0.8930 0.3908 0.0778 6.1164
0.3 0.7 0.4 0.2615 0.6311 0.7950 0.1282 0.4109 3.7589
2.5 1.1 2.1 3.2405 1.2652 1.7064 1.2870 0.6377 4.2860
1.0 1.0 1.0 1.0914 1.2013 1.2527 0.3901 0.5492 1.3846
0.2 0.5 1.5 0.2606 0.5550 1.2747 0.1173 0.1842 3.9691

30 0.6 1.0 0.2 0.7988 0.6666 0.1397 0.2009 0.4976 0.5541
1.0 1.2 1.0 1.0318 1.3712 1.3004 0.3149 0.4991 1.3962
0.1 0.2 0.5 0.3446 0.1829 0.6371 0.3794 0.0691 2.8381
0.3 0.7 0.4 0.2514 0.6060 0.7650 0.1140 0.3707 3.4513
2.5 1.1 2.1 3.0149 1.1743 1.7622 1.0964 0.3927 2.9208
1.0 1.0 1.0 1.0466 1.1260 1.2523 0.3296 0.3831 1.3701
0.2 0.5 1.5 0.2446 0.5318 1.2913 0.0969 0.1363 1.3302

40 0.6 1.0 0.2 0.7948 0.6463 0.1307 0.1986 0.4601 0.5467
1.0 1.2 1.0 1.0128 1.3138 1.2735 0.2855 0.3892 1.3833
0.1 0.2 0.5 0.3207 0.1835 0.5788 0.3691 0.0661 2.0529
0.3 0.7 0.4 0.2505 0.5911 0.7099 0.1088 0.3443 2.5766
2.5 1.1 2.1 2.8909 1.1424 1.8507 1.0181 0.3141 1.5175
1.0 1.0 1.0 1.0285 1.0942 1.2641 0.3036 0.3070 1.3592
0.2 0.5 1.5 0.2326 0.5216 1.3513 0.0858 0.1104 1.3274

50 0.6 1.0 0.2 0.7946 0.6359 0.1269 0.1964 0.4450 0.5423
1.0 1.2 1.0 1.0037 0.6211 0.4066 0.1727 0.5680 1.0623
0.1 0.2 0.5 0.3088 0.1833 0.5842 0.3647 0.0649 1.2393
0.3 0.7 0.4 0.2477 0.5806 0.6368 0.1044 0.3362 1.6333
2.5 1.1 2.1 2.8522 1.1244 1.8733 0.9546 0.2751 1.5140
1.0 1.0 1.0 1.0144 1.0723 1.2502 0.2831 0.2647 1.3529
0.2 0.5 1.5 0.2270 0.5144 1.3846 0.0798 0.0969 1.3273

100 0.6 1.0 0.2 0.7941 0.6112 0.0926 0.1912 0.4106 0.4077
1.0 1.2 1.0 0.9978 0.6444 0.3096 0.1346 0.4431 0.8330
0.1 0.2 0.5 0.3016 0.1764 0.5559 0.3691 0.0669 0.6401
0.3 0.7 0.4 0.2461 0.5641 0.5677 0.0952 0.3104 1.3488
2.5 1.1 2.1 2.6656 1.1021 2.0777 0.8103 0.1869 1.5002
1.0 1.0 1.0 0.9895 1.0286 1.2264 0.2363 0.1790 1.2841
0.2 0.5 1.5 0.1559 0.2888 0.6507 0.0743 0.2057 1.1702

150 0.6 1.0 0.2 0.7960 0.6006 0.0718 0.1882 0.3997 0.3421
1.0 1.2 1.0 0.9969 0.6429 0.3005 0.1213 0.4317 0.8044
0.1 0.2 0.5 0.3456 0.1667 0.4991 0.3974 0.0720 0.5946
0.3 0.7 0.4 0.2470 0.5562 0.4975 0.0921 0.3005 0.8660
2.5 1.1 2.1 2.5933 1.0939 2.1501 0.7259 0.1521 1.4292
1.0 1.0 1.0 0.9887 1.0200 1.1904 0.2093 0.1483 1.1919
0.2 0.5 1.5 0.1514 0.2810 0.6537 0.0689 0.2018 1.1586

### Table 3

MLEs, (θ), AIC, CAIC, K-S and p-value for the remission times (in months) of bladder cancer patients.

Model α β λ (θ) AIC CAIC K-S p-value
GHLP 0.0642 1.3909 4.4696 −410.24 826.48 820.53 0.0427 0.9737
OGHL 0.9217 0.1250 - −412.18 828.36 824.40 0.0661 0.6304
PwHL 0.8880 0.2015 - −415.10 834.19 830.22 0.0751 0.4652
HLP 0.0574 - 3.8836 −413.17 830.34 826.37 0.0963 0.1863
PHL 0.8880 - 0.2015 −415.10 834.19 830.22 0.0761 0.4489
GHL 0.1440 0.9527 - −416.64 837.27 833.30 0.0950 0.1994
GE 0.1212 1.2180 - −413.08 830.16 826.19 0.0725 0.5113
EP 0.3341 4.3342 −417.04 835.61 831.64 0.9533 0.0000
POEU 1.5503 100 6.7528 −415.15 834.30 830.33 0.0808 0.3742
NH 0.3341 - 4.3342 −417.04 838.09 820.53 0.5709 0.0000
HL 0.1479 - - −416.73 835.45 833.47 0.0989 0.1631

References
1. Abdel-Hamid, AH (2016). Properties, estimations and predictions for a Poisson-half-logistic distribution based on progressively type-II censored samples. Applied Mathematical Modelling. 40, 7164-7181.
2. Adamidis, K, Dimitrakopoulou, T, and Loukas, S (2005). On an extension of the exponential-geometric distribution. Statistics & Probability Letters. 73, 259-269.
3. Adamidis, K, and Loukas, S (1998). A lifetime distribution with decreasing failure rate. Statistics & Probability Letters. 39, 35-42.
4. Ali, MM, Pal, M, and Woo, JS (2007). Some exponentiated distributions. Communications for Statistical Applications and Methods. 14, 93-109.
5. Arora, SH, Bhimani, GC, and Patel, MN (2010). Some results on maximum likelihood estimators of parameters of generalized half logistic distribution under Type-I progressive censoring with changing failure rate. International Journal of Contemporary Mathematical Sciences. 5, 685-698.
6. Barreto-Souza, W, and Cribari-Neto, F (2009). A generalization of the exponential-Poisson distribution. Statistics & Probability Letters. 79, 2493-2500.
7. Chung, Y, and Kang, Y (2014). The exponentiated Weibull-geometric distribution: properties and estimations. Communications for Statistical Applications and Methods. 21, 147-160.
8. Cordeiro, GM, Alizadeh, M, and Ortega, EM (2014). The exponentiated half-logistic family of distributions: properties and applications. Journal of Probability and Statistics. 2014, 1-21.
9. Kang, SB, and Seo, JI (2011). Estimation in an exponentiated half logistic distribution under progressively type-II censoring. Communications for Statistical Applications and Methods. 18, 657-666.
10. Kantam, RRL, Ramakrishna, V, and Ravikumar, MS (2013). Estimation and testing in type I generalized half logistic distribution. Journal of Modern Applied Statistical Methods. 12, 198-206.
11. Krishnarani, SD (2016). On a power transformation of half-logistic distribution. Journal of Probability and Statistics. 2016, 1-10.
12. Kuş, C (2007). A new lifetime distribution. Computational Statistics & Data Analysis. 51, 4497-4509.
13. Lee, ET, and Wang, JW (2003). Statistical Methods for Survival Data Analysis. New York: Wiley
14. Mahmoudi, E, and Jafari, AA (2012). Generalized exponential-power series distributions. Computational Statistics & Data Analysis. 56, 4047-4066.
15. Mudholkar, GS, and Srivastava, DK (1993). Exponentiated Weibull family for analyzing bathtub failure-rate data. IEEE Transactions on Reliability. 42, 299-302.
16. Muhammad, M (2016a). A generalization of the BurrXII-Poisson distribution and its applications. Journal of Statistics Applications & Probability. 5, 29-41.
17. Muhammad, M (2016b). Poisson-odd generalized exponential family of distributions: theory and Applications. Hacettepe University Bulletin of Natural Sciences and Engineering Series B: Mathematics and Statistics.https://doi.org/10.15672/hjms.2016.393
18. Muhammad, M (2017). The complementary exponentiated BurrXII-Poisson distribution: model, properties and application. Journal of Statistics Applications & Probability. 6, 33-48.
19. Muhammad, M, and Yahaya, MA (2017). The half logistic-Poisson distribution. Asian Journal of Mathematics and Applications. 2017, 1-15.
20. Nadarajah, S, and Haghighi, F (2011). An extension of the exponential distribution. Statistics. 45, 543-558.
21. Olapade, AK (2014). The type I generalized half logistic distribution. Journal of Iranian Statistical Society. 13, 69-82.
22. Pappas, V, Adamidis, K, and Loukas, S (2015). A generalization of the exponential-logarithmic distribution. Journal of Statistical Theory and Practice. 9, 122-133.
23. Raja, TA, and Mir, AH (2011). On extension of some exponentiated distributions with application. International Journal of Contemporary Mathematical Sciences. 6, 393-400.
24. Seo, JI, Kim, Y, and Kang, SB (2013). Estimation on the generalized half logistic distribution under Type-II hybrid censoring. Communications for Statistical Applications and Methods. 20, 63-75.
25. Seo, JI, Lee, HJ, and Kan, SB (2012). Estimation for generalized half logistic distribution based on records. Journal of the Korean Data and Information Science Society. 23, 1249-1257.
26. Silva, RB, Barreto-Souza, W, and Cordeiro, GM (2010). A new distribution with decreasing, increasing and upside-down bathtub failure rate. Computational Statistics & Data Analysis. 54, 935-944.
27. Silva, RB, and Cordeiro, GM (2015). The Burr XII power series distributions: a new compounding family. Brazilian Journal of Probability and Statistics. 29, 565-589.
28. Tahmasbi, R, and Rezaei, S (2008). A two-parameter lifetime distribution with decreasing failure rate. Computational Statistics & Data Analysis. 52, 3889-3901.