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

net/publication/279516790

PARAMETERS ESTIMATION METHODS OF THE WEIBULL DISTRIBUTION: A

COMPARATIVE STUDY

Article · April 2014

CITATION READS

1 1,915

2 authors:

Nwosu Chinwe R Osarumwense Osabuohien-Irabor

Nnamdi Azikiwe University, Awka Ambrose Alli University
23 PUBLICATIONS   29 CITATIONS    17 PUBLICATIONS   65 CITATIONS   

SEE PROFILE SEE PROFILE

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

Works on Regression View project

article co-authored by Ebuh G.U. View project

All content following this page was uploaded by Nwosu Chinwe R on 01 July 2015.

The user has requested enhancement of the downloaded file.

23177 Osabuohien-Irabor Osarumwense and Nwosu Chinwe Rose/ Elixir Statistics 69 (2014) 23177-23184

Available online at www.elixirpublishers.com (Elixir International Journal)

Statistics
Elixir Statistics 69 (2014) 23177-23184

Parameters estimation methods of the Weibull distribution: A comparative

study
Osabuohien-Irabor Osarumwense1 and Nwosu Chinwe Rose2
1
Department of Mathematics/Statistics, Ambrose Alli University, Nigeria.
2
Department of Applied Statistics, Nnamdi Azikiwe University, Nigeria.
ARTICLE INFO
A B S T RA CT
Art i cl e h i st ory :
In this study, a comparison of performance among four different methods (Probability
Received: 27 January 2014;
Weighted Moments (PWM), Maximum Likelihood Methods (MLM), the Methods of
Received in revised form:
Moments (MOM), and the least Square Method (LSM)) of estimating the parameters of the
29 March 2014;
Weibull distribution was done, using different values of α and β. This study contains two
Accepted: 12 April 2014;
different experiments; experiment I and experiment II. Values of α > β, α < β, α = β were
used in experiment I, and another varied values of α < β, α < β, α = β in line with some
Ke y w or d s
modeled distribution of the Weibull distribution, were used in experiment II. These values
Distribution, were generated by simulation. The aim is to find the best methods of estimating the two-
Prediction, parameter Weibull distribution. Based on the sample sizes and parameters considered, the
Estimate, method which gave the best estimate for the two-parameter Weibull distribution, is the
Parameter, method of Moments (MOM), taking into consideration the total deviation as a measurement
Weibull. for comparison. This study will help to estimate parameters associated with various models
of a component or system such as reliability problems, maintainability and predictions such
as MTTF, MTTFF, MTTR, MTBF, Hazard Rate etc.
© 2014 Elixir All rights reserved.

Introduction
The Weibull distribution is today seen as the best distribution in the family of failure distribution (Exponential, Rayleigh,
lognormal, normal etc). The Shape parameter of the Weibull distribution is what gives the Weibull distribution its flexibility, therefore
the need to evaluate precisely, accurately, and efficiently its parameters. The Weibull distribution most frequently provides the best fit
of life data. This is due in part to the broad range of distribution of the shape parameters that are included in the Weibull family. The
Weibull distribution is mainly used in predicting or forecasting failure rates in numerous applications in various areas; for example,
life expectancy, breaking strength, survival analysis, and animal bioassay [1], [2], [3]. One of the numerous advantages of the Weibull
distribution is the ability to provide reasonably accurate failure analysis and failure forecasts with extremely small sample. Solutions
are possible at the earliest indications of a problem without having to “crash a few more”. Although the Weibull distribution was first
identified by Frechet Maurice (1927), and first applied by Rosin and Rammler (1933) to describe the size of distribution of particles,
but it was not accepted as they failed to prove the logic behind the distribution. Similarly, Emil Gumbel (1958), a statistician, also
studied the distribution but he is best remembered for his confrontation with the Nazis. Again, a Swedish Engineer named Walodi
Weibull (1939), proposed the use of the distribution but the method also did not attract much attention unti 1951 when he presented
his most famous paper to the American Society of Mechanical Engineer (ASME) on the Weibull distribution. The Weibull
distribution is a continous distribution with a general probability density function (pdf), [4], is given as;
    
  


 

  0,   0,   0 (1)

Where ,    are the shape, scale and location paremeters respectively of the Weibull distribution.
But if   0, then the two-parameter Weibull distribution [5] given as,
 
  

   
    0,   0,   0 (2)

The two-parameter Weibull distribution can also be be stated as given by [3]

Tele: +2348034861146
E-mail addresses: [email protected],
[email protected]
© 2014 Elixir All rights reserved
23178 Osabuohien-Irabor Osarumwense and Nwosu Chinwe Rose/ Elixir Statistics 69 (2014) 23177-23184

      !"  0,   0,   0
 
(3)
Where t, is the time to failure (life length of a component), and α and β are the scale and shape parameter respectively. [6], also put the
two-parameter Weibull distribution which has been used in this study as;
#

    !

 $  0,   0,   0 (4)

The cummulation distribution function (CDF) of the two-parameter Weibull distribution also known in reliability study as Unit
unreliability written as F(t), given by [7] as,
#
%  : ,    1 (   $  0,   0,   0 (5)
#
Where %  : ,   is the unreliability unit and   $ is the reliability unit or survival function of the Weibull distribution
equating to 1.
The instantaneous failure rate at any point in line of the Weibull distribution, which is also known as the hazard function )* is
given by [8],

+    !

 0,   0,   0 (6)

and the mean and variance of the two-parameter Weibull distribution from (4) are given as respectively as,


,-    Γ 1 / 
(7)

2
3 
01-    Γ 1 / 
( Γ 3 1 / 
(8)

The shape parameter  , gives the Weibull distribution its flexibility, and as shown by [9], if β = 1, the Weibull distribution is
identical to the exponential distribution, if β = 2, it is identical to the Rayleigh distribution and if β = 3.5 and 2.5, the Weibull
approximates the normal and lognormal distribution respectively. These and other unknown parameters of the Weibull distribution
have to be estimated, hence, the need for this paper which aims at finding the best method of estimating these parameters.
Many papers have been published in the 1960s and 1970s concerning parameter estimation of the Weibull distribution based on
complete and censored sample. [10], described the Maximum likelihood estimation method and derived the variance-Covariance
matrix of the parameter of a two-parameter Weibull distribution based on complete and censored sample.
[11], developed an iterative procedure to find the Maximum likelihood estimate for a three-parameter Weibull distribution and
illustrated numerical examples for; one-, two-, and three-parameter Weibull model. [12], derive the elements of the information matrix
of the maximum likelihood estimate for a three-parameter Weibull distribution.
[13], derived the confidence limits of a two-parameter Weibull distribution.
[14], applied a two-parameter Weibull model for regional flood quantile estimation based on the index flood assumption and
compared the method of moments (MOM), probability Weighted Moments (PWM) and Maximum likelihood (ML) by simulation
experiment.
[15], derived the asymptotic variance of the quantile obtained from regional analysis for three estimation techniques for a two-
parameter Weibull Model. They also applied the regional Weibull Model to fit the frequency distribution of annual flood data.
[9], gave the confidence interval for the shape, scale and location parameters based on the method of Maximum likelihood and linear
estimation.
[16], proposed a method that depends on two-step iteration procedure. In the first step, he assumed that the location parameter is
known, and the scale and shape parameters are estimated using the graphical methods. In the second step, the shape parameter is
assumed known and the scale and location parameter are estimated by transforming the data using law transformation so that the
transformation data can be modeled to a two parameter exponential distribution.
23179 Osabuohien-Irabor Osarumwense and Nwosu Chinwe Rose/ Elixir Statistics 69 (2014) 23177-23184

[17], presented methods for estimating the parameters of the two-parameter Weibull distribution. He put his computational
experiments in two categories; Graphical methods (Weibull probability plotting and Hazard plotting technique) and analytical
methods (Maximum likelihood Estimator, Methods of Moments, and the Least squares Methods). Based on the computational results,
he concluded that the best estimating methods for two-parameter Weibull distribution, is the Method of Moments.
[18], estimated the parameter of the Weibull distribution graphically using the Weibull probability plot to transform the Weibull
transformation into x
[5], compared the three well-known methods (MLM, MOM, LSM) of parameter estimation of the Weibull distribution. A scale (α)
and shape (β) parameter of 1.0 and 3.0 were used respectively for the two-parameter Weibull distribution and values of 1, 2, and 3
were used for the three parameter Weibull distribution respectively.
[19], estimated the parameters of the modified Weibull distribution.
Material and methods
In this study, the data used were generated by simulation with the help of statistical software known as Minitab. Sets of
parameters used for this study were put in two different experiments. In experiment I, a case where the   3.0, 0.8,  7
 0.5, 4.0 and    2. Experiment II, which is in line with the values of some modeled distributions of the Weibull
distribution as shown in [9],   1,   1,   1,   2, and   1,   3.5. These six sets of parameters were used
in seven different sample sizes of 10, 20, 30,50, 60,70 and 100 giving a total of fifteen different sample data. With the help of
Microsoft Excel 2005, the probability density functions of these four different estimating methods were compared. And the data
generated plus the assumed parameters were used in evaluating these estimators.
Greenwood et al,[20], explained that a probability distribution function, F(   ;1<=. - >  can be characterized by the
probability weighted moments (PWM) given as;

?@,A,B C ,D @ %A 1 ( %B E  FG D %@ % A 1 ( % B E %  (9)

Where H, I  J are real numbers, but if I  J are non-negative integers, then;

?@,K,B  ∑BAMG BA

(1A ?@,A G (10)

A
?@,A,G  ∑GDBA E(1B ?@,G,B (11)

Where, if ?@,K,B exists and is a continuous function of % , ?@,A,G exists. And if,  %  is in inverse form, then,


%   N 


O":!,
(12)

If ?B  ?,G,B , then the probability weighted moments (PWM) of two-parameters Weibull distribution, without loss of generality
is given as,

!PQ

?B  
 (13)
RS T
QB 

Therefore,
VW
U  NYZW /Z  (14)
PX \
NY2

] 
N^3
N^VW /3V 
(15)

Where;
23180 Osabuohien-Irabor Osarumwense and Nwosu Chinwe Rose/ Elixir Statistics 69 (2014) 23177-23184

∑x
_G 
?

n i

∑x
i =1
^ ^
i

(16)
i =1

∑ x (F )
_  
i

?
n −1 i

∑ xi( n − i )
i =1
^^ ^

(17)
i =1

n 2
∑ x ( F i)
_3 
?

n −1 i

^^^3 ∑ x
i =1
( n − i )(n − i − 1) i


^
(18)
i =1

In the maximum Likelihood method, if  , 3 … … … ^ is a random sample of n independent observation whose probability density

is %  : ,  , then the likelihood can be written as,

N  , 3 … … … ^; ,    ∏  : , 
n
(19)
i=1

Then, the estimators are;

n

∑t N^"c
β

ά
i =1
(20)
n
d

Q ∑
i =1
N^"c


 ∑^@M 
^
(21)

Another analytical method of estimation used was the method of moments with estimators as,

If the J "e moment of the sample is;

g
fB  , B     Γ1 / h  (22)

Γn 
∞
n −1 −t
∫t e
0
dt (23)

Therefore, the estimators are;


 d
∑ "2 2
ά j d ck c
2 l
PQ 
(24)

 d 
∑ "
 j d ck c

PQ  
l (25)

For the least square estimator, double natural logarithms were “taken” to put the function (relationship between the cdf and the
parameters) in linear order.
#
% H   1 (   $ (26)
@
Where % H   ^Q
is the mean rank failure [17], H is the number of failure, n  number of data points, and N Natural

logarithms
Therefore,
23181 Osabuohien-Irabor Osarumwense and Nwosu Chinwe Rose/ Elixir Statistics 69 (2014) 23177-23184

U   "m no (27)
 
p^ ∑d
ckN^"c qN^jN^
lst  pq∑d
ck N^ jN^
ls ∑d
ck N^" c t
] 
 c⁄dS  c⁄dS
2 (28)
^ ∑d d
ckN^"c   u∑ckN ^"c v
2

Each of the different estimators for the four different methods was resolved simultaneously using the error predictor numerical
iteration as described by [7]. The iteration modifies the initial value of the shape parameter (β) and starting with a modification of ±
0.1, the values of  ( ά decreases in steps of 0.1, 0.01, 0.001…………etc as its converges. The iteration in this study used an
! ά
accuracy of 0.01 with a termination criterion of
3
> w , where ξ is the desired level of accuracy and  ( ά is the difference

 ] and U 
!Q ά
between the two estimators. And this set 
3
And the stirling approximation formula for gamma function was used in expressing values associated with gamma, and given by [21],
as,
| |
Γ / 1 x √2πα  }
(29)


! x  !  ! 2€ 1 / 3!

(30)

(30) is the modification of (29) which was used in this study.

In statistical analysis, the total deviation (TD) and the mean square error (MSE) are two measurements that can give an indication of
accuracy of parameter estimation [5]. The total deviation for the two-parameter Weibull distribution can be calculated as given by
[17],
ƒ !
! _ 

-  ‚ !
‚/ ‚ 
‚ (31)

The total deviation (TD) was used in this study as a measurement to indicate the degree of accuracy. The smaller the variance (Total
deviation), the better the method.
Results and discussion
The different estimators obtained from the four different methods of estimating the two parameter Weibull distribution were
evaluated and the different estimates at different sample sizes calculated.
In experiment I (see tables 1, 2, and 3), the values of α and β were varied, while in experiment II, as shown in tables 4, 5, and 6,
the values of α remain unchanged while β varied in line with some modeled distributions from the Weibull distribution. The
computational analysis of experiment I for the four compared estimation methods (Maximum likelihood Method, Method of Moments
and the least Square Method) for the two-parameter Weibull distribution revealed that the method of moments (MOM) is the best
estimating method with a total deviation (TD) of 0.190 and 0.432 when a sample size of 20 and 100 were used for values of   3.0
and   0.8 respectively as shown in Table I.
Table 1: Comparison of MLM, MOM and LSM when α = 3.0 and β = 0.8
Prob. Weighted
Parameters/ Sample Max. Likelihood Moments Method Least Square Method
Moments
sizes Method (MLM) (MOM) (LSM)
(PWM)
α β n ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD
3.0 0.8 20 3.621 0.889 0.318 2.245 0.706 0.369 3.197 0.900 0.190 4.010 1.051 0.649
60 3.809 0.943 0.447 1.706 0.850 0.493 1.028 0.950 0.844 6.401 1.021 1.409
100 4.000 0.716 0.438 1.857 0.841 0.432 1.740 0.790 0.432 6.990 1.007 1.588
Source: Researcher’s Calculation
23182 Osabuohien-Irabor Osarumwense and Nwosu Chinwe Rose/ Elixir Statistics 69 (2014) 23177-23184

Table 2: Comparison of MLM, MOM and LSM when α = 0.5 and β = 4.0
Parameters/ Sample Prob. Weighted Moments Max. Likelihood Moments Method Least Square Method
sizes (PWM) Method (MLM) (MOM) (LSM)
α β n ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD
0.5 4.0 30 0.777 6.174 1.097 0.081 4.000 0.838 0.082 4.000 0.836 1.141 0.308 2.205
50 0.816 5.410 0.984 0.066 4.000 0.868 0.066 4.000 0.867 1.761 1.004 3.271
Source: Researcher’s Calculation
Table 3: Comparison of MLM, MOM and LSM when α = 2.0 and β = 2.0
Parameters/ Sample Prob. Weighted Moments Max. Likelihood Moments Method Least Square Method
sizes (PWM) Method (MLM) (MOM) (LSM)
α β n ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD
2.0 2.0 70 4.298 3.169 1.733 5.066 2.420 1.743 4.952 2.440 1.696 4.412 1.009 1.701

Source: Researcher’s Calculation

However, the Probability Weighted moments (PWM) with a deviation of 0.354 was best when a sample size of 10 was used. Also, for
case where β > α, where sample sizes of 30 and 50 were used as shown in Table 2, the method of Moments (MOM) had the least
deviation of 0.836 and 0.867 respectively and therefore consider as the best estimating method for this sample sizes. However, the
MLM came close with a total deviation of 0.838 and 0.868 respectively. Results from Table 1- 3, show that the least Squares method
is consistently the worst estimating method for the sample sizes and parameter values. When a sample size of 70 was applied to a case
of α= β=2, the method of moments had the least total deviation (TD) of 1.696.
Table 4: Comparison of MLM, MOM and LSM when α = 1.0 and β = 1.0
Parameters/ Prob. Weighted Moments Max. Likelihood Moments Method Least Square Method
Sample sizes (PWM) Method (MLM) (MOM) (LSM)
α β n ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD
1.0 1.0 10 2.067 0.886 1.181 1.149 2.070 1.219 1.127 2.000 1.127 3.010 1.653 2.663
20 1.121 0.977 0.144 0.975 1.090 0.115 0.969 1.070 0.101 1.540 1.011 0.551
60 1.097 0.875 0.222 0.857 0.960 0.183 0.884 1.010 0.126 5.231 1.004 4.235
Source: Researcher’s Calculation

Table 5: Comparison of MLM, MOM and LSM when α = 1.0 and β = 2.0
Prob. Weighted Moments Max. Likelihood Moments Method
Parameters/ Sample sizes Least Square Method (LSM)
(PWM) Method (MLM) (MOM)
α β n ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD
1.0 2.0 10 1.246 2.216 0.354 1.429 2.330 0.594 1.415 2.300 0.565 2.940 1.407 2.236
20 1.161 2.011 0.166 1.126 1.910 0.171 1.151 1.990 0.156 1.840 1.017 1.331
60 1.096 2.321 0.256 0.859 1.920 0.181 0.861 1.980 0.149 3.452 1.005 2.959
Source: Researcher’s Calculation
Table 6: Comparison of MLM, MOM and LSM when α = 1.0 and β = 3.5
Parameters/ Prob. Weighted Moments Max. Likelihood Moments Method Least Square Method
Sample sizes (PWM) Method (MLM) (MOM) (LSM)
α β n ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD ƒ
„ _
… TD
1.0 3.5 10 0.098 3.713 0.152 1.246 3.800 0.211 1.077 3.400 0.105 1.765 2.400 1.079
20 1.239 3.255 0.309 0.954 5.100 0.503 0.973 3.510 0.022 2.643 1.018 2.352
60 1.144 3.866 0.248 0.913 3.350 0.129 0.906 3.200 0.179 3.119 1.005 2.831
Source: Researcher’s Calculation
Table 4, 5 and 6 shows the computational analysis of experiment II, as described by [9]. In Table 4, where values of α= β=1, the
method of Moments had the least minimum variance of 1.127, 0.101 and 0.126 for sample sizes of 10, 20, and 60 respectively. The
MLM was close with a TD of 1.219 and 0.183 when a sample size of 10 and 60 were used. The least Squares method has again shown
to be poor estimator for the two-parameter Weibull distribution. For α=1, β= 2 which modeled the Rayleigh distribution, the method
of moments (MOM) was also the best estimating method for sample sizes of 20 and 60 with a total deviation (TD) of 0.156 and 0.149
respectively as shown in Table 5. But for α=1, β= 3.5 as shown in Table 6, with least deviation of 0.105, and 0.022 the Method of
Moments (MOM) came best when a sample sizes of 10,and 20 were used respectively. However, the MLM was the best method for
estimating the parameter of the Weibull distribution when a sample size of 60 was used.
23183 Osabuohien-Irabor Osarumwense and Nwosu Chinwe Rose/ Elixir Statistics 69 (2014) 23177-23184

Conclusion
Based on the computational results from the sets of parameters and sample sizes considered in this study, and as used in the two
different experiments as shown in Tables 1, 2, 3, 4, 5 and 6, the following results shows that;
(i) The best method for estimating the two-parameter Weibull distribution, taking into consideration the total deviation as a
measurement for comparison, is the method of Moments (MOM).
(ii) The Method of Moments (MOM) is better than the Greenwood’s Probability Weighted Moments (PWM) as a Method of
estimating the two-parameter Weibull distribution.
(iii) Results also show that the Maximum likelihood method (MLM) and the probability Weighted Moments (PWM) can be a close
substitute when estimating the two-parameter Weibull distribution.
(iv) While the least squares method (LSM) is a poor estimating method for the two-parameter Weibull distribution as revealed in the
computational analysis.
References
[1] Weibull W. (1938). Investigate into strength properties of brittle material. Proc. The Royal Swedish Institute for Engr. #149.
[2] Weibull W. (1981). Fatique testing and analysis of results. New York: Pergamum press.
[3] Henderson A.J. (1965). The Weibull distribution, unpublished M.S. thesis, Lowa State University, Ames, lowa.
[4] Mann N.R., Schafer R.E. and Singpurwalla N.D. (1974). Methods for Statistical Analysis of Reliability and Life Data, John
Wiley and sons, New York.
[5] Ahmad Mahir Razali, Ali A. Salih and Asaad A. M., (2009). Estimation Accuracy of Weibull Distribution Parameters, Journal of
Applied Science Research 5(7) page 790 † 795.
[6] Walpole R.E. and R.H. Myers (1972). Probability and Statistics for Engineers and Scientist, the Macmillan international, New
York.
[7] Krishna B.M. (1992). Reliability Analysis and Prediction (A Methodology Oriented Treatment) Elservier Science Publisher
Amsterdam.
[8] Ansell J.I. and Phillip M.J. (1994). Practical Methods for Reliability DaTA Analysis, Oxford Science Publication.
[9] Dodson B. (1994). Weibull Analysis ASQC Press, Wisconsin.
[10] Cohen A.C. (1965). Maximum Likelihood Estimation in the Weibull Distribution Based on Complete and on Censored Sample,
Technometrics, Vol.7, No. 4, November, 1965.
[11] Harter H.L. and Moore A.H. (1965a). Point and Interval Estimators Based on order Statistics for the Scale Parameter of a
aWeibull Population with known Shape Parameter, Technometrics, Vol. 7, No.3, August 1965a.
[12] Harter H.L. and Moore A.H. (1967). Asymptotic variance and Covariances of Maximum likelihood Estimators, from Censored
Samples of the Parameters of Weibull and Gamma Population. Annals of Math. Statis., 38: 557 † 570.
[13] Thoman D.R. and Bain L.J. (1969). Inferences on the Parameters of the Weibull Distribution, Techno., 11:445 † 460.
[14] Boes D.C., Heo J.H., and Salas J.D. (1989). Regional Flood Quantile Estimation for a Weibull Model. Water Resour, Res., 25:
979 † 990.
[15] Heo J.H., Boes D.C, and Salas J.D. (1990). Regional Flood Frequency Modelling and Estimation, Water Resour., Paper No. 101,
Colorado State University, Fort Collins, Colorado.
[16] Li Y.M. (1994). A General Linear Regression Analysis Applied to the 3-Parameter Weibull Distribution, IEEE, Transction on
Reliability, 43: 255 † 263.
[17] Al-Fawzan M.A., (2000). Methods for Estimating the Parameters of the Weibull Distribution, King Abdulaziz city for Science
and Technology, Riyadh, Saudi Arabia.
[18] Murthy D.N., xie M., and Jiang R. (2004). Weibull Models, Wiley Series in Probability and Statistics, New York.
[19] Mazen Z. and Ammar S.S. (2009), Applied Mathematical Science Vol. 3, No. 11, 541 † 550.
23184 Osabuohien-Irabor
Irabor Osarumwense and Nwosu Chinwe Rose/ Elixir Statistics 69 (2014) 23177
23177-23184

[20]] Greenwood, J.A., Landwehr, J.M., and Matalas, N.C. (1979). Probability Weighted moments:
ents: definition and relation of
parameters of several distributions expressible in inverse form. Water Resources Research, 15, 1049 – 1054.
[21] Erwin Kreysizig, (2006). Advanced Engineering Mathematics (8th Edition), John Wiley Publication Ltd, India.
Biographies

Osabuohien-Irabor
Irabor Osarumwense, holds a Bachelor degree in Statistics and Masters degree in Applied
Statistics from Ambrose Alli University, Nigeria and Nnamdi Azikiwe University, Nigeria, respectively.
Currently, he is pursuing a doctorate degree in Statistics with specialty in Time Ser
Series Analysis. He is a
Lecturer at Ambrose Alli University, Ekpoma, Nigeria.

Dr. Nwosu Chinwe Rose,, holds a B.Sc Statistics, M.Sc Statistics from the prestigious University of Nigeria,
Nsukka, and a PhD Statistics from Abia State University, Nigeria. She is currently the Head of Department and
a Senior Lecturer in department of Statistics, Nnamdi Azikiwe University, Nigeria. She has held high
administrative positions in and outside the University system including the position of the Registrar, Federal
Polytechnic, Oko, Nigeria.

View publication stats