 # Section III. Statistical Models

## A Difference-Cum-Exponential Type Efficient Estimator of Population Mean

### INTRODUCTION

Precision is the reason for most of the research work in statistical theory and computation. Auxiliary information usually enhances the precision in survey sampling. It can be used at the selection stage or in estimation or on both. In the literature, auxiliary information is used by authors to construct various efficient estimators of different population parameters. The ratio estimator works better when there is a positive correlation between study and auxiliary variables, while the product estimator performs well in case of the negative correlation. Generally, a linear regression estimator is better than both ratio and product estimators. The equality in the efficiencies is obtained if the intercept of the regression line of у on x is zero. Since the regression estimator either performs better or equal to ratio and product estimators, it is convenient to use linear regression estimator than classical ratio and product estimator to gain in the efficiency. Many authors, such as [1-16], etc., worked on ratio and product type estimator to get the better results of such type estimator. Further, to get a better result, researchers worked on difference and exponential type estimators and showed its usefulness. Some of them are [17-20], etc.

Motivated by Rao  and Ekpenyong and Enang , in this paper we propose a difference-cum-exponential type estimator and studied their properties. In section 13.2, we made a review of the existing estimator related to the proposed estimator. Later, we suggest a difference cum exponential type estimator and derive the expression for the bias and mean squared error and given the conditions for minimum mean square error. Last, we do practical analysis by a simulation study and real-life populations to illustrate the performance of the proposed estimator.

### NOTATIONS AND LITERATURE REVIEW

Let a finite population u = {U[,U1,..,UN} of size TV. A sample of size n is taken under simple random sampling without replacement technique. Let у be the study variable and x be the auxiliary variable, y, and xh i = 1,2,..,7V be the observations of у and x at /"' unit in the population. We will use the following notations in the subsequent work.

_ i ^ n

Sample mean for study variable у = — > v,, sample mean for auxiliary variable

n ‘—‘M

1 " — 1 x-1 N

x=~2^ population mean forstudy variable F =t; / , V/, population mean for

1 x-1 N У

auxiliary variable X = ~2^ X’’ P°Pu'atn rat*° Я = -=, population variance for y,

. 1 X-1 N ~ 7 7 1 X-1 'V ~7

S;:= ——-2^ ()’i -F)',populationvarianceforA,5; = ——(•*/ -ЗО’.рор-

1 x-1 N — — S, 5, S„.

ulation covariance S,v =-> (у,- - У)(jc, - X), Cv = —, C, = —. p = ——,

" y Y X SxSy

p,x=c,v = pc,cv, кух = hL,f=±x=IzL'

Sx R N n

Some estimators of population mean and their mean square error (MSE) from literature are presented here.

Sample mean у is the usual unbiased estimator of population mean with variance. The classical ratio estimator to estimate the population mean F w'hen the population mean of auxiliary variable .v is known, defined as The MSE of the classical ratio estimator is The expression for linear regression estimator is written as and MSE of regression estimator is given as In 1991, Bahl and Tuteja  suggested the ratio type exponential estimator as and obtained MSE as Rao  proposed a difference estimator as where h{ and h2 are suitable constants.

MSE of the defined estimator for optimum values of /г, and h2 is obtained as Further, Grover and Kaur  suggested an exponential type estimator as where bt and b2 are suitably chosen constants.

For the optimum values of fr, and b2, the MSE of Ток is derived as Ekpenyong and Enang  proposed two exponential-ratio type estimators as and The MSE of TPR for optimum values of a: and a2 are calculated as Also, the MSE of TPR1 for optimum values of 8, and S2 is given as ### SUGGESTED ESTIMATOR

To get a better and efficient estimate of population mean Y when X is known, we suggest a difference-cum-exponential type estimator as where к, t are suitably chosen constants and 8(* 0) is either constant or any parametric value such as pvv, Cx, p, etc.

To derive the bias and MSE of Ts, the error terms are defined as with £(г,) = 0, £(e2) = 0 and £(er) = X.C^, £(e2) = A.C?, £(£|£2) = Cxy.

Express Ts in terms of errors. Eq. (13.16) becomes Assuming |e,| < 1, i = 1,2, expand expression in terms of e’s and ignoring terms having e’s degree greater than two to get first order approximation, we have On simplifying, we get To get the bias of Ts to the first order of approximation, take expectation on both sides of Eq. (13.17), we have By squaring Eq. (13.17) on both sides and terminating the terms having e’s degree more than two, we have To get the MSE of Ts to the first order of approximation, take expectation on both sides of Eq. (13.19), we have or or where, To minimize MSE(Ts), we have to differentiate MSE(Ts) in Eq. (13.22) partially with respect to к and t, and equating to zero, we get and On solving simultaneous Eqs. (13.23) and (13.24) for к and t, the optimal values of к and t are obtained as The minimum MSE(Ts) can be obtained by using the optimum values of к and t from Eq. (13.25) in Eq. (13.22), as ### SPECIAL CASES

In this section, we do consider the different estimators with minimum of MSE(Ts) by assigning the different values to 8 to show the variability of the proposed estimator.

13.4.1 For 5 = 1

For 6 = 1, the proposed estimator T, becomes with where, 13.4.2 For 8 = Cx

For 6 = Cx, the estimator Ts becomes The optimum MSE of Ts{2) can be obtained as where, 13.4.3 For 6 = (Зу*

For 8 = pv.t, the estimator Ts becomes The minimum MSE of Tv(3) is where, 13.4.4 For 8 = p

For 8 = p, the estimator Ts becomes The minimum MSE of TsW where, ### EFFICIENCY COMPARISON

The proposed estimator Ts will be more efficient than other estimators if minimum MSE(Ts) is less than MSEs of the other estimators.

For efficiency comparison, first we express the MSEs of the estimators у, TR, TRE, TexpR, Trao, Tgk, TPRl and TPR2 in terms of £‘s, we have where, The estimator T, will be more efficient than estimators y, TR, TRE, TafR, Trao. TGK, 7>R1 and TPR2 whenever the following conditions satisfied, or or or or or or or or ### EMPIRICAL STUDY

In this section, we have considered some real-life populations to show the efficiency of the proposed estimator over existing estimators. The description of the seven different population parameters are stated in Table 13.1.

PRE of estimators are calculated by It is observed from Table 13.2 that the proposed estimator 7, for different values of 8 i.e. 8 = 1, C,-, руд and p performs efficiently than the usual unbiased estimator у, ratio estimator TR, linear regression estimator TRE, ratio type exponential estimator TexpR, Rao’s difference type estimator TRM>, Grover and Kaur’s estimator TGK, Ekpenyong and Enang’s estimators 7>R1 and TPR2, respectively, for different populations. It is worth to mention that the proposed estimator Ts{3) is more efficient than the other considered estimator(s) for Population 1 and 3. The estimator Ts{2) is efficient for Populations 2, 4, 5, 6 and 7. Also, it is worth to mention that correlation coefficients in all considered cases are different with least in Population-6 as p = 0.2522 to highest in Population-7 as p = 0.9410 still the results are good in all cases. This shows that the proposed estimator performs better irrespective of the correlation coefficient is mild, moderate or high.

TABLE 13.1

Population Parameters

 Population Source Study Variable у Auxiliary Variable x N n Y X c, C, P 1. Steel and Torrie  Logarithm of leaf burn in seconds Chlorine percentage 30 6 0.6860 0.8077 0.7001 0.7493 0.4996 2. Maddala  Per capita consumption of veal Price of veal per pound 16 4 7.6375 75.4343 0.2278 0.0986 0.6823 3. Kadilar and Cingi  Apple production amount Number of apple trees 204 50 966 26441 2.4739 1.7171 0.71 4. Srivstava, Srivastava and Khare  Weight of children Mid-arm circumference of children 55 30 17.08 16.92 0.12688 0.07 0.54 5. Cochran  Food cost Income of the family 10 4 101.1 58.8 0.1449 0.1281 0.6515 6. Cochran  Food cost Income of the family 33 16 27.49 72.545 0.3685 0.1458 0.2522 7. Murthy  Fixed capital Output of factory 80 20 11.264 51.826 0.750 0.354 0.9410

TABLE 13.2

Comparison of Estimators Through PRE with Respect to у

 Estimators PRE(., у) Source-1 Source-2 Source-3 Source-4 Source-5 Source-6 Source-7 100 100 100 100 100 100 100 92.10 167.59 201.55 141.13 158.82 104.49 298.97 133.26 187.10 201.65 141.16 173.74 106.79 873.21 133.04 133.06 159.32 128.50 161.43 106.45 163.52 139.79 188.07 210.89 141.18 174.06 107.22 875.32 142.72 188.16 213.41 141.19 174.17 107.24 876.48 149.47 158.72 176.87 128.03 150.25 106.05 361.28 154.55 189.45 242.00 141.21 175.01 107.32 925.56 181.95 1.177.37 407.76 597.24 227.90 1,138.57 3,204.70 205.89 2.691.97 256.52 1,338.97 478.13 2,223.96 6,277.01 256.57 2,521.67 1,014.39 602.37 193.86 2,036.63 5.744.44 239.09 1.529.56 515.67 856.18 293.71 2,017.26 3,393.64

### SIMULATION

Here we execute a simulation using R software to show the performance of proposed estimators over others by calculating percent relative efficiency of all estimators concerning у as defined in Eq. (13.39). For this, we have generated four hypothetical populations of size N = 5000 and sample size n = 200. Consider a linear relationship between study variable у and auxiliary variable л with slop 0.5 and drift 1. We also uses an error factor in the model Y = 1 + 0.5X by random t, which is independent of X, follows N(0,1). We have considered four cases, in the first three, the auxiliary variable X follows normal distribution as N(2,0.5), N(3,1) and N(4,1.5). In the latter case, X is taken from Uniform distribution L/(0,1). To make results more accurate, we have replicated it 10,000 times.

From Table 13.3, it is envisaged that the proposed class of estimators 7*(i)> TS(Y,, TS0) and 7j(4) performs efficiently in all the cases, which also supports the result obtained in numerical study in section 6.

### CONCLUSION

In this paper, we propose a difference-cum-exponential type estimator for the estimation of population mean Y of the study variable у when auxiliary information is available. The bias and MSE formulae of the proposed estimator are obtained and compared with that of the usual unbiased estimator y, classical ratio estimator TR, regression estimator TRE, exponential estimator TexpR, difference estimator Grover and Kaur’s estimator T(;K > Ekpenyong and Enang’s estimator TpR and TpR2- We have examined the performance of the proposed estimator by considering seven different real-life populations and also by a simulation study. It is interesting

TABLE 13.3

PRE of Estimators with Respect to у

 Estimators PRE(., у) Case-1 Case-2 Case-3 Case-4 100 100 100 100 98.3515 109.3995 132.2530 76.2459 105.3458 122.7273 152.1649 101.9720 105.3127 122.2299 148.0307 97.2554 105.4733 122.8233 152.2483 102.2855 105.4813 122.8404 152.2759 102.3278 105.9562 119.7086 139.3933 111.1496 105.5044 122.9139 152.4357 102.3529 770.9312 326.5865 222.2267 404.7270 1.346.4636 529.0413 351.6292 522.1004 1.126.0685 468.4702 319.4904 555.5838 1.376.1330 487.9484 295.5479 726.9252

to remark that the proposed class of estimators performs efficiently than the other considered estimators. Thus, our conclusion is to recommend the proposed estimator for future study.

### REFERENCES

• 1. Kadilar, C. and Cingi. H. 2004. Ratio estimators in simple random sampling. Applied Mathematics and Computation. 151(3):893—902.
• 2. Upadhyaya. L. N. and Singh, H. P. 1999. Use of transformed auxiliary variable in estimating the finite population mean. Biometrical Journal. 41(5):627—636.
• 3. Ekpenyong. E. J. and Enang, E. I. 2015. Efficient exponential estimator for estimating the population mean in simple random sampling. Hacettepe Journal of Mathematics and Statistics. 40(3):689—705.
• 4. Kadilar. C. and Cingi, H. 2006. Improvement in estimating the population mean in simple random sampling. Applied Mathematics Letter. 19(1):75—79.
• 5. Singh. H. P. and Solanki. R. S. 2012. An alternative procedure for estimating the population mean in simple random sampling. Pakistan Journal of Statistics and Operations Research. VIII(2):213-232.
• 6. Ray. S. K. and Sahai, A. 1980. Efficient families of ratio and product type estimators. Biometrika. 67(1):211—215.
• 7. Pandey, B. N. and Dubey, V. 1988. Modified product estimator using coefficient of variation of auxiliary variate. Assam Statistical Review. 2(2):64—66.
• 8. Reddy, V. N. 1973. On ratio and product method of estimation. Sankhya. Series B. 35:307-316.
• 9. Singh. H. P. and Espejo. M. R. 2003. On linear regression and ratio-product estimation of a finite population mean. The Statistician. 52(1):59—67.
• 10. Singh. H. P. and Tailor. R. 2005, Estimation of finite population mean with known coefficient of variation of an auxiliary character. Statistica. 65(3):301—313.
• 11. Singh. H. P. and Agnihotri, N. A. 2008. A general procedure of estimating population mean using auxiliary information in sample survey. Statistics in Transition-new series. 9, 71-87.
• 12. Tailor. R. and Sharma, B. A. 2009. Modified ratio-cum-product estimator of finite population mean using known coefficient of variation and coefficient of kurtosis. Statistics in Transition-New Series. 10, 15-24.
• 13. Singh, H. P. and Solanki, R. S. 2011. Generalized ratio product methods of estimation in survey sampling. Pakistan Journal of Statistics and Operation Research. 7(2):245—264.
• 14. Solanki, R. S„ Singh. H. P. and Pal, S. K. 2013. Improved estimation of finite population mean in sample survey. Journal of Advance Computing. 1, 70-78.
• 15. Kumar. S. 2013. Estimation in finite population surveys: Theory and applications. Journal of Modern Applied Statistical Method. 12(1)120-127.
• 16. Pal. S. K.. Singh. H. P.. Kumar, S. and Chatterjee. K. A. 2017. A family of efficient estimators of the finite population mean in simple random sampling. Journal of Statistical Computation and Simulation. 88(5):920-934.
• 17. Bahl, S„ Tuteja, R. K. 1991. Ratio and product type exponential estimators. Journal of Information and Optimization Sciences. 12, 159-164.
• 18. Gupta. S. and Shabbir, J. 2008. On the improvement in estimating the population mean in simple random sampling. Journal of Applied Sciences. 35(5 ):559—566.
• 19. Solanki. R. S„ Singh. H. P. and Rathour, A. 2012. An alternative estimator for estimating finite population mean using auxiliary information in sample surveys. ISRN Probability and Statistics. 1-14.
• 20. Yadav, S. K. and Kadilar. C. 2013. Efficient familiy of exponential estiamtors for the population mean. Hacettepe Journal of Mathematics and Statistics. 42(6):671—677.
• 21. Rao, T. J. 1991. On certain methods of improving ratio and regression estimators. Communications in Statistics-Theory and Methods. 20( 10):3325—3340.
• 22. Grover. L. K. and Kaur. P. 2011. An improved estimator of the finite population mean in simple random sampling. Model Assisted Statistics and Applications. 6( 1 ):47—55.
• 23. Steel, R.. Torrie, J. and Dickey D. 1960. Principles and procedures of statistics. Detroit. MI: McGraw-Hill Companies.
• 24. Maddala, G. S. 1977. Econometrics, economics handbook series. New York. NY: McGraw-Hills Publication Company.
• 25. Kadilar. C. and Cingi. H. 2005. A new estimator using two auxiliary variables. Applied Mathematics and Computation. 151. 893-902.
• 26. Srivastava, R. S„ Srivastava, S. P. and Khare. В. B. 1989. Chain ratio type estimator for ratio of two population means using auxiliary characters. Communication in Statistic Theory and Methods. 18( 10):3917—3926.
• 27. Cochran, W. G. 1977. Sampling techniques. 3rd ed. New York, NY: Wiley.
• 28. Murthy, M. N. 1967. Sampling theory and methods. Calcutta, India: Statistical Publishing Society.