Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Evaluating the performance of memory type logarithmic estimators using simple random sampling

  • Shashi Bhushan ,

    Contributed equally to this work with: Shashi Bhushan, Anoop Kumar, Amani Alrumayh, Hazar A. Khogeer, Ronald Onyango

    Roles Conceptualization, Writing – review & editing

    Affiliation Department of Statistics, University of Lucknow, Lucknow, U.P., India

    ⨯
  • Anoop Kumar ,

    Contributed equally to this work with: Shashi Bhushan, Anoop Kumar, Amani Alrumayh, Hazar A. Khogeer, Ronald Onyango

    Roles Conceptualization, Methodology, Validation

    anoop.asy@gmail.com

    Affiliation Department of Statistics, Amity University, Lucknow, India

    ⨯
  • Amani Alrumayh ,

    Contributed equally to this work with: Shashi Bhushan, Anoop Kumar, Amani Alrumayh, Hazar A. Khogeer, Ronald Onyango

    Roles Writing – review & editing

    Affiliation Department of Mathematics, College of Science, Northern Border University, Arar, Saudi Arabia

    ⨯
  • Hazar A. Khogeer ,

    Contributed equally to this work with: Shashi Bhushan, Anoop Kumar, Amani Alrumayh, Hazar A. Khogeer, Ronald Onyango

    Roles Writing – review & editing

    Affiliation Department of Mathematical Sciences, College of Applied Sciences, Umm Al-Qura University, Makkah, Saudi Arabia

    ⨯
  • Ronald Onyango

    Contributed equally to this work with: Shashi Bhushan, Anoop Kumar, Amani Alrumayh, Hazar A. Khogeer, Ronald Onyango

    Roles Formal analysis, Funding acquisition, Investigation, Methodology, Resources, Writing – review & editing

    Affiliation Department of Applied Statistics, Financial Mathematics and Acturial Science, Jaramogi Oginga Odinga University of Science and Technology, Bondo, Kenya

    ⨯

Abstract

In survey research, various types of estimators have been suggested that consider only the current sample information to compute the unknown population parameters. Therefore, we utilize the past sample information along with the current sample information in the form of hybrid exponentially weighted moving averages to suggest the memory type logarithmic estimators for time-based surveys. The expression of the mean square error of the suggested estimators is determined to the first order of approximation. A relative comparison of the suggested estimators with the existing estimators is performed and efficiency conditions are obtained. Further, a simulation study is accomplished using a hypothetically rendered population and a real data illustration to improve the theoretical results. The results of the simulation study and the real data application exhibit that the consideration of past and current sample information meliorates the efficiency of the suggested estimators.

Citation: Bhushan S, Kumar A, Alrumayh A, Khogeer HA, Onyango R (2022) Evaluating the performance of memory type logarithmic estimators using simple random sampling. PLoS ONE 17(12): e0278264. https://doi.org/10.1371/journal.pone.0278264

Editor: Nadia Hashim Al-Noor, Mustansiriyah University - College of Science, IRAQ

Received: May 23, 2022; Accepted: November 11, 2022; Published: December 15, 2022

Copyright: © 2022 Bhushan et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All the data used for this study are provided within the paper.

Funding: The author(s) received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

1 Introduction

In sample surveys, usually, the information on an auxiliary variable, namely, population mean , population standard deviation Sx, population correlation coefficient ρxy, population coefficient of skewness β1(x), population coefficient of kurtosis β2(x) etc., are known in advance which are positively or negatively correlated with the study variable. The utilization of this information helps to enhance the efficiency of the suggested estimators. [1] introduced the ratio estimator when the auxiliary information is positively correlated with the study variable as (1) [2] investigated the classical product estimator when the auxiliary information is negatively correlated with the study variable as (2) where and are the sample means of the auxiliary variable x and study variable y respectively.

The mean square error (MSE) of the above ratio and product estimators is given as (3) (4) where f = 1/n, is the population mean of study variable y, Cx and Cy are the population coefficient of variations of the auxiliary and study variables, respectively.

[3] suggested a class of logarithmic type estimator as (5) where α is a duly opted scalar.

The MSE of the estimator up to the first order of approximation is given by (6) The minimum MSE at optimum value of α(opt) = −ρxy(Cy/Cx) is given by (7) Over the recent years, utilizing the current sample information, a large number of ratio, product, regression and exponential type estimators consist of different auxiliary information have been suggested by several authors. [4] utilized the known correlation coefficient for estimating the population mean. [5] suggested a general estimation procedure of population mean using known values of population parameters. [6] introduced estimation of entropy with application to test of fit under RSS. [7] suggested the estimation of proportion in presence of tie using RSS. [8] discussed interval estimation of P(X < Y) using RSS. [9] suggested estimation of a symmetric distribution function in multistage RSS, whereas [10] examined a new goodness of fit tests for the Cauchy distribution. [11] suggested improved modified ratio estimators using robust regression estimators. [12] suggested some efficient classes of estimators under stratified random sampling, whereas [13] suggested some novel class of estimators of population mean under ranked set sampling (RSS). [14, 15] developed some log type class of population mean estimators under RSS. [16] suggested quantile regression ratio type estimators of population mean using complete and partial auxiliary information, whereas [17] proposed quantile robust regression type estimators based on minimum covariance determinant for estimating population mean. [18] suggested two auxiliary variable based robust regression ratio type estimators of the population mean. [19] analyzed the covid-19 risk using an exponential estimators. [20] developed double sampling based robust ratio estimators in the presence of outliers. [21] considered an improved class of robust ratio estimators by using the minimum covariance determinant estimation. [22] investigated robust regression type estimators to estimate population mean using simple and two-stage sampling. [23] introduced robust ratio type estimators for estimating population mean under simple random sampling. [24] examined the Poisson regression ratio estimators for population mean using double sampling. [25] proposed compromised imputation based mean estimators using robust quantile regression method.

It has been well established that the consideration of auxiliary information helps to meliorate the efficacy of the estimators. Due to this reason, only the current sample at time t is utilized to obtain the knowledge of the study and auxiliary variables. The efficiency of the estimator can further be enhanced by utilizing the knowledge of the current and the previous samples such as from time t − 1, from time t − 2 and all this. It becomes essential in situations when a particular time interval based survey such as monthly or annual is performed. [26] introduced memory type estimators for population mean utilizing exponentially weighted moving averages (EWMA) for time scaled surveys under SRS, whereas [27] developed the memory type product and ratio estimators for population mean utilizing hybrid exponentially weighted moving averages (HEWMA) for time-based surveys under SRS. Later on, [28] suggested memory type product and ratio estimators of population mean in stratified sampling, whereas [29] investigated the memory type product and ratio estimators of population mean utilizing ranked-based sampling schemes. [30] studied a memory type shrinkage estimator of population mean in the quality control process. [31] suggested the ratio and product type estimators for estimating population mean for time-based survey. Recently, [32] suggested memory-type ratio and product estimators for population variance utilizing EWMA statistics for time-scaled surveys.

In the theory of survey sampling, the classical estimators are based on functional forms, like chains, exponential functions, and linear combinations. However, the construction of estimators using the logarithmic function is novel. Development of advance data analysis tools has enhanced the computational limitations of these functions. In the real world, logarithms are utilized for the measurement of earthquakes on Richter scales, acidity on pH scale, and sound on decibels. In addition, logarithms can also be utilized to measure exponential decay and growth, like interest rates, bacterial growth in a Petri dish, and radioactive decay in carbon dating. Some real life examples are also given in Section 6 and a numerical study is performed. Therefore, it is important to study logarithmic type estimators for the estimation of population parameters and this is the motivation behind the construction of the suggested class of estimators. This study aims to proffer a memory type logarithmic estimator based on the knowledge of the past as well as current sample information. To achieve this aim, the HEWMA statistic is considered in this study.

The plan of this article is described in few sections. In Section 2, HEWMA statistic is described along with the review of the existing memory type estimators along with their properties. The developed memory type logarithmic estimator is given in Section 3 along with its properties. In Section 4, a comparative study is performed by comparing the MSE expressions of the suggested and existing estimators. A simulation study is accomplished in Section 5 and a real data illustration is given in Section 6. The conclusion is outlined in Section 7.

2 Reviews of estimators based on HEWMA statistic

[33, 34], respectively, introduced the concept of the EWMA statistic and the HEWMA statistic for efficiently detecting the changes in the fundamental process mean in control charting. Let X1, X2, …., Xn be an independent and identically distributed random variables from which the sequence HE1, HE2, …, HEn are defined by utilizing the undermentioned recurrence expressions. (8) (9) where λi, i = 1, 2 are duly chosen scalars. Also, Et and HEt denote the EWMA and HEWMA statistic, respectively. The initial values of these statistics are considered as expected mean which can be estimated from a pilot survey. In this study, it is considered as zero i.e. HE0 = Et0 = 0. The mean and variance of HEt statistic were computed by [34]. Later on, [35] observed that the expressions of mean and variance derived by [34] were incorrect, then [35] provided the correct expressions of mean and variance of HEt statistic as (10) (11) where t ≥ 1, and are the mean and variance of the study variable. The limiting form of the variance is expressed by (12) (13) where .

[27] considered the HEWMA statistic to propose the memory type ratio and product estimators. The HEWMA statistic for the study variable is given by (14) (15) and the HEWMA statistic for the auxiliary variable is given by (16) (17) The statistic Zt and Qt are unbiased for population mean and respectively.

Utilizing the statistic Zt and Qt, [27] suggested the memory type ratio and product estimators under simple random sampling as (18) (19) To deduce the properties of the suggested estimators, we consider the following annotations.

, such that E(e0) = E(e1) = 0 and , E(e0e1) = fδ{(λ1λ2)2/(λ1 − λ2)2}ρxyCxCy.

The MSE of the ratio and product estimators to the first order of approximation are given by (20) (21) It is to be noted that the value of δ will be replaced by δ1 which is given by (22)

3 Suggested memory type logarithmic estimators

The logarithmic function has a few important characteristics and furnishes an essential role in various dimensions of scientific and non-scientific disciplines. In this article, we have developed the logarithmic relationship between the study variable Y and the auxiliary variable X for computing the population mean . The developed estimators would perform in situations where the variables Y and X are logarithmically associated. Motivated by the works of [26, 27], we propose the memory type logarithmic estimator using the HEWMA statistic as (23) where β is a duly opted scalar.

Using the notations described in the preceding section, we rewrite the estimator in terms of e′s as (24) Squaring and taking expectation both sides of (24), we get the MSE of the estimator to the first order of approximation as (25) Minimizing the with respect to β, we get (26) The minimum MSE at the optimum value of β is given by (27) The above MSE expression is important to perform the comparative study in the next section.

4 Comparative study

To derive the efficiency conditions, we perform a comparative study by comparing the minimum MSE of the developed memory type logarithmic estimator with the minimum MSE of the existing conventional and memory type estimators.

  1. By comparing the minimum MSE of the developed memory type logarithmic estimator from (27) with the minimum MSE of the usual mean estimator, we get (28)
  2. By comparing the minimum MSE of the developed memory type logarithmic estimator from (27) with the minimum MSE of the classical ratio estimator from (3), we get (29)
  3. By comparing the minimum MSE of the developed memory type logarithmic estimator from (27) with the minimum MSE of the classical product estimator from (4), we get (30)
  4. By comparing the minimum MSE of the developed memory type logarithmic estimator from (27) with the minimum MSE of the conventional logarithmic type estimator from (7), we get (31)
  5. By comparing the minimum MSE of the developed memory type logarithmic estimator from (27) with the minimum MSE of the memory type ratio estimator from (20), we get (32)
  6. By comparing the minimum MSE of the developed memory type logarithmic estimator from (27) with the minimum MSE of the memory type product estimator from (21), we get (33)

Under the above efficiency conditions, the suggested memory type logarithmic estimator dominates the existing conventional and memory type estimators. Furthermore, these conditions are substantiated with a simulated study and a real data illustration.

5 Simulation study

To illustrate the effectiveness of the suggested memory type logarithmic estimators against the existing conventional and memory type ratio and product estimators, we conduct a simulation study over an artificially rendered bivariate normal population of size N = 1000 units using R software with parameters , , σx = 6 and σy = 5. Using 40,000 iterations, the MSE and relative efficiency (RE) of various estimators (t) are computed by utilizing the undermentioned expressions: (34) (35) The performance of the developed memory type logarithmic estimator is studied by considering various amounts of correlation coefficients ρxy = ±0.1, ±0.3, ±0.5, ±0.7, ±0.9 and sample sizes n = 10, 20, 50, 100, 200, 300. The performance of the existing and suggested memory type estimators is also examined by considering various amounts of λ2 = 0.05, 0.25, 0.5, 0.75, 1 and a fixed value of λ1 = 0.1. The simulation results are revealed from Tables 1 to 4 for different values of ρxy, n and λ2. From the simulation results, we draw the interpretation in point-wise fashion as follows:

  1. From Tables 1 and 2, it has been observed that the suggested memory type logarithmic estimator dominates the usual mean estimator , classical ratio estimator , logarithmic estimator , memory type ratio estimator in terms of lesser MSE and greater RE for different amount of ρxy = +0.1, +0.3, +0.5, +0.7, +0.9.
  2. From Tables 3 and 4, it has been observed that the suggested memory type logarithmic estimator dominates the usual mean estimator , classical product estimator , logarithmic estimator , memory type product estimator in terms of lesser MSE and greater RE for different amount of ρxy = −0.1, −0.3, −0.5, −0.7, −0.9.
  3. From Table 1, it has been observed that the MSE of the suggested memory type logarithmic estimator decreases as the values of λ2 increase, whereas a reverse inclination can be seen from the results of Table 2 reported in terms of RE.
  4. The similar inclination can be seen from the results of Tables 3 and 4 reported, respectively, in terms of MSE and RE.
  5. From Tables 1 and 3, it is noticed that the MSE of the various estimators decreases as the sample size increase for each value of ρxy, whereas from Tables 2 and 4, it is seen that the RE increases as the sample size increases for each value of ρxy.
  6. From Tables 1 and 3, it is also observed that the MSE of various estimators decreases as the values of correlation coefficient increases, whereas from Tables 2 and 4, it is seen that the RE increases as the values of ρxy increases.
thumbnail
Download:
Table 1. MSE of existing and suggested estimators for various values of ρxy, λ and n.

https://doi.org/10.1371/journal.pone.0278264.t001

thumbnail
Download:
Table 2. RE of existing and suggested estimators for various values of ρxy, λ and n.

https://doi.org/10.1371/journal.pone.0278264.t002

thumbnail
Download:
Table 3. MSE of existing and suggested estimators for various values of ρxy, λ and n.

https://doi.org/10.1371/journal.pone.0278264.t003

thumbnail
Download:
Table 4. RE of existing and suggested estimators for various values of ρxy, λ and n.

https://doi.org/10.1371/journal.pone.0278264.t004

6 Real data illustration

In the present section, we exemplify the suggested estimators using a real data from [36], pp. 31. The real data is based on the density as the study variable Y and stiffness as the auxiliary variable X. The descriptive statistics are given as follows: N = 30, , , , and ρxy = 0.89. The execution of the real data illustration is given below.

  1. We select 25 samples each of size 5 from the above discussed population using simple random sampling without replacement, which are listed in Table 5 for ready reference.
  2. The usual mean estimator , usual ratio estimator , logarithmic type estimator , memory type ratio estimator and suggested memory type logarithmic estimator are computed using their respective expressions.
  3. Using λ1 = 0.10 and λ2 = 0.25, the HEWMA statistic is also tabulated for each sample and the time-varying MSE for different estimators is also listed in Table 5.
  4. The MSE for the memory type ratio estimator and the suggested memory type estimator is computed as 0.0692 and 0.0109, respectively, with the help of expressions (20) and (27) which shows that the suggested memory type logarithmic estimator dominates the existing estimators.
thumbnail
Download:
Table 5. Evaluation of the memory type logarithmic estimator consist of HEWMA statistic.

https://doi.org/10.1371/journal.pone.0278264.t005

7 Conclusion

In this article, we have developed a memory type logarithmic estimator by incorporating the past sample information along with the current sample information in the form of HEWMA statistic. The mathematical property, namely, MSE of the developed memory type logarithmic estimators is obtained to the first order of approximation. Further, a simulation study is performed over an artificially generated population and presented the simulation results from Tables 1 to 4 that confirm the dominance of the suggested memory type logarithmic estimators over the usual mean estimator, classical ratio and product estimators, the conventional logarithmic estimator envisaged by [3] and memory type ratio and product estimators suggested by [27]. Moreover, a real data illustration is also presented in support of the theoretical results and the results are reported in Table 5 that also show the dominance of the suggested estimators over their entrants. Therefore, the suggested memory type logarithmic estimator is enthusiastically recommended to the survey practitioners for time-based surveys.

In the forthcoming article, we intend to develop the suggested memory type logarithmic estimator using the HEWMA statistic under stratified random sampling and ranked set sampling.

Acknowledgments

The authors would like to express their heartfelt thanks to the hon’ble reviewers for their constructive suggestions and to the editor-in-chief.

References

  1. 1. Cochran WG. The estimation of the yields of cereal experiments by sampling for the ratio of grain to total produce. Journal of Agricultural Science. 1940;30(2):262–275.
  2. 2. Robson DS. Applications of multivariate polykays to the theory of unbiased ratio-type estimation. Journal of American Statistical Association. 1957;52(280): 511–522.
  3. 3. Bhushan S, Gupta R. Some log-type classes of estimators using auxiliary information. International Journal of Agricultural and Statistical Sciences. 2015;11(2):487–491.
  4. 4. Singh HP, Tailor R. Use of known correlation coefficient in estimating the finite population mean. Statistics in Transition. 2003;6(4):555–560.
  5. 5. Khoshnevisan M, Singh R, Chauhan P, Smarandache F. A general family of estimators for estimating population mean using known value of some population parameter(s). For East Journal of Mathematics. 2007;22: 181–191.
  6. 6. Zamanzade E, Mahdizadeh M. Entropy estimation from ranked set samples with application to test of fit. Revista Colombiana de Estadística. 2017;40(2):223–241.
  7. 7. Zamanzade E, Wang X. Proportion estimation in ranked set sampling in the presence of tie information. Computational Statistics. 2018;33:1349–1366.
  8. 8. Mahdizadeh M, Zamanzade E. Interval estimation of P(X<Y) in ranked set sampling. Computational Statistics. 2018;33:1325–1348.
  9. 9. Mahdizadeh M, Zamanzade E. Estimation of a symmetric distribution function in multistage ranked set sampling. Statistical Papers. 2019;61(2):851–867.
  10. 10. Mahdizadeh M, Zamanzade E. New goodness of fit tests for the Cauchy distribution. Journal of Applied Statistics. 2019;44(6):1106–1121.
  11. 11. Zaman T. Improvement of modified ratio estimators using robust regression methods. Applied Mathematics and Computation. 2019;348: 627–631.
  12. 12. Bhushan S, Kumar A, Singh S. Some efficient classes of estimators under stratified sampling, Communications in Statistics—Theory and Methods. 2021:1–30.
  13. 13. Bhushan S, Kumar A, Lone SA. On some novel classes of estimators under ranked set sampling. AEJ-Alexandria Engineering Journal. 2022;61:5465–5474.
  14. 14. Bhushan S, Kumar A. Log type estimators of population mean under ranked set sampling. Predictive Analytics using Statistics and Big Data: Concepts and Modeling. 2020;28:47–74.
  15. 15. Bhushan S, Kumar A. Novel log type class of estimators under ranked set sampling. Sankhya, B. 2022;84:421–447.
  16. 16. Shahzad U, Hanif M, Sajjad I, Anas MM. Quantile regression-ratio-type estimators for mean estimation under complete and partial auxiliary information. Scientia Iranica. 2020.
  17. 17. Shahzad U, Al-Noor NH, Afshan N, Alilah DA, Hanif M, Anas MM. Minimum Covariance Determinant-Based Quantile Robust Regression-Type Estimators for Mean Parameter. Mathematical Problems in Engineering. 2021:2021.
  18. 18. Zaman T, Dunder E, Audu A, Alilah DA, Shahzad U, Hanif M. Robust regression-ratio-type estimators of the mean utilizing two auxiliary variables: A simulation study. Mathematical Problems in Engineering. 2021;2021.
  19. 19. Zaman T, Sagir M, Sahin M. A new exponential estimators for analysis of COVID‐19 risk. Concurrency and Computation: Practice and Experience. 2022;34(10):e6806.
  20. 20. Zaman T, Bulut H. A simulation study: Robust ratio double sampling estimator of finite population mean in the presence of outliers. Scientia Iranica. 2021.
  21. 21. Bulut H, Zaman T. An improved class of robust ratio estimators by using the minimum covariance determinant estimation. Communications in Statistics-Simulation and Computation. 2022;51(5): 2457–2463.
  22. 22. Ali N, Ahmad I, Shahzad U, Al–Noor NH, Hanif M. Robust regression type estimators to determine the population mean under simple and two-stage random sampling techniques. Electronic Journal of Applied Statistical Analysis. 2022;15(2):357–373.
  23. 23. Zaman T, Bulut H, Yadav SK. Robust ratio‐type estimators for finite population mean in simple random sampling: A simulation study. Concurrency and Computation: Practice and Experience. 2022;e7273.
  24. 24. Koc H, Tanis C, Zaman T. Poisson regression-ratio estimators of the population mean under double sampling, with application to Covid-19. Mathematical Population Studies. 2022:1–15.
  25. 25. Anas MM, Huang Z, Shahzad U, Zaman T, Shahzadi S. Compromised imputation based mean estimators using robust quantile regression. Communications in Statistics-Theory and Methods. 2022;1–16.
  26. 26. Noor-ul-Amin M. Memory type estimators of population mean using exponentially weighted moving averages for time scaled surveys. Communications in Statistics-Theory and Methods. 2021;50(12):2747–2758.
  27. 27. Noor-ul-Amin M. Memory type ratio and product estimators for population mean for time-based surveys. Journal of Statistical Computation and Simulation. 2020;90(17):3080–3092.
  28. 28. Aslam I, Noor-ul-Amin M, Hanif M, Sharma P. Memory type ratio and product estimators under ranked-based sampling schemes. Communications in Statistics-Theory and Methods. 2021;1–23.
  29. 29. Aslam I, Noor-ul-Amin M, Yasmeen U, Hanif M. Memory type ratio and product estimators in stratified sampling. Journal of Reliability and Statistical Studies. 2020;1–20.
  30. 30. Singh GN, Bhattacharyya D, Bandyopadhyay A, Khalid M. Study of a memory type shrinkage estimator of population mean in quality control process. IEEE Access. 2021;9:161555–161564.
  31. 31. Chhaparwal P, Kumar S. Improving Efficiencies of Ratio-and Product-type Estimators for Estimating Population Mean for Time-based Survey. Journal of Reliability and Statistical Studies. 2022:325–340.
  32. 32. Qureshi MN, Tariq MU, Hanif M. Memory-type ratio and product estimators for population variance using exponentially weighted moving averages for time-scaled surveys. Communications in Statistics-Simulation and Computation. 2022:1–10.
  33. 33. Roberts S. Control chart tests based on geometric moving averages. Technometrics. 1959;1(3):239–250.
  34. 34. Haq A. A new hybrid exponentially weighted moving average control chart for monitoring process mean. Quality and Reliability Engineering International. 2013;29(7):1015–1025.
  35. 35. Haq A. A new hybrid exponentially weighted moving average control chart for monitoring process mean: discussion. Quality and Reliability Engineering International. 2016;33(7):1629–1631.
  36. 36. Yan X, Su X. Linear regression analysis: Theory and Computing. World Scientific; 2009.