© All Rights Reserved

0 views

© All Rights Reserved

- estimators1
- Lancaster University - MS Data Science Handbook
- State Estimation
- Analysis of the Mayfield method for estimating nest success.
- MA1022_TUTORIAL_5_6_2010
- 26362-29029-1-PB
- Cambridge Time Series Notes
- Half Rate in GSm
- AGarcia_HW8
- Personal Characteristics and There Effect on Juror Verdict
- C4BMobileRobots
- ENG5001Tute_07
- Midterm 2
- MLE of Utility Function by Using Stata
- Guidelines for Prediction of Co2 Corrosion in Oil and Gas Production Systems
- ife-kr-e-2009-003
- 2.8 Restricted and Unrestricted Regression
- Statistics
- Short Introduction to the GMM
- ARCH Models Definition

You are on page 1of 20

net/publication/267704005

CITATIONS READS

13 527

1 author:

Hanaa H Abu-Zinadah

University of Jeddah

26 PUBLICATIONS 101 CITATIONS

SEE PROFILE

All content following this page was uploaded by Hanaa H Abu-Zinadah on 14 December 2015.

Journal of Applied Sciences Research, 6(4): 358-376, 2010

© 2010, INSInet Publication

Hanaa H. Abu-Zinadah

P.O. Box 32691, Jeddah 21438, Saudi Arabia.

Abstract: This article is considered with the mixture model of exponentiated Pareto and exponential

distributions (MEPED). First, some properties of the model with some graphs of the density and hazard

function are discussed. Next, the maximum likelihood and Bayes methods of estimation are used for

estimating the parameters, reliability and hazard functions of the model under complete and type II

censored samples. An approximation form due to Lindley[1 8] is used for obtaining the Bayes estimates

under the squared error loss and LINEX (linear-exponential) loss functions. The performance of findings

in the article is showed by demonstrating some numerical illustrations through Monte Carlo simulation

study. Also, applications of mixed models are included.

Key words: Finite mixture; Statistical properties; M aximum likelihood; Bayesian analysis; Squared error

loss function; LINEX loss function; Lindley's approximation; Exponentiated Pareto

distribution; Exponential distribution.

maximum likelihood estimators of the parameters of

Mixtures of lifetime distributions occur when two two component mixture generalized gamma distribution.

different causes of failure are present, each with the Also, Ahmad et al.[1 ] deduced approximate Bayes

same parametric form of lifetime distributions. In estimation for mixtures of two W eibull distributions

recent years, the finite mixtures of lifetime distributions under type II censoring. On finite mixture of two

have proved to be of considerable interest both in component Gompertz distribution considered by Al-

terms of their methodological development and Hussaini et al.[3 ] . Several papers discussed normal

practical applications Titterington et al [3 6 ] . McLachlan mixtures, for example, Hosmer[1 3 ] and Holgersson and

and Basford [2 2 ] , Lindsay [1 9] , McLachlan and Peel[2 3 ] , Jorner[1 2 ] . Moreover, John [1 6 ] briefly outlines the use of

McCulloch and Searle [2 1 ] and Demidenko [7 ] . On the methods of moments and maximum likelihood in

characterizations of mixtures were studied by, Nassar estimating the parameters of two component gamma

and Mahmoud [25 ] , Nassar [2 4] , Gharib [1 0 ,1 1 ] and Ismail and distribution. Further, a mixture of two inverse W eibull

El Khodary[14 ] . Many authors interested with inferences distributions was studied by Sultan et al. [3 5 ]. Also,

on mixtures of exponential distributions among them Elsherpieny[8 ] estimated the parameters of mixed

Rider[2 8 ] , Everitt and Hand [9 ] , Al-Hussaini[2 ] , generalized exponentionally distributions.

[4 ]

Bartoszewicz and

The mixture of exponentiated Pareto and exponential distributions (MEPED) has its probability density function

(p.d.f.) as

(1.1)

where and are the mixing proportions, is the p.d.f. of exponential distribution (E(á)), which is

given by

(1.2)

Corresponding Author: Hanaa H. Abu-Zinadah, King Abdulaziz University, Girls College of Education, Department of

Mathematics, P.O. Box 32691, Jeddah 21438, Saudi Arabia.

358

J. Appl. Sci. Res., 6(4): 358-376, 2010

(1.3)

(1.4)

(1.5)

(1.6)

(1.7)

(1.8)

For more details about exponentiated Pareto distribution see Shawky and Abu-Zinadah (2006, 2008a, b, 2009).

A mixture is identifiable if there exist a one-to-one correspondence between the mixing distribution and a

resulting mixture. That is, there is a unique characterization of the mixture. Therefore, a mixture of exponentiated

Pareto and exponential distributions is identifiable.

This article is considered with finite mixture of exponentiated Pareto and exponential distributions. Statistical

properties of the model are derived in section 2. Maximum likelihood estimators of the parameters, reliability and

hazard functions are studied from complete and type II censored samples in section 3. In section 4, Bayes

estimators of the parameters, reliability and hazard functions are obtained under the squared error loss and LINEX

(linear-exponential) loss functions. Also, Monte Carlo simulation study are made in section 5. Concluding remarks

about comparisons between these estimators and the maximum likelihood ones are considered in section 6. Finally,

Applications of mixed models are presented in section 7.

2. Statistical Properties:

M oments and Some M easures: The rth moment about the origin, , of a MEPED with p.d.f.

(2.1)

where (2.2)

359

J. Appl. Sci. Res., 6(4): 358-376, 2010

is beta function.

The above closed form of allows us to derive the following forms of statistical measures for the MEPED:

- Mean, denoted by ì,

(2.3)

- Variance, denoted by ó 2 ,

(2.4)

where .

- Coefficient of variation:

(2.5)

- Skewness:

(2.6)

where

- Kurtosis:

(2.7)

where

M edian and M ode: The median of the mixture of exponentiated Pareto and exponential distributions cannot be

found in explicit form.

360

J. Appl. Sci. Res., 6(4): 358-376, 2010

(2.8)

Next, to find the mode for the mixture of exponentiated Pareto and exponential distributions, we differentiate

f(x) with respect to x, so (1.1) gives

(2.9)

Then by equating (2.9) with zero, we get mode. W e observe that the MEPED, may be unimodal with mode

can be found numerically by solving (2.9), or has reverse 'J' shaped (see Fig. 1).

Figure 1 shows some densities of E(á) and EP(è), components and their mixtures.

Figure 2 shows some hazard curves of E(á) and EP(è) and the hazard curves of their mixtures.

W e note about the HF of the MEPED that as the mixing proportion (p) decreased the HF curve closed to the

HF curve of EP(è). Also, as p increased the HF of the mixture closed to the HF of E(á).

is the time of the ith component to fail. This sample of failure times are obtained and recorded from a life test of

n items whose life times have the mixture of exponentiated Pareto and exponential distributions with p.d.f. and

c.d.f. given, respectively, by (1.1) and (1.4). The likelihood function in this case (see, Lawless (1982)) can be

written as:

, (3.1)

(3.2)

Assuming that the parameters á and è are unknown and ë, is known , the likelihood equations are given by

(3.3)

(3.4)

(3.5)

361

J. Appl. Sci. Res., 6(4): 358-376, 2010

f (x) f (x)

f (x) f (x)

f (x) f (x)

Fig. 1: Shapes of E(á ) and EP(è ),components and their MEPED with (p,á,è,ë).

The non-dashed curve for MEPED, The --------- curve for E(á), The - - - - curve for EP(è).

(3.6)

The solution of the two nonlinear likelihood equations (3.3) and (3.4) yields the maximum likelihood estimate

(MLE) of . The MLE’s of R(t) and h(t) are given, respectively, by (1.7) and

362

J. Appl. Sci. Res., 6(4): 358-376, 2010

h (x) h (x)

h (x) h (x)

h (x) h (x)

Fig. 2: Shapes of HF of E(á ) and EP(è ), components and the HF of MEPED with (p,á,è,ë).

The non-dashed curve for HF of the mixture, The _ _ _ _ curve for HF of E(á ), The ---- curve for HF

of EP(è).

4. Bayes Estimation: Let A and È be independent random variables, where ë is known. The joint prior density

W e choose the random variables A, È, to follow gamma distributions with shape parameter

363

J. Appl. Sci. Res., 6(4): 358-376, 2010

(4.1)

It is well known that the posterior density function of given the observations, denoted by is

given by

(4.2)

It then follows that the Bayes estimator of a function of the parameter is given by the ratio

(4.3)

where is given by (3.1), by (4.1) and Ù is the region in the AÈ plane on which the posterior

density is positive.

The ratio of the integrals (4.3) may thus be approximated by using a form due to Lindley [1 8 ] which reduces,

in the case of two parameters, to the form

(4.4)

where For

is as given by (3.1).

364

J. Appl. Sci. Res., 6(4): 358-376, 2010

For

and

Now, we apply Lindley's form (4.4), we first obtain the elements which can be obtained as

and , (4.5)

where (4.6)

(4.7)

(4.8)

(4.9)

Furthermore,

(4.10)

(4.11)

(4.12)

(4.13)

365

J. Appl. Sci. Res., 6(4): 358-376, 2010

(3.5).

Using prior density function (4.1), we obtain

(4.14)

In Bayesian estimation, we consider two types of loss functions. The first is the squared error loss function

(quadratic loss) which is classified as a symmetric function and associates equal importance to the losses for

overestimation and underestimation of equal magnitude. The second is the LINEX (linear-exponential) loss function

which is asymmetric, was introduced by Varian[3 7 ] . These loss functions were widely used by several authors;

among of them Rojo[29 ] , Basu and Ebrahimi[5 ] , Pandey[2 6] , and Soliman [3 4 ] .

The quadratic loss function for Bayes estimate of a parameter â, say, is the posterior mean assuming that

(4.15)

where The sign and magnitude of the shape parameter c reflects the direction and degree of

asymmetry respectively. (If the overestimation is more serious than underestimation, and vice-versa). For

c closed to zero, the LINEX loss is approximately squared error loss and therefore almost symmetric.

The posterior expectation of the LINEX loss function Equation (4.15) is

(4.16)

where denoting posterior expectation with respect to the posterior density of â. By a result of Zellner

(1986), the (unique) Bayes estimator of â, denoted by under the LINEX loss is the value which

minimizes (4.16), is given by

(4.17)

4.1 Estimation under Squared Error Loss Function: Using Equations (4.5)-(4.14), the Bayes estimate of the

function relative to squared error loss function given by Equation (4.4), can be shown to be

(4.18)

366

J. Appl. Sci. Res., 6(4): 358-376, 2010

where

All functions of the right-hand side of Equation (4.18) are to be evaluated at the maximum likelihood estimates

of á and è, when ë is known.

4.1.1 Bayes Estimation of the Vector of Parameters: The two parameters á and è, when ë is known, can be

approximately estimated using Lindley's approximation from (4.18), and their estimates are obtained as follows:

(4.19)

(4.20)

4.1.2 The Bayes Estimate of the RF: Set in (4.18), where R(t) is given as in (1.7). Then

(4.21)

where

4.1.3 The Bayes Estimate of the HF: Set in (4.18), where h(t) is given in (1.8). Then

(4.22)

where

367

J. Appl. Sci. Res., 6(4): 358-376, 2010

4.2 Estimation Under LINEX Loss Function: On the basis of the LINEX loss function (4.17), the Bayes estimate

(4.23)

where

(4.24)

4.2.1 Bayes Estimation of the Vector of Parameters: The two parameters á and è, when ë is known, can be

approximately estimated using Lindley's approximation from (4.18), and their estimates are obtained as follows:

(4.25)

(4.26)

368

J. Appl. Sci. Res., 6(4): 358-376, 2010

4.2.2 The Bayes Estimate of the RF: Set in (4.18), where R(t) is given in (1.7). Then

(4.27)

where

4.2.3 The Bayes Estimate of the HF: Set in (4.18), where h(t) is given in (1.8). Then

(4.28)

where

369

J. Appl. Sci. Res., 6(4): 358-376, 2010

5. Simulation Study: W e obtained, in the above Sections, Bayesian and non-Bayesian estimates of the vector

parameters è reliability R(t) and hazard h(t) functions of the MEPED. W e adopted the squared error loss and

LINEX loss functions. The MLE’s are also obtained.

In order to assess the statistical performances of these estimates, a simulation study is conducted. The

computations are carried out for censoring percentages of 80% and 100% (complete sample case), for each sample

size. The root mean square errors (RM SE’s) using generated random samples of different sizes are computed for

each estimator. The random samples are generated as follows:

1. For given values of the prior parameters and generate a random values for á and è from the

2. Using á and è, obtained in step (1), generate random samples of different sizes: n=10, 20 and 40 from a

MEPED as given by (1.1). The computations are carried out for such sample sizes and censored samples of sizes:

8, 16 and 32, respectively.

(3.3) and (3.4). The estimators and of the functions R(t) and h(t) are then computed at some

values t 0

4. The Bayes estimates relative to squared error loss, and are computed, using (4.18) after considering

the appropriate changes according to Subsections (4.1.1), (4.1.2) and (4.1.3). Also, the Bayes estimates relative to

LINEX loss computed, using (4.18) after considering the appropriate changes according to

5. The above (2-4) steps are repeated 1000 times and the biases and the root mean square errors are computed

for different sample sizes n and censoring sizes r where the had-symbol ^ stands for an estimate

and

The computational (our) results were computed by using M athematica 5.0. In all above cases the prior

parameters chosen as d 1=2 and d 2=4 which yield the generated values of á=1.20777 and è=2.55506 (as the true

values). The true values of R(t) and h(t) when t=t o=0.5, are computed to be R(0.5)=0.743149 and h(0.5)=0.582644.

The bias (first entries) and RMSE’s (second entries) are displayed in Tables 1-4. The computations are achieved

under complete and censored samples.

Tables 1, 2, 3 and 4 contains estimated biases and RMSE’s of MLE’s, Bayes estimators under quadratic and

LINEX loss functions of á, è, R(t) and h(t) respectively.

370

J. Appl. Sci. Res., 6(4): 358-376, 2010

Table 1: Estim ated biases (first entries) and RM SE’s (second entries) of various estim ators of á for different sam ple sizes.

n r

10 8 -0.294552 0.890992 0.541597

--------------------------------------------------------------------------------------------------------------------------

0.680967 1.36858 0.767958

-----------------------------------------------------------------------------------------------------------------------------------------------------

10 -0.286597 0.84836 0.534034

--------------------------------------------------------------------------------------------------------------------------

0.616013 1.30924 0.745842

20 16 -0.411872 0.648193 0.471918

--------------------------------------------------------------------------------------------------------------------------

0.532008 0.999404 0.609553

------------------------------------------------------------------------------------------------------------------------------------------------------

20 -0.382741 0.665582 0.449921

--------------------------------------------------------------------------------------------------------------------------

0.495303 1.06784 0.617458

40 32 -0.446398 0.544141 0.395552

--------------------------------------------------------------------------------------------------------------------------

0.502533 0.88921 0.515145

------------------------------------------------------------------------------------------------------------------------------------------------------

40 -0.418157 0.563135 0.399096

--------------------------------------------------------------------------------------------------------------------------

0.473756 0.931012 0.512992

Table 2: Estim ated biases (first entries) and RM SE’s ( second entries ) of various estim ators of è for different sam ple sizes.

n r

--------------------------------------------------------------------------------------------------------------------------

0.952666 1.41808 0.976376

------------------------------------------------------------------------------------------------------------------------------------------------------

10 -0.317354 1.1974 0.947593

--------------------------------------------------------------------------------------------------------------------------

0.84756 1.35159 0.960699

20 16 -0.447461 0.825674 0.758241

--------------------------------------------------------------------------------------------------------------------------

0.807332 0.992831 0.778909

------------------------------------------------------------------------------------------------------------------------------------------------------

20 -0.28316 0.791217 0.739698

--------------------------------------------------------------------------------------------------------------------------

0.739949 0.959501 0.763499

40 32 -0.331014 0.425055 0.517037

--------------------------------------------------------------------------------------------------------------------------

0.691088 0.530619 0.533921

------------------------------------------------------------------------------------------------------------------------------------------------------

40 -0.206503 0.411675 0.501686

--------------------------------------------------------------------------------------------------------------------------

0.603667 0.514166 0.524908

Table 3: Estim ated biases (first entries) and RM SE’s ( second entries ) of various estim ators of R(t) for different sam ple sizes.

n r

--------------------------------------------------------------------------------------------------------------------------

0.076901 0.947203 0.172072

------------------------------------------------------------------------------------------------------------------------------------------------------

10 0.0242503 -0.154821 0.0557544

--------------------------------------------------------------------------------------------------------------------------

0.076351 0.917943 0.159091

20 16 0.0370652 -0.185319 0.0242676

--------------------------------------------------------------------------------------------------------------------------

0.0648364 0.73519 0.12628

------------------------------------------------------------------------------------------------------------------------------------------------------

20 0.0405856 -0.150803 0.023912

--------------------------------------------------------------------------------------------------------------------------

0.0622027 0.65092 0.134988

371

J. Appl. Sci. Res., 6(4): 358-376, 2010

Table 3: Continue

40 32 0.04958 -0.163039 -0.00431204

--------------------------------------------------------------------------------------------------------------------------

0.0617596 0.609119 0.0767005

------------------------------------------------------------------------------------------------------------------------------------------------------

40 0.0516474 -0.186897 -0.00983854

--------------------------------------------------------------------------------------------------------------------------

0.062659 0.683589 0.0837048

Table 4: Estim ated biases (first entries) and RM SE’s ( second entries ) of various estim ators of h(t) for different sam ple sizes.

n r

--------------------------------------------------------------------------------------------------------------------------

0.154876 1.09067 0.251928

------------------------------------------------------------------------------------------------------------------------------------------------------

10 -0.0477734 -0.0635128 0.177017

--------------------------------------------------------------------------------------------------------------------------

0.158033 1.01581 0.25978

20 16 -0.0633164 0.108436 0.14551

--------------------------------------------------------------------------------------------------------------------------

0.13375 0.796672 0.223293

------------------------------------------------------------------------------------------------------------------------------------------------------

20 -0.0746503 0.114652 0.14623

--------------------------------------------------------------------------------------------------------------------------

0.129789 0.794805 0.224273

40 32 -0.0909308 0.175892 0.125347

--------------------------------------------------------------------------------------------------------------------------

0.12506 0.617904 0.19323

------------------------------------------------------------------------------------------------------------------------------------------------------

40 -0.0976032 0.198083 0.117098

--------------------------------------------------------------------------------------------------------------------------

0.127798 0.661755 0.186228

Figure 3 shows boxplot of the bias of the estimates Also, Table 2 shows that the Bayes estimates under the

of the vector parameters è reliability R(t) and hazard quadratic loss function often have the smallest

h(t) functions of the MEPED. estimated RMSE’s as compared with the Bayes

estimates under LINEX loss function or MLE’s in the

6. Concluding Remarks: In this paper we have case of large sample size for both complete and

presented the Bayesian and maximum likelihood censored samples. Further, the boxplot in Figure 3

estimates of the vector parameters è , reliability R(t) shows that among the simulated estimates there is not

and hazard h(t) functions of the lifetimes follow the outlier for estimating.

MEPED. The estimation are conducted on the biases From the previous observations, the estimation

and RMSE’s of complete and type-II censored samples. from a finite MEPED data is possible and flexible

Bayes estimators, under squared error loss and LINEX using non-Bayesian approach to estimate parameter á,

loss functions, are derived in approximate forms by reliability function R(t) and hazard function h(t) as

using Lindley’s method . The MLE’s are also obtained. shown from this article. W e also, recommend to use

Our observations about the results are stated in the Bayes estimates under quadratic loss function to

following points: estimate the parameter è especially for large samples.

Table 1, Table 3 and Table 4 show that the Furthermore, we recommend non-Bayesian approach to

maximum likelihood estimates usually have the estimate the parameter è especially for small and

smallest estimated RMSE’s as compared with the median sample size.

RMSE’s of estimates under the quadratic loss function

or LINEX loss function. This is True for both complete 7. Applications: Mixture models play a vital role in

and censored samples. It is immediate to note that many practical applications. For example, direct

RM SE’s decrease as sample size increases. On the applications of finite mixture models are in fishers

other hand Table 2 shows that the MLE’s often have r e s e a r c h , e c o n o m ic s , m e d ic in e , p syc h o l o g y ,

the smallest estimated RMSE’s as compared with the palaeoanthropology, botany, agriculture, zoology, life

Bayes estimates under LINEX loss function or testing and reliability, among others. Indirect

quadratic loss function in the case of small and median applications include outliers, Gaussian sums, cluster

sample size for both complete and censored samples. analysis, latent structure models, modeling prior

372

J. Appl. Sci. Res., 6(4): 358-376, 2010

373

J. Appl. Sci. Res., 6(4): 358-376, 2010

Fig. 3: Boxplot of the bias of the estimates.

374

J. Appl. Sci. Res., 6(4): 358-376, 2010

densities, empirical Bayes method and nonparametric 11. Gharib, M., 1996. Characterizations of the

(kernel) density estimation. In many applications, the exponential distribution via mixing distributions.

available data can be considered as data coming from Microelectron. Reliab., 36(3): 293-305.

a mixture population of two or more distributions. This 12. H olgersson, M . and U. Jorner, 1979.

idea enable us to mix statistical distributions to get a D ecomposition of a mixture into normal

new distribution carrying the properties of its components: A review. Int. J. Biomed. Comput., 9:

components. For example, a mixture of two gamma 367-392.

distributions applied to rheumatoid arthritis was 13. Hosmer, D.W ., 1973. On MLE of the parameters

discussed by Masuyama (1977). in a mixture of two normal distributions when the

sample size is small. Commun. Statist., 1(3): 217-

ACKNOW LEDGM ENTS 227.

14. Ismail, S.A. and I.H. El-Khodary, 2001.

I would like to thank the Deanship of Scientific Characterization of mixtures of exponential family

Research of King Abdulaziz University for their distributions through conditional expectation.

supported to the researcher. Also, I would like to thank Annual Conference on Statistics and Computer

Prof. Dr. Shawky Ahmed for his consultant and helpful Modeling in Human and Social Sciences, 13: 64-

comments which improved the presentation of the 73.

paper. 15. Jaheen, Z.F., 2005. On record statistics from a

REFERENCES mixture of two exponential distributions. Journal of

Statistical Computation and Simulation, 75(1): 1-

1. Ahmad, K.E., H.M. Moustafa and A.M . Abd-El-

11.

Rahman, 1997. Approximate Bayes estimation for

16. John, S., 1970. On identifying the population of

mixtures of two W eibull distributions under type

origin of each observation in a mixture of

II censoring. Journal of Statistical Computation and

observations from two gamma populations.

Simulation, 58: 269-285.

Technometrics, 12(3): 565-568.

2. Al-Hussaini, E.K., 1999. Bayesian prediction under

17. Lawless, J.F., 1982. Statistical Models and

a mixture of two exponential components model

Methods for Lifetime Data. W iley, New York.

based on type I censoring. Journal of Applied

18. Lindley, D.V., 1980. Approximate Bayesian

Statistical Science, 8: 173-185.

3. Al-Hussaini, E.K., G.R. Al-Dayian and S.A. methods. Trabajos de Estadistica, 31: 223-245.

Adham, 2000. On finite mixture of two-component 19. Lindsay, B.G., 1995. Mixture Models: Theory,

Gompertz lifetime model. Journal of Statistical Geometry, and Applications. Instit. Math. Statist.,

Computation and Simulation, 67: 1-20. Hayward, CA.

4. Bartoszewicz, J., 2002. Mixtures of exponential 20. Masuyama, M., 1977. A mixture of two gamma

distributions and stochastic orders. Statistics and distributions applied to rheumatoid arthritis. Rep.

Probability Letters, 57: 23-31. Stat. Appl. Res., JUSE, 24(2): 28/82-31/85.

5. Basu, A.P. and N. Ebrahimi, 1991. Bayesian 21. McCulloch, C.E. and S.R. Searle, 2001.

approach to life testing and reliability estimation Generalized, Linear, and Mixed Models. W iley,

using asymmetric loss function. J. Stat. Plan. New York.

Infer., 29: 21-31. 22. McLachlan, G.J. and K.E. Basford, 1988. Mixture

6. Calabria, R. and G. Pulcini, 1996. Point estimation Models: Inferences and Applications to Clustering.

under asymmetric loss functions for lift-truncated Marcel Dekker, New York.

exponential samples. Commun. Statist.-Theory 23. McLachlan, G.J. and D. Peel, 2000. Finite Mixture

Meth., 25(3): 585-600. Models. W iley, New York.

7. Demidenko, E., 2004. Mixed Models: Theory and 24. Nassar, M.M., 1988. Two properties of mixtures of

applications. W iley, New Jersey. exponential distributions. IEEE Transactions on

8. Elsherpieny, E.A., 2007. Estimaton of parameters Reliability, 37(4): 383-385.

of mixed generalized exponentionally distributions 25. Nassar, M.M. and M.R. M ahmoud, 1985. On

from censored type I samples. Journal of applied characterizations of a mixture of exponential

sciences Research, 3(12): 1696-1700. distributions. IEEE Transactions on Reliability,

9. Everitt, B.S. and D.J. Hand, 1981. Finite Mixture 34(5): 484-488.

Distributions. Chapman and Hall, London. 26. Pandey, B.N., 1997. Testimator of the scale

10. Gharib, M., 1995. Two characterizations of a parameter of the exponential distribution using

gamma mixture distribution. Bull. Austral. Math. LINEX loss function. Commun. Statist.-Theory

Soc., 52: 353-358. Meth., 26(9): 2191-2202.

375

J. Appl. Sci. Res., 6(4): 358-376, 2010

27. Radhakrishna, C., A.V. Dattatreya Rao and 33. S ha w ky, A .I. and A bu-Zina d a h, 2009.

G .V .S.R. Anjaneyulu, 1992. Estimation of Exponentiated Pareto distribution: different method

parameters in a two-component mixture generalized of estimations. IJCMS, 4(14): 677-693.

gamma distribution. Commun. Statist.-Theory 34. Soliman, A.A., 2000. Comparison of LINEX and

Meth., 21(6): 1799-1805. quadratic Bayes estimators for the Rayleigh

28. Rider, P.R., 1961. The method of moments applied distribution. Commun. Statist.-Theory Meth., 29(1):

to amixture of two exponential distributions. Ann. 95-107.

Math. Statist., 32: 143-147. 35. Sultan, K.S., M.A. Ismail and A.S. Al-Moisheer,

29. Rojo, J., 1987. On the admissibility of c +d with 2007. Mixture of two inverse W eibull distributions:

respect to the LINEX loss function. Commun. properties and estimation. Computational Statistics

Statist.-Theory Meth., 16(12): 3745-3748. and Data Analysis, 51: 5377-5387.

30 Shawky, A.I. and Abu-Zinadah, 2006. General 36. Titterington, D.M., A.F.M. Smith and U.E. Makov,

recurrence relations and characterizations of certain 1985. Statistical Analysis of Finite Mixture

distributions based on record values. JATA, 2(2): Distributions. W iley, London.

149-159. 37. Varian, H.R., 1975. A Bayesian Approach to Real

31. Shawky, A.I. and Abu-Zinadah, 2008a. General Estate Assessment. Amsterdam: North Holland,

recurrence relations and characterizations of certain 195-208.

distributions based on order statistics. JSTA, 7(1): 38. Zellner, A., 1986. Bayesian estimation and

93-117. prediction using asymmetric loss functions. J. A.

32. Shawky, A .I. and Abu-Zinadah, 2008b. S. A., 81: 446-451.

Characterizations of the exponentiated Pareto

distribution based on record values. AMS, 2(26):

1283-1290.

376

- estimators1Uploaded byBalaganesh Nallathambi
- Lancaster University - MS Data Science HandbookUploaded bypaul
- State EstimationUploaded byTesfahun Girma
- Analysis of the Mayfield method for estimating nest success.Uploaded byArthur Kalverboer
- MA1022_TUTORIAL_5_6_2010Uploaded byNuwan Bandara
- 26362-29029-1-PBUploaded byLusiana
- Cambridge Time Series NotesUploaded bytemoujin
- Half Rate in GSmUploaded bysberisha
- AGarcia_HW8Uploaded byAndrés García Arce
- Personal Characteristics and There Effect on Juror VerdictUploaded byaxpdjcmu
- C4BMobileRobotsUploaded byMichael Davis
- ENG5001Tute_07Uploaded byJustin
- Midterm 2Uploaded byWeiWei
- MLE of Utility Function by Using StataUploaded bymeriem
- Guidelines for Prediction of Co2 Corrosion in Oil and Gas Production SystemsUploaded byMartín Diego Mastandrea
- ife-kr-e-2009-003Uploaded byAnonymous J37Ljisq
- 2.8 Restricted and Unrestricted RegressionUploaded byNguyen Hoang Thao
- StatisticsUploaded byFrancisco De Real Onde
- Short Introduction to the GMMUploaded byJuanfer Subirana Osuna
- ARCH Models DefinitionUploaded byHuy Hai Vu
- Reliability Estimation in Inverse Rayleigh Distribution using Precautionary Loss FunctionUploaded byIWNEST
- 2010_AS_Q1Uploaded byBob
- ESTRUCTURAS HIDRAULICAS HYFRAN.docxUploaded byFavio Espinoza Pari
- Extrem UmUploaded byRoger Asencios Nuñez
- 1-s2.0-S1877750317300443-mainUploaded byDhika Ira Riswana
- JohnstonUploaded byyash231987
- Find Me If You Can: Improving Geographical Prediction with Social and Spatial ProximityUploaded byJonathan Chang
- 2Uploaded byNandN
- Recreation- HS Chamang 27-9-2008Uploaded byAwang Noor Abd Ghani
- st-infUploaded byJM Palana

- act-1985-vol19-iss08-poppel.pdfUploaded byDickson Siele
- A_Class_of_Weibull_Mixture_Distributions (1).pdfUploaded byDickson Siele
- A Study on Mixture of Exponentiated Pareto and Exponential Distributions.pdfUploaded byDickson Siele
- A mixture of exponentials distribution.pdfUploaded byDickson Siele
- 2002 Formulae and TablesUploaded byDaizy Rono
- 19-03.pdfUploaded byDickson Siele
- Edu 2009 Spring Exam FmUploaded byDickson Siele
- 59Uploaded bySilviu
- 342-Intro-to-Risk.pdfUploaded byNitin Nath Singh
- Mixture Distributions Loss ModelsUploaded byDickson Siele
- CGUIDEUploaded byJustine Powell
- Colin Raban Risk ManagementUploaded byDickson Siele
- Variance Derivation for the Overdispersion Parameter-Theses Final DocUploaded byDickson Siele
- 9781420035889Uploaded byDickson Siele
- notes12_aUploaded byDickson Siele

- Linear Regression Correlation AnalysisUploaded bySalami Jama
- Studenmund_Ch14_v2Uploaded byHasanul Banna
- 366Uploaded byTony
- Berger Could Fisher Jeffreys and Neyman Have Agreed on Testing With CommentaryUploaded bymike
- t SeriesUploaded byFrancisco Faner
- B. J. Winer-Statistical Principles in Experimental Design-NY c._ McGraw-Hill Book Company (1962)Uploaded byniki098
- ECN-720, 603 Econometric Theory Spring 2014- OutlineUploaded bySumera Bokhari
- Relevant CourseworkUploaded byFernando ML
- 3-Seasonality.pdfUploaded byjustinlondon
- ps2_akUploaded byeurdone
- MCA First Year Syllabus Manipur UniversityUploaded byMeitei opubunq
- 27Uploaded byInayah Gorjess Pazzer
- 76197355 Ken Black QA 5th Chapter15 SolutionUploaded byManish Khandelwal
- PUBH 6000 Statistical Inference - Spring 2017 - Handout (4)Uploaded byaastha93
- Demand Estimation (1)Uploaded byAnonymous FmpblcLw0
- PRINT2Uploaded byAnonymous rNmOyeE
- Panel Data Analysis Sunita AroraUploaded bylukgv,h
- LECTURE9-NOVEMBER18&19(1)Uploaded byDevon Moody
- #3 - Current Practices in Shelf Life Estimation - SchwenkeUploaded byعبدالعزيز بدر
- 74853-ID-pengaruh-volatilitas-laba-perataan-laba (1).pdfUploaded byAmy Kurniasari
- Chapter 01Uploaded byHannah Watson
- MLE Ordinal Model R PackageUploaded byApoorva Khandelwal
- 01 - Nature - Lucid-dreams - Reporting-ChecklistUploaded byLucaBertini
- Chapter 05 - MulticollinearityUploaded bymgayanan
- GaiseCollege FullUploaded bylondonbradley
- hasil lat2Uploaded bylucas17
- Answer to Econometric for Finance.pdfUploaded byJame
- iahiuahuisahUploaded byPedro Pires
- B. Statistics (Final)Uploaded bysumu_821
- Annova TestUploaded bykeertheswaran