Volume 9, Issue 11, November – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Consistency of Stratified Random Sampling

Estimators in Repetive Sampling
Adekunle Nurudeen MASOPA1*; Adisa Anthony AGBONA2;
Sabiu MUBARAK3; Adefunke Rukayat MASOPA4
1,2,3
Department of Statistics,Federal Polytechnic Ede
4
Ogun State Teaching Service Commission

Correspondence Author:- Adekunle Nurudeen MASOPA1*

Abstract:- This study focuses on consistency of stratified without any recourse to a well stated and known principle of
random sampling in repeated sampling processes within selection (Lohr,2021; Wu,2022). Kanaki & Kalogiannakis
a population with heterogeneous characteristics. The (2023) posited that process of selection under non-random
data for the study is a real life data on number of sampling usually undermine the estimation and precision
students in schools where the stratification is on the basis due to implementation procedure. On the other hand, the
of ownership (public or private). Proportional allocation sampling methods operating under the principle of known
method was used to determine the number of units and non-zero probability of inclusion have advantage over
(schools) to be chosen from each stratum for a given the non-random sampling methods particularly in the areas
sample size and four (4) independent samples of equal of procedure, estimation and precision assessment (Pawar,
sample sizes were chosen and estimates of mean Verma, Daniel, & Sayyad,2023). Karunarathna, De Alvis,
,variance as well as confidence interval obtained with Gunasena and Jayawardana (2024) posited that researchers
the estimators of stratified random sampling. The have preference for random sampling methods over non
estimates obtained for each sample sizes were subjected random sampling methods in order to explore the
to a test of significance to test the null hypothesis of no advantages of estimation and efficiency. Random sampling
significance difference between the estimates and the can be based on equal probability of selection for units in a
actual value using the t-statistic. The analysis revealed given population as it is the case in simple random sampling
that the estimates obtained for different samples differs (Rahman, Tabash, Salamzadeh, Abduli, & Rahaman, 2022),
but the test of significance revealed that there is no interval selection of units as related to systematic
significant difference in the estimates across the sampling(Iliyasu & Etikan,2021), selection of units from a
independent samples as the P-values are less the level of population where there is intra-group heterogeneity as well
significance   0.05 except for n  30 which could be as inter-group homogeneity as it is applicable in cluster
considered as an outlier. Also, there is no significant sampling(Lohr, 2021) and selection of units from
difference in the estimates of variances for the various homogeneous sub-groups obtained after stratification of a
sample sizes considered for this study with a Pvalue of 0. heterogeneous population as it is in stratified
2344. sampling(Sharma,2023; Shah, 2024).Stratified random
sampling is a method applicable where a given
Keywords:- Resampling, Stratified Sampling, Proportional heterogeneous population is partitioned into a set of non-
Allocation, Estimators, P-value. overlapping homogenous sub-groups called strata (Verma,
Verma & Abhishek, 2024). Enzo(2021) mentioned that the
I. INTRODUCTION population is usually partitioned into sub-groups on the
basis of a stratification variable capable of grouping the
Sampling is obviously the most widely used concepts units in the population in a way that the groups are mutually
in day to day human activities as it forms the basis of exclusive. Pandey (2024) opined that stratified sampling
operations in virtually all areas of endeavour. It has an specifically have enormous areas of applications. Stratified
enormous range of application in very many real life sampling is known to have a good precision when adopted
situations which makes it a concept of concern to in selection of units from heterogeneous populations in the
researchers and stakeholder in the specific areas of need presence of a characteristic which could be used for
with the aim of maximizing benefits in its usage by ensuring stratification (Ahamed, Gupt, & Phukon, 2021). The
appropriateness, efficiency and cost economy. Samples principal concern of samplers and researchers is to deploy
survey methods have wide range of applications in both an appropriate method in order to ensure accuracy of
physical, natural, health science and economy (Olayiwola, estimates obtained through the sampling process. The major
Apantaku, Bisira &Adewara,2013).Sampling could be done concern of stratified sampling is to reduce the variability at a
in a probabilistic context where every units in the population given cost or reduce the cost for a given variance(Lawal,
has a known and non-zero probability of selection(Mweshi, Salami, Obisesan,Yusuff & Owolabi,2018; Adebola &
& Sakyi, 2020; Clark & Steel, 2022,). Non – probabilistic Ajayi,2014).Some of the issues earlier raised by researchers
sampling involves selection of units from the population on stratified sampling include the determination of
stratification boundary. Horgan (2006) considered the issues

IJISRT24NOV1055 www.ijisrt.com 2635

Volume 9, Issue 11, November – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
of stratification boundaries and opined that coefficient of equal sample size were drawn from the population and
variation would be uniform if units are selected in geometric estimates of mean, variance, standard error and confidence
progression from a skewed distribution. Kareem, interval were calculated based on the estimators of stratified
Oshungade, Oyeyemi and Adejumo (2015) proposed the sampling. Let a population of N units be stratified into L
method of moving average stratification algorithm for
determination of stratification boundary. The choice of strata of sizes N h from where nh units is chosen with
method of allocation is another issue of concern in stratified respect to a method of allocation.
sampling. Equal allocation method in stratified sampling
gives no attention to the strata weights which has the Suppose X ij be a random variable observed on i
th

tendency to affect the precision (Oyoo,2021 ; Aubry,2024).

The proportional allocation strategy takes care of the effect unit selected from the j th stratum in the k th sample of size
of each stratum weight in the selection of units for 

estimation. Lawal, Salami, Obisesan, Yusuff and Owolabi

2
n ( nh  n) . Suppose  kj is the j th stratum mean
L 1
(2018) established that optimum allocation outperform other
( j ) k is the j th stratum variance. The stratification
2
methods based on the variances of the allocation procedures and
when the underlined distribution is known to be skewed. On
the contrary Fayose and Adebara (2018) established that variable is type of school (public /private) which implies
proportional allocation is the best for income data though j  1,2 and four (k=4) independent samples of equal
without any attention to the distribution exhibited by the sample sizes n were drawn from the population with the
data. More importantly, it is a general phenomenon for adoption of proportional allocation to determine the number
every concept to be prone to issues of concern to researchers of units to be selected in each of the strata. The use of
as it usually forms the basis of investigation into the proportional allocation is to ensure that the variation in the
adequacy, accuracy and appropriateness of the procedure strata population sizes is taken into consideration.
under consideration. In other words, all sampling methods Proportional allocation is a method that considers individual
when critically examined have conventional or procedural stratum weights in the determination of number of units to
issues that require attention of researchers. be chosen from each stratum for a given sample size
(Fayose & Adebara,2018).
Bolarinwa (2020) posited that a common practice in
sampling is to compute estimates on the basis of the  Estimators of mean, Variance , Standard Error and
estimators of chosen method either on a single sample or Confidence Interval
over varying sample sizes with no attention to the possibility
of obtaining different estimates from independent samples nh
taken from the same population even as the sample sizes
vary. Resampling of units from a larger sample or with

 xij
partial replacement is often adopted in some sampling
methods such as double sampling and sampling on
j n
k  i 1
j
successive occasions (Udumboso, Akanbi & Afolabi, 2019).
It is a rare instance for any estimator to produce same
estimates over independent samples since the units in  nh

x
L
different samples a bound to have unique measurements for
the characteristics of interest. The central idea of this paper X st  N  N 1
j i 1
ij

is to examine the consistency of stratified random sampling j 1 nj

over samples independently selected from a given
population for a particular size. The sample size was varied
to check whether the consistency behavior is uniform across
sample sizes or changes as sample size varies. The rest of
this paper is arranged as follow; section 2.0 deals with (1)
methodology, 3.0 results and discussions, 4.0 conclusions
 L 
V ( X st )  V (W j  )
and 5.0 recommendations k
j
j 1
II. METHODOLOGY

 Selection and Estimation procedure L

The data used for this study is the enrolment data into =(  (W j )2 (1  f j )V ( kj ))
j 1
both public and private schools in Ede North local
government, Osun state Nigeria. The population consists of
a total of 74 schools comprising 28 public schools and 46
L nh 

 (W j )2 (1  f j )  (n j  1)1 (   kj )2 )
private schools in Ede North Local Government, Osun State
Nigeria. The data consists of the number of students in the =(
j 1 i 1
various grade levels in the schools. The summary of the data
is presented in Table 1. Four independent samples of the

IJISRT24NOV1055 www.ijisrt.com 2636

Volume 9, Issue 11, November – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
L
=(  (W ) (1  f j
2
j ) j2 U k X  L
 Z1 / 2[( (W j )2 (1  f j ) j2 ]0.5
j 1 j st
j 1
(5)

(N ) n
1
=N
2
j
2
j ( N j  n j ) j2 (2) The proportional allocation method was adopted to
j 1 ensure fairness in the determination of number of units to be
chosen from each stratum for a given sample size ( n ) such
 that the stratum sample size depends on the stratum weight (
0.5
Se = [V ( X st )] Wh ).The estimators in 1, 2, 3 ,4 and 5 are used in the
computation to obtain the estimates on the k-samples shown
L

(N ) n
1 in table 2 to 5. A test of significance was performed using
=[N
2
j
2
j ( N j  n j ) j2 ]0.5 (3) the F-statistic and Pvalue to test the hypothesis in (6) to
j 1
determine whether the estimated value of the statistic(s) at a
particular sample size (n ) is significantly different.
The lower and upper boundaries of the confidence
interval respectively can be expressed as H 0 : 1  2  3   4   vs H1 : 1   2   3   4  
for at least two pairs of k.

L k X  L

st  Z1 / 2 [(  (W j ) (1  f j ) j ]
2 2 0.5 Decision rule: Reject if the P<0.05, otherwise do not
j j 1
(4) reject.

Table 1 Showing Summary of the Data

Type of school Number of schools Stratum weight
Public 28 0.3784
Private 46 0.6216
Total 74 1.0000

Table 1 shows the distribution of schools with respect III. RESULTS AND DISCUSSIONS
to ownership. It shows that there are 28 and 46 public and
private schools respectively with corresponding weights of Table 2 shows the estimates of mean, variance and
0.3784 and 0.6216 respectively. The distribution of schools confidence interval for the four independently selected
suggests that the two categories do not have the same samples of equal sizes. It was observed that the estimates of
weights as shown in column 3 where the number of private the statistic(s) are reasonably close. The estimates for
schools is almost twice the number of public schools. In various sample sizes were subjected to test of significance
order to take into consideration the variations in the stratum using the t-statistic with overall mean of 85.2466 and the
weights, the proportional allocation method was adopted Pvalue obtained was as presented in table IX. The results
such the stratum sample size is directly proportional to the show that the estimates are not significantly different from
stratum weight and expressed as nh  nWh . The various the overall mean of 85.2466 indicating that the deviation of
the estimates from the population mean is not significant
sample sizes leads to different values of nh in the two strata except for n=30 which can be considered as an outlier.
while approximation was observed in cases where the value
is a non-integer.

Table 2 Showing Estimates of mean, Variance and 95% Confidence Interval

Sample size Sample  
X V ( X st )
L k , U k
(n) number (k) st j j

25 1 82.36 187.5105 55.5207 , 109.1993

2 83.56 135.3285 60.7591 , 106.3609
3 69.28 137.0947 46.3307 , 92.2293
4 77.04 106.2560 56.8361 , 97.2439
30 1 78.63 138.3470 55.5763 , 101.6837
2 79.43 93.9289 60.4343 , 98.4257
3 71.00 132.0982 48.4729 , 93.5271
4 75.37 103.3938 55.4401 , 95.2999
35 1 73.09 93.8435 54.1029 , 92.0771
2 92.26 99.9712 72.6627 , 111.8573
3 75.11 107.0401 54.8318 , 95.3882
4 70.11 70.3275 53.6761 , 86.5469

IJISRT24NOV1055 www.ijisrt.com 2637

Volume 9, Issue 11, November – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
40 1 71.78 64.2678 56.0673 , 87.4927
2 87.38 69.9323 70.9893 , 103.7707
3 76.25 84.0253 58.2837 , 94.2163
4 84.15 77.2908 66.9187 , 101.3813
45 1 73.89 45.8211 60.6226 , 87.1574
2 86.45 49.9972 72.5911 , 100.3089
3 81.62 85.5797 63.4888 , 99.7512
4 82.18 54.1798 67.7531 , 96.6069
50 1 80.24 56.8281 65.4647 , 95.0153
2 84.18 35.9509 72.4281 , 95.9319
3 81.62 59.6646 66.4804 , 96.7596
4 92.62 39.2463 80.3412 , 104.8988
55 1 83.93 37.0770 71.9954 , 95.8646
2 87.49 30.3473 76.6927 , 98.2873
3 80.09 39.9746 67.6977 , 92.4823
4 92.62 39.2463 80.3412 , 104.8988

Table 2 Showing the Pvalue for the t-Statistics

PvalueSample size (n) 25 30 35 40 45 50 55
 0.1141 0.0119 0.2242 0.2309 0.2056 0.8474 0.7873
P( X st )

IV. CONCLUSIONS [3]. Aubry, P. (2024). On the implementation of stratified

two-stage simple random sampling without
The results of the analysis for this study as shown in replacement, with possible collapsed
table 1 revealed that different samples though yielded strata. MethodsX, 13, 102928.
different estimates of the mean under stratified random [4]. Bolarinwa, O. A. (2020). Sample size estimation for
sampling approach which are essentially not statistically health and social science researchers: The principles
significantly different from the population mean. The and considerations for different study
implication is that the estimators of stratified random designs. Nigerian Postgraduate Medical
sampling method may give different estimates on Journal, 27(2), 67-75.
resampling samples of equal size from same population but [5]. Clark, R. G., & Steel, D. G. (2022). Sample design
enjoys consistency in terms of accuracy of estimates since for analysis using high-influence probability
the divergence in the estimates are not not significantly sampling. Journal of the Royal Statistical Society
different from the overall estimates of means and variances. Series A: Statistics in Society, 185(4), 1733-1756.
[6]. Enzo, V. (2021). The use of stratification in a mixed
RECOMMENDATION abilities class to facilitate note-taking in Italian L2
during Geography lessons: A case study conducted in
Based on the result of this study, it is recommended an after-school laboratory for migrant students.
that in studies where stratified sampling is to be adopted and [7]. Fayose,T.S., &Adebara,L(2018). Efficiency of
resampling is also required, the process of resampling Proportional Allocation Procedure over other
should involve minimum number of repeated samples as the Allocation Procedures in Stratified Random
stratified sampling procedure with its estimators have Sampling. International Journal of Advanced
proven to be consistent and capable of yielding estimates Research and Publications.2 (4),pp.13-15.
that would be close to the true value. More so, it is [8]. Horgan,J.M.(2006).Stratification of skewed
necessary that samplers direct attention for further studies population. International statistical
towards examining the consistency of stratified sampling in review.74(1),pp.67-76
larger or lesser population with or without attention on the [9]. Iliyasu, R., & Etikan, I. (2021). Comparison of quota
distribution of the variable of interest and other methods of sampling and stratified random sampling. Biom.
allocation. Biostat. Int. J. Rev, 10(1), 24-27.
[10]. Kanaki, K., & Kalogiannakis, M. (2023). Sample
REFERENCES design challenges: an educational research
paradigm. International Journal of Technology
[1]. Adebola, F.B & Ajayi ,O.S (2014).Comparison of Enhanced Learning, 15(3), 266-285.
allocation procedures in a stratified random [11]. Kareem,A.O.,Oshungade, I.O.,& Oyeyemi,
sampling of skewed population under different G.M(2015). A note on the precision of Stratified
distributions and sample sizes. Sampling. Open Journal of Statistics,5(2)
[2]. Ahamed, M. I., Gupt, B. K., & Phukon, M. (2021). [12]. Karunarathna, I., De Alvis, K., Gunasena, P., &
Methods of stratification for a generalised auxiliary Jayawardana, A. (2024). Designing and conducting
variable optimum allocation. Mathematics and clinical research: Methodological approaches.
Statistics, 9(5), 617-629.

IJISRT24NOV1055 www.ijisrt.com 2638

Volume 9, Issue 11, November – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[13]. Lawal,M.,Salami,A.,Obisesan,K.O.,Yusuff,K.M.,&O
wolabi,A.A.(2018).On the comparison of some
methods of allocation in stratified random sampling
for skewed population. International Journal of
Mathematics and Statistics studies. 6(3).pp.29-42.
[14]. Lohr, S. L. (2021). Sampling: Design and analysis.
Chapman and Hall/CRC.
[15]. Mweshi, G. K., & Sakyi, K. (2020). Application of
sampling methods for the research design. Archives
of Business Review–Vol, 8(11), 180-193.
[16]. Olayiwola,O.M.,Apantaku,F.S.,Bisira,H.O.,&Adewa
ra,A.A.(2013).Efficiency of Neyman allocation
procedure over other allocation procedures in stratif
ied sampling.American Journal of Theoretical and A
pplied Statistics. 2(5).pp.122-127
[17]. Oyoo, D. O. (2021). Ratio Estimation of Finite
Population Total in Stratified Random Sampling
Under Non-response (Doctoral dissertation,
University of Nairobi).
[18]. Pandey, G. P. (2024). Advancing English Language
Teaching through Survey Research: Methodologies,
Impacts and Applications. Nepal Journal of
Multidisciplinary Research, 7(2), 127-141.
[19]. Pawar, P. B., Verma, R., Daniel, C. O., & Sayyad, L.
(2023). Foundation of Research Methodology: A
Comprehensive Guide. Red Unicorn Publication.
[20]. Rahman, M. M., Tabash, M. I., Salamzadeh, A.,
Abduli, S., & Rahaman, M. S. (2022). Sampling
techniques (probability) for quantitative social
science researchers: a conceptual guidelines with
examples. Seeu Review, 17(1), 42-51.
[21]. Shah, M. Y.(2024). Some New Aspects of Post
Stratification in Sample Surveys (Doctoral
dissertation, Sher-e-Kashmir University of
Agricultural Sciences & Technology of Jammu).
[22]. Sharma, L. R. (2023). Choosing appropriate
probability sampling designs in research. Health, 2,
4-21.
[23]. Udomboso,C.G.,Akanbi,O.B.,Afolabi,S.A.(2019).Ap
plication of regression type estimator in double
sampling skills enrolment in Oyo State. Global
Scientific Journal.7 (3).pp. 853-863.
[24]. Verma, R., Verma, S., & Abhishek, K.
(2024). Research methodology. Booksclinic
Publishing.
[25]. Wu, C. (2022). Statistical inference with non-
probability survey samples. Survey
Methodology, 48(2), 283-311.

IJISRT24NOV1055 www.ijisrt.com 2639