Journal of Taibah University for Science

ISSN: (Print) 1658-3655 (Online) Journal homepage: https://www.tandfonline.com/loi/tusc20

A new one parameter distribution: properties and

estimation with applications to complete and type
II censored data

R. Alshenawy

To cite this article: R. Alshenawy (2020) A new one parameter distribution: properties and
estimation with applications to complete and type II censored data, Journal of Taibah University for
Science, 14:1, 11-18, DOI: 10.1080/16583655.2019.1698276

To link to this article: https://doi.org/10.1080/16583655.2019.1698276

© 2020 The Author(s). Published by Informa

UK Limited, trading as Taylor & Francis
Group

Published online: 10 Dec 2019.

Submit your article to this journal

Article views: 933

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tusc20
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE
2020, VOL. 14, NO. 1, 11–18
https://doi.org/10.1080/16583655.2019.1698276

A new one parameter distribution: properties and estimation with applications

to complete and type II censored data
R. Alshenawy a,b

a Faculty of Science, Department of Mathematics and Statistics, King Feisal University, Al-Ahsaa, Saudi Arabia; b Faculty of Commerce,
Department of Applied Statistics and Insurance, Mansoura University, Mansoura, Egypt

ABSTRACT ARTICLE HISTORY

The aim of this paper is to propose a new flexible model with one parameter. Several statisti- Received 24 August 2019
cal properties and reliability characteristics of the proposed model are discussed. The maximum Revised 26 October 2019
likelihood method is used to estimate the model parameter under complete and Type II cen- Accepted 6 November 2019
sored samples. Simulation is conducted to estimate both the relative absolute bias and the KEYWORDS
relative mean square error for the model parameter. Finally, three real data sets are analyzed Hazard rate function; lifetime
to investigate the applicability of the proposed model. expectancy; maximum
likelihood method; censored
data

1. Introduction 2. A new one parameter model

Numerous classical distributions have been used exten- The random variable Xis said to have a new (A) distribu-
sively over the past decades for modelling univariate tion with one scale parameter β(X ∼ A(β), for short) if
and bivariate data in areas such as engineering, actu- its cumulative distribution function (CDF) is given by
arial, environmental and medical sciences, biological    
1 β
studies, demography, economics, finance and insur- FX (x; β) = exp 1 − exp , x > 0, β > 0,
ance. Gompertz and its inverse distributions are of these β x
(1)
models. The last two models with their extensions are
where the CDF can be obtained as a special case from
widely used in several fields such as reliability, actuarial
the inverse Gompertz model. The probability density
science, queuing problems and biological sciences. For
function (PDF) of the A distribution can be expressed as
example, El-Bassiouny et al. [1–5].
    
In the statistical literature, there are many distribu- 1 1 β β
fX (x; β) = 2 exp 1 − exp + , x > 0,
tions have only one parameter such as exponential, x β x x
Rayleigh, Lindley, and its inverse and modified versions (2)
d
(univariate and bivariate). These models can be used for where fX (x; β) = dx FX (x; β). Figures 1 and 2 show the
modelling lifetime data in various fields because these CDF and PDF of the A distribution for various values of
models have several desirable properties and nice phys- the parameter β, respectively. From Figure 2, it is clear
ical interpretations. For example, Kundu and Raqab [6], that the PDF of the A distribution has uni-model shape.
Cordeiro et al. [7], Ghitany et al. [8], Merovci [9–11],
Elbatal and Elgarhy [12], Adepoju et al. [13], Ahmad 3. Statistical properties
et al. [14], Oguntunde et al. [15], Sarhan et al. [16],
3.1. The quantile and mode
Oguntunde and Adejumo [17], El-Gohary et al. [18],
Merovci and Elbatal [19], Haq [20], El-Morshedy et al. If X ∼ A(β), then the quantile, say xq , is given by
[21–23], Eliwa et al. [24], El-Morshedy and Eliwa [25],
β
Basheer [26], Alizadeh et al. [27], Eliwa and El-Morshedy xq = , 0 < q < 1, (3)
[28–30], among others. ln(1 − β ln q)
In this paper, we propose a new flexible one- where FX (xq ; β) − q = 0. Setting q = 12 in Equation (3),
parameter lifetime distribution. This model is more flex- we get the median of the A distribution as follows
ible when it is compared to well-known models in the
statistical literature. Moreover, it can be used to model β
Med(X) = . (4)
different types of data sets. ln(1 − β ln( 12 ))

CONTACT R. Alshenawy [email protected] Faculty of Science, Department of Mathematics and Statistics, King Feisal University, O. Box 400,
Al-Ahsaa 31982, Saudi Arabia; Faculty of Commerce, Department of Applied Statistics and Insurance, Mansoura University, Mansoura 35516, Egypt
© 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted
use, distribution, and reproduction in any medium, provided the original work is properly cited.
12 R. ALSHENAWY

∞
where (ϑ) = 0 e−x x ϑ−1 dx and

     
dk (1 + r) β 2β
vr(k) = exp 1 − exp + |x=x0 ,
dx k β x x
ϑ = m − k − s + 1 and x0 ∈]0, 0.1[,

On the other hand, it is important to get the moments

to calculate some statistical properties like expectation,
variance, skewness, and kurtosis measures. So, if X ∼
A(β), the rth moment of X is given as follows

∞ 
 k
Figure 1. The CDF of A distribution for different values of β.  (−x0 )m w(k) k −ξ
μr = Cm β (ξ ), / {0} ∪ Z − ,
ξ∈
k!
k=0 m=0
∞ (8)
where (ξ ) = 0 e−x x ξ −1 dx and

     
dk 1 β 2β
w(k) = exp 1 − exp + |x=x0 ,
dx k β x x
ξ = m − k − r + 1, x0 ∈ ]0, 0.1[.

Setting r = 1 in Equation (8), we get the mean of X.

2
The skewness and kurtosis can be calculated from the
ordinary moments of X.

Figure 2. The PDF of A distribution for different values of β.

4. Reliability properties

Furthermore, we can derive the mode of the A distri- 4.1. Reliability and reversed (hazard) rate
bution by solving the following equation with respect functions
to x If X ∼ A(β), then the reliability function of X is given by
 
β    
exp − 2x − β = 0, (5) 1 β
x RX (x; β) = 1 − exp 1 − exp , x, β > 0.
β x
d
where dx fX (x; β) = 0. In the general case, it is not pos- (9)
sible to get an explicit solution to Equation (5). It has to The hazard rate (HR) function of X is
be obtained numerically.
exp( βx )
hX (x; β) = , x > 0. (10)
3.2. Probability-weighted moments and moments x 2 exp( β1 (exp( βx ) − 1))

The probability-weighted moment (PWM) can be used

to estimate the parameters of a distribution whose where hX (x; β) = RfXX(x;β)
(x;β) . Figure 3 shows the HR func-
inverse form cannot be expressed explicitly. The (s, r)th tion of the A model for various values of the param-
PWM of non-negative X following any FX (.), say ρs,r , is eter β. From Figure 3, it is clear that the HR function
defined by of the A distribution is an upside-down bathtub shape,
which makes the A distribution more flexible to fit dif-
 ∞ ferent lifetime data sets. The reversed hazard rate (RHR)
ρs,r = E(X s F r (x)) = x s FXr (x)fX (x) dx. (6)
0 function of X is
 
Assume X ∼ A(β), and using Taylor and binomial 1 β
rX (x; β) = 2 exp , x > 0, (11)
expansions, the ρs,r of X is given as x x

∞ 
 k (k)
(−x0 )m vr k −ϑ
fX (x;β)
where rX (x; β) = 1−R . From Equation (11), it is
ρs,r = Cm β (ϑ), / {0} ∪ Z − ,
ϑ∈ X (x;β)
k! immediate that the shape of the RHR function is
k=0 m=0
(7) decreasing.
 JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 13

follow

Mr (t) = E(T − t | T > t). (15)

Therefore, if T ∼ A(β), then

⎛
 ∞ l
1 ⎝μr − t +
Mr (t) =
RT (t; β)
l=0 i=0

(−t0 )i w∗(l) l l−i−1
× C t , (16)
l! (l − i + 1) m

Figure 3. The HR function of the A distribution for different where RT (t; β) = 1 − exp( β1 (1 − exp( βt ))) and
values of β.
    

dl 1 β

w∗(l) = l exp 1 − exp

4.2. Rényi and δ− entropies dt β t

t=t0

If X ∼ A(β), then the Rényi entropy is defined as and t0 ∈]0, 0.1[.

 ∞
1
Iδ (X) = log fXδ (x) dx, δ ∈ ]0, ∞[ − {1} Also, the mean inactive time (MIT) or the mean wait-
1−δ 0
⎛ ing time (MWT) for describing different maintenance
∞ 
1
k
strategies, say Mw (t) is given as follow
= log ⎝
1−δ
k=0 m=0
Mw (t) = E(t − T | T ≤ t). (17)
(−x0 )m u(k) k −τ
× Cm (δβ) (τ ) , τ ∈ / {0} ∪ Z − , Using Equations (1), (8 ), and (16), we get
k!
(12) 1 
Mw (t) = (t + Mr (t)R(t) − μr ), (18)
where FT (t; β)
     

u(k) =
dk δ
1 − exp
β
+
2δβ
where FT (t; β) = exp( β1 (1 − exp( βt ))). The Mw (t) can
exp
,
dx k β x x
be used to get the CDF of the A model as follows
x=x0

τ = 2δ − k + m − 1, x0 ∈ ]0, 0.1[.   ∞  
1 − Mw (t)
FX (x; β) = exp − r(t) dt , r(t) = ,
Equation (12) can be obtained by using Taylor and bino- x Mw (t)
(19)
mial expansions. The δ−entropy, say Hδ (X), is given
where Mw (t) is differentiable.
as
1
Hδ (X) = log {1 − [(1 − δ) Iδ (X)]} . (13) 4.5. Stochastic orderings
δ−1
Shaked and Shanthikumar [31] state that for the random
4.3. Stress-strength reliability variable X and Y
If X ∼ A(β1 ) represents the stress which is applied to
if X ≤lr Y then X ≤hr Y =⇒ X ≤mrl Y =⇒ X ≤st Y,
an item or a component and Y ∼ A(β2 ) represents the
(20)
strength to sustain the stress, then the probability that
where lr :likelihood ratio order, hr :hazard rate order,
the system is strong enough to overcome the stress is
mrl :mean residual life order and st :stochastic order.
given by
 ∞ Assume X ∼ A(β1 )and Y ∼ A(β2 ), if β1 < β2 then
β1 X ≤lr Y, hence X ≤hr Y =⇒ X ≤mrl Y =⇒ X ≤st Y for
R= RY (x; β2 ) dFX (x; β1 ) = , (14)
0 β1 + β2 x ∈ (0, 1)where

where RY (x; β2 ) = 1 − exp( β12 (1 − exp( βx2 ))) and d fX (x, β1 )

ln < 0.
FX (x; β1 ) = exp( β11 (1 − exp( βx1 ))). dx fY (x, β2 )

4.4. Mean residual and mean inactive lifetimes 4.6. Reliability to parallel and series systems
The mean residual lifetime (MRL) of T−t given that the Assume parallel and series systems with independent n
component has survived to time t, say Mr (t) is given as components, each component has A model, then the
14 R. ALSHENAWY

reliability function in case of parallel system is given by n!  k

    l∗ = (RX (xk ))n−k fX (xi ). (26)
n β (n − k)!
RP−S (x) = 1 − exp 1 − exp , x > 0. i=1
β x
(21) If X1 , X2 , . . . , Xn is a random sample from the A distribu-
The reliability function in case of series system is given tion, and X(1) , X(2) , . . . , X(k) , k ≤ nrepresent the ordered
by sample obtained from Type II right-censored data, then
    n the log-likelihood function L∗ is
1 β  
RS−S (x) = 1 − exp 1 − exp , x > 0. n!
β x L∗ = ln + (n − k)
(22) (n − k)!
    
1 β
× ln 1 − exp 1 − exp
5. Parameter estimation β xk
 k    
1 
k k
5.1. Maximum likelihood estimator based on 1 β
+β − exp −1 −2 ln xi .
complete data xi β xi
i=1 i=1 i=1
Assume X1 , X2 , . . . , Xn be a random sample of size n from (27)
the A distribution, then the likelihood function is
Differentiating Equation (27) with respect to β, we get

n
∂ ∗ (n − k) ∂
= fX (xi ; β). (23) L =
∂β 1 β ∂β
i=1 1 − exp( β (1 − exp( x )))
k
    
Substituting from Equation (2) into Equation (23), we 1 β
× 1 − exp 1 − exp
get β xk
         
 1 
n k k
1 1 β β 1 β
= exp 1 − exp + . + + 2 exp −1
xi2 β xi xi xi β xi
i=1 i=1 i=1


k  
The log-likelihood function is given by 1 1 β
− exp . (28)
 n     β xi xi
1 
n n
1 β i=1
LL = β − exp −1 −2 ln (xi ) .
x β xi The solution has to be obtained numerically.
i=1 i i=1 i=1
(24)
By differentiating Equation (24) with respect to β, and
6. Simulation
equating it to zero, we get
n       In this section, we estimate the relative absolute bias
n
1 1  β β β
− exp − exp + 1 = 0. (RAB), variance (Var), mean square error (MSE) and rel-
x (β)2 xi xi xi ative mean square error (RMSE) for the parameter β
i=1 i i=1
(25) using simulations under complete and Type II-censored
It is not possible to get an explicit solution to samples. The population parameter is generated using
Equation (25). So, the solution has to be obtained software “Mathcad 15” package programme. The sam-
numerically. pling distributions are obtained for different sample
sizes n = [20, 50, 100, 400] from N = 1000 replications
5.2. Maximum likelihood estimation based on for different values of the parameter β.
type II right-censored data
6.1. Simulation study based on complete data
In reliability, medical, and engineering research, cen-
soring is a case in which the value of data is only par- Table 1 lists an assessment of the MLE properties for the
tially known. There are many applications ranging from parameter βin terms of bias, RAB, Var, MSE and RMSE.
accelerated life testing through to earthquakes, floods, We note that the bias and MSE are reduced as the
rainfall, sea currents, and wind speeds, see Kotz and sample size is increased.
Nadarajah [32]. One of the most common censored data
sets, the Type II right-censored data. If we have a set
6.2. Simulation study based on type II
number of items and stop the experiment when a fixed
right-censored data
number is observed to have failed, the residual items are
then right-censored. Many distributions were studied Table 2 reports an assessment of the MLE properties
based on Type II right-censored data sets; see Balakr- for the parameter βin terms of bias, RAB, Var, MSE and
ishnan and Aggarwala [33], Castro-Kuriss [34] and Eliwa RMSE. The sampling distributions are obtained for dif-
and El-Morshedy [28] . The likelihood function l∗ is ferent sample sizes n = [20, 50, 100, 400], and k = np,
 JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 15

Table 1. The MLE, Bias, RAB, Var, MSE, and RMSE based on Table 3. The strength of glass of the aircraft window data.
complete data.
18.83 20.8 21.657 23.03 23.23 24.05 24.321 25.5 25.52 25.8 26.69
∧ 26.77 26.78 27.05 27.67 29.9 31.11 33.2 33.73 33.76 33.89 34.76
n β β Bias RAB Var MSE RMSE 35.75 35.91 36.98 37.08 37.09 39.58 44.045 45.29 45.381
20 0.125 0.232604 0107604 0.86084 0.06523 0.07681 0.61449
0.6 0.70478 0.10478 0.17463 0.09402 0.10500 0.17500
1.0 1.10622 0.10622 0.10622 0.12228 0.13357 0.13357
50 0.125 0.16879 0.04379 0.35035 0.01668 0.01859 0.14875
Table 4. The CDF, MLE, −LL, K-S and p-values for data set I.
0.6 0.64149 0.04149 0.06916 0.03167 0.03339 0.05567 Model CDF MLE −LL K-S p-value
1.0 1.04807 0.04807 0.04807 0.04045 0.04276 0.04276
E 1 − exp(−βx) 0.032 137.27 0.459 4.340 × 10−6
100 0.125 0.14117 0.01617 0.12939 0.007474 0.00774 0.06189
0.6 0.61738 0.01738 0.02896 0.014876 0.01518 0.02529 IE exp(−β/x) 29.215 137.26 0.475 1.682 × 10−6
1.0 1.02365 0.02365 0.02365 0.01979 0.02036 0.03036 L 1− 1+β+βx
exp(−βx) 0.063 126.99 0.365 5.173 × 10−4
1+β
400 0.125 0.13039 0.00539 0.04314 0.001699 0.00173 0.01383
0.6 0.60728 0.007282 0.01214 0.003197 0.00325 0.00542 R 1 − exp(−βx 2 ) 9.997 × 10−4 118.22 0.319 3.657 × 10−3
1.0 1.00613 0.00613 0.00613 0.00465 0.00469 0.00469
IR exp(−β/x 2 ) 810.503 118.20 0.325 2.864 × 10−3
A exp( β1 (1 − exp( βx ))) 125.662 107.95 0.162 0.393
Table 2. The MLE, Bias, RAB, Var, MSE, and RMSE based on
censored data.
∧
n p β β Bias RAB Var MSE RMSE Table 5. The AIC, CAIC, BIC, and HQIC values
20 0.3 0.25 0.0691 −0.0558 0.4466 0.0365 0.9996 0.3174 for data set I.
0.6 0.3975 −0.2024 0.3374 0.0660 0.1070 0.1784
Model AIC CAIC BIC HQIC
1 0.7368 −0.2631 0.2631 0.1114 0.1806 0.1806
0.5 0.25 0.1131 −0.0118 0.0947 0.0459 0.0460 0.3684 E 276.54 276.67 277.97 277.00
0.6 0.5432 −0.0567 0.0946 0.0763 0.0795 0.1326 IE 276.52 276.66 277.95 276.99
1 0.9354 −0.0645 0.0645 0.1455 0.1497 0.1497 L 255.99 256.13 257.42 256.46
0.8 0.25 0.2167 0.0917 0.7343 0.0625 0.0709 0.5679 R 238.44 238.58 239.91 238.91
0.6 0.7037 0.1037 0.1728 0.1072 0.1179 0.1966 IR 238.40 238.54 239.84 238.87
1 1.0860 0.0860 0.0860 0.1358 0.1432 0.1432 A 217.90 218.04 219.33 218.39
50 0.3 0.25 0.0182 −0.9817 0.9817 0.0099 0.9737 0.9737
0.6 0.3314 −0.2685 0.4475 0.0205 0.0926 0.1543
1 0.6618 −0.3381 0.3381 0.0359 0.1502 0.1502
0.5 0.25 0.0722 −0.0527 0.4217 0.0141 0.0169 0.1352 (IR), and Lindley (L) distributions. These models have
0.6 0.4802 −0.1197 0.1995 0.0231 0.0374 0.0624
1 0.8501 −0.1498 0.1498 0.0349 0.0573 0.0573 only one parameter in the standard case. In order
0.8 0.25 0.1477 0.0227 0.1822 0.0161 0.0167 0.1336 to do this comparison, some criteria are used like
0.6 0.6165 0.0165 0.0276 0.0315 0.0318 0.0531
1 1.0209 0.0209 0.0209 0.0441 0.0446 0.0446
−LL, Kolmogorov-Smirnov (K-S) distance test statistic
(see, Kolmogorov [35] and Smirnov [36]), Akaike infor-
100 0.3 0.25 0.0070 −0.1179 0.9438 0.0029 0.0168 0.1346
0.6 0.3108 −0.2891 0.4818 0.0085 0.0921 0.1535 mation criterion (AIC) and corrected AIC (CAIC) (see,
1 0.6308 −0.3691 0.3691 0.0153 0.1516 0.1516 Hurvich and Tsai [37]), Bayesian information criterion
0.5 0.25 0.0479 −0.0770 0.6166 0.0077 0.0136 0.1094
0.6 0.4579 −0.1420 0.2368 0.0112 0.0313 0.0523 (BIC) (see, Zhao [38]), and Hannan-Quinn information
1 0.8203 −0.1796 0.1796 0.0187 0.0510 0.0510 criterion (HQIC) (see, Hannan and Quinn [39]).
0.8 0.25 0.1354 0.0104 0.0837 0.0079 0.0080 0.0641
0.6 0.5970 −0.0029 0.0049 0.0138 0.0138 0.0231
Data set I: The data set in Table 3 gives the strength
1 0.9887 −0.0112 0.0112 0.0192 0.0193 0.0193 of a glass of the aircraft window (see Fuller et al. [40]).
400 0.3 0.25 0.0015 −0.1235 0.9880 0.0001 0.0152 0.1222 Table 4 lists the MLE, −LL, K-S and p-values for data
0.6 0.2932 −0.3067 0.5112 0.0020 0.0961 0.1603 set I.
1 0.6100 −0.3899 0.3899 0.0032 0.1552 0.1552
0.5 0.25 0.0390 −0.0859 0.6872 0.0015 0.0089 0.0712 Table 5 reports goodness-of-fit measures for data
0.6 0.4478 −0.1521 0.2535 0.0025 0.0256 0.0427 set I.
1 0.8030 −0.1969 0.1969 0.0042 0.0430 0.0430
0.8 0.25 0.1172 −0.0077 0.0623 0.0017 0.0018 0.0146
From Tables 4 and 5, we can conclude that the A dis-
0.6 0.5802 −0.0197 0.0329 0.0031 0.0035 0.0059 tribution is the best distribution among all the tested
1 0.9686 −0.0313 0.0313 0.0046 0.0056 0.0056 distributions, because it has the smallest value of −LL,
K-S, AIC, CAIC, BIC, and HQIC. For the A distribution, the
approximate 95% two-sided confidence interval (CI) for
where 0 < p < 1 for different values of the parameter β.
the parameter β is [77.124, 135.415]. Figures 4 and 5
We note that the bias is reduced as the sample size is
show estimated reliability function, and HR function for
increased.
data set I, respectively.
Figure 6 shows the profile of log-likelihood function
7. Data analysis to parameter β for data set I.
It is explained that the likelihood equation has a
7.1. Data analysis based on complete data
unique solution.
In this section, the proposed model is compared with Dataset II: The data set in Table 6 provided by
some competitive models such as exponential (E), Balakrishnan and Aggarwala [33] in a city located in
inverse exponential (IE), Rayleigh (R), inverse Rayleigh the south part of Chile. This data is regarding with
16 R. ALSHENAWY

Table 7. The MLE, −LL, K-S and p-values for data set II.
Model MLE −LL K-S p-value
E 0.04 564.13 0.533 9 × 10−10
IE 23.99 564.89 0.454 1 × 10−10
L 0.077 520.82 0.455 9 × 10−10
R 0.00154 481.02 0.425 9 × 10−10
IR 556.30 479.12 0.322 6 × 10−10
A 107.354 404.28 0.129 0.022

Table 8. The AIC, CAIC, BIC, and HQIC values for data set II.
Model AIC CAIC BIC HQIC
E 1132.26 1132.29 1135.16 1133.44
Figure 4. The empirical and estimated reliability functions of IE, IE 1131.77 1131.80 1134.67 1132.95
L 1043.65 1043.68 1046.54 1044.82
R, IR, L, E, and A distributions for data set I. R 964.04 964.07 966.94 965.22
IR 960.23 960.26 963.13 961.41
A 810.55 810.58 813.45 811.73

Figure 5. The empirical and HR functions of IE, R, IR, L, E, and A

distributions for data set I.

Figure 7. The empirical and estimated reliability functions of IE,

R, IR, L, E, and A distributions for data set II.

Figure 6. The profile of the log-likelihood functions of β for Figure 8. The empirical and HR functions of IE, R, IR, L, E, and A
data set I. distributions for data set II.

Table 6. The behavioural sciences data.

Table 8 reports goodness-of-fit measures for data
19(16) 20(15) 21(14) 22(9) 23(12) 24(10) 25(6) 26(9) set II.
27(8) 28(5) 29(6) 30(4) 31(3) 32(4) 33(1) 34(1)
35(4) 36(2) 37(2) 39(1) 42(1) 44(1) From Tables 7 and 8, we can conclude that, the A
distribution is the best distribution, as the A model has
the smallest value of −LL, K-S, AIC, CAIC, BIC, and HQIC
behavioural sciences: The scale “General Rating of Affec- among all tested distributions. For the A distribution,
tive Symptoms for Preschoolers (GRASP)” measures the the approximate 95% two-sided CI for the parameter β
emotional problem and the behavioural problem of is [93.339, 121.368]. The empirical and estimated relia-
children, which can be classified with depressive status bility functions for data set II is plotted in Figure 7, and
or not based on this scale. the empirical and HR function is plotted in Figure 8.
Table 7 lists the MLE, −LL, K-S and p-values for data Figure 9 shows the profile of the log-likelihood func-
set II. tion of β for data set II.
 JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE 17

Funding
The authors acknowledge the Deanship of Scientific Research
at King Faisal University for their support under grant No.
180132.

ORCID
R. Alshenawy http://orcid.org/0000-0002-6570-4733

References
[1] El-Bassiouny AH, EL-Damcese M, Mustafa A, Eliwa MS.
Figure 9. The profile of the log-likelihood function of β for data Exponentiated generalized Weibull-Gompertz distribu-
set II. tion with application in survival analysis. J Stat Appl Prob.
2017;6(1):7–16.
[2] El-Bassiouny AH, EL-Damcese M, Mustafa A, Eliwa MS.
Characterization of the generalized Weibull-Gompertz
Table 9. The MLEs, L, KS and p-values. distribution based on the upper record values. Int J Math
Model MLEs −L∗ K-S p-value Appl. 2015;3(3):13–22.
[3] El-Bassiouny AH, Tahir MH, Elmorshedy M, Eliwa MS.
A 1.297 102.364 0.157 0.897
Univariate and multivariate double slash distribution:
properties and application. J Stat Appl Prob. 2019. To
appear.
[4] El-Bassiouny AH, EL-Damcese M, Mustafa A, Eliwa MS.
From Figure 9, it is clear that the likelihood equation Mixture of exponentiated generalized Weibull-Gompertz
has a unique solution. distribution and its applications in reliability. J Stat Appl
Prob. 2016a;5(3):1–14.
[5] El-Bassiouny AH, EL-Damcese M, Abdelfattah M, Eliwa
7.1.1. Data analysis based on type II censored data MS. Bivariate exponentaited generalized Weibull-
The censored data have been obtained from McCool Gompertz distribution. J Appl Probab Stat. 2016b;11(1):
[41] which obtains the fatigue life of 10 bearings of a cer- 25–46.
[6] Kundu D, Raqab MZ. Generalized Rayleigh distribution:
tain type in hours. Assume type II right-censored sample
different methods of estimation. Comput Stat Data Anal.
of size k = 8. Table 9 represents the MLE, −L∗ , K-S and 2005;49:187–200.
p-values for the A distribution. [7] Cordeiro GM, Cristino CT, Hashimoto EM, Ortega EM.
From Table 9, it is clear that the A distribution pro- The beta generalized Rayleigh distribution. Stat Pap.
vides a good fit for this data. 2013;54:133–161.
[8] Ghitany ME, Al-Mutairi DK, Balakrishnan N, Al-Enezi LJ.
Power Lindley distribution and associated inference.
Comput Stat Data Anal. 2013;64:20–33.
8. Conclusions [9] Merovci F. Transmuted exponentiated exponential distri-
bution. Math Sci Appl E-Notes. 2013a;1(2):112–122.
In this paper, we have proposed a new lifetime distri-
[10] Merovci F. Transmuted lindley distribution. Int J Open
bution with one parameter. This distribution exhibits an Probl Comput Sci Math. 2013b;6:63–72.
upside-down bathtub-shaped hazard function, which [11] Merovci F. Transmuted Rayleigh distribution. Austrian J
makes it a suitable distribution for fitting several real Stat. 2013c;42(1):21–31.
lifetime data sets. Some statistical and reliability proper- [12] Elbatal I, Elgarhy M. Statistical properties of
Kumaraswamy quasi Lindley distribution. Int J Math
ties have been studied. The MLE method has been used
Trends Technol. 2013;4:237–246.
to estimate the model parameter based on complete [13] Adepoju KA, Chukwu AU, Wang M. The beta power expo-
and Type II censored data. We have conducted two sim- nential distribution. J Stat Sci Appl. 2014;2:37–46.
ulations, one of them based on complete data whereas [14] Ahmad A, Ahmad SP, Ahmed A. Transmuted inverse
the other based on Type II right-censored data to esti- Rayleigh distribution: a generalization of the inverse
mate both RAB and RMSE for the model parameter, Rayleigh distribution. Math Theory Model. 2014;4(7).
[15] Oguntunde PE, Babatunde OS, Ogunmola AO. Theoret-
and it is found that the MLE is a good method to esti-
ical analysis of the Kumaraswamy inverse exponential
mate the model parameter. Finally, the empirical results distribution. Int J Stat Appl. 2014;4(2):113–116.
have shown that the A distribution provides a better fit [16] Sarhan M, Lotfi T, Hamilton DC. A new lifetime distribu-
than exponential, inverse exponential, Rayleigh, inverse tion and its power transformation. J Probab Stat. 2014.
Rayleigh, and Lindley distributions. article ID 532024.
[17] Oguntunde PE, Adejumo AO. Transmuted inverse expo-
nential distribution. Int J Adv Stat Prob. 2015;3(1):1–7.
[18] El-Gohary A, El-Bassiouny AH, El-Morshedy M. Inverse
Disclosure statement flexible Weibull extension distribution. Int J Comput
No potential conflict of interest was reported by the authors. Appl. 2015;115:46–51.
18 R. ALSHENAWY

[19] Merovci F, Elbatal I. Weibull-Rayleigh distribution: the- [29] Eliwa MS, El-Morshedy M. Bivariate odd Weibull-V family
ory and applications. Appl Math Inf Sci. 2015;9(4):2127– of distributions: properties, Bayesian and non-Bayesian
2137. estimation with bootstrap confidence intervals and
[20] Haq MA. Kumaraswamy exponentiated inverse Rayleigh application. J Taibah Univ Sci. 2020a. To appear.
distribution. Math Theory Model. 2016;6(3):93–104. [30] Eliwa MS, El-Morshedy M. Discrete flexible distribution
[21] El-Morshedy M, El-Bassiouny AH, El-Gohary A. Exponenti- for over-dispersed data: statistical and reliability proper-
ated inverse flexible Weibull extension distribution. J Stat ties with estimation approaches and applications. J Appl
Appl Prob. 2017;6(1):169–183. Stat. 2020b. To appear.
[22] El-Morshedy M, Eliwa MS, El-Gohary A, Khalil AA. Bivari- [31] Shaked M, Shanthikumar JG. Stochastic orders and their
ate exponentiated discrete Weibull distribution: statisti- applications. New York: Academic Press; 1994.
cal properties, estimation, simulation and applications. [32] Kotz S, Nadarajah S. Extreme value distributions. Theory
Math Sci. 2019b. doi:10.1007/s40096-019-00313-9. and applications. London: Imperial College Press; 2000.
[23] El-Morshedy M, Eliwa MS, Nagy H. A new two-parameter [33] Balakrishnan N, Aggarwala R. Censoring theory. Methods
exponentiated discrete Lindley distribution: proper- and applications. Boston, MA: Birkhaauser; 2000.
ties, estimation and applications. J Appl Stat. 2019a. [34] Castro-Kuriss C. On a goodness-of-fit test for censored
doi:10.1080/02664763.2019.1638893. data from a location-scale distribution with applications.
[24] Eliwa MS, El-Morshedy M, Afify AZ. The odd Chen gener- Chil J Stat. 2011;2:115–136.
ator of distributions: properties and estimation methods [35] Kolmogorov A. Confidence limits for an unknown distri-
with applications in medicine and engineering. J Natl Sci bution function. Ann Math Stat. 1941;4:461–463.
Found Sri Lanka. 2019. To appear. [36] Smirnov N. Approximate laws of distribution of ran-
[25] El-Morshedy M, Eliwa MS. The odd flexible Weibull-H dom variables from empirical data. Uspehi Matem Nauk.
family of distributions: properties and estimation with 1944;10:179–206.
applications to complete and upper record data. Filomat. [37] Hurvich CM, Tsai CL. A corrected Akaike information cri-
2019;33(9). To appear. terion for vector autoregressive model selection. J Time
[26] Basheer AM. Alpha power inverse Weibull distribu- Ser Anal. 1993;14:271–279.
tion with reliability application. J Taibah Univ Sci. [38] Zhao XM. Bayesian information criterion (BIC). Encycl Syst
2019;13(1):423–432. Biol. 2013;73–73.
[27] Alizadeh M, Afify AZ, Eliwa MS, Ali S. The odd log-logistic [39] Hannan EJ, Quinn BG. The determination of the order of
Lindley-G family of distributions: properties, Bayesian an autoregression. J R Stat Soc Ser B. 1979;41:190–195.
and non-Bayesian estimation with applications. Comput [40] Fuller EJ, Frieman S, Quinn J, Quinn G, Carter W. Frac-
Stat. 2019. doi:10.1007/s00180-019-00932-9. ture mechanics approach to the design of glass aircraft
[28] Eliwa MS, El-Morshedy M. Bivariate Gumbel-G family of windows: a case study. SPIE Proc, 1994. p. 419–430.
distributions: statistical properties, Bayesian and non- [41] McCool JI. Inference techniques for Weibull population.
Bayesian estimation with application. Ann Data Sci. 2018. Aerospace research laboratries report ARLTR; 2000. p.
doi:10.1007/s40745-018-00190-4. 74–80.