COMSATS University Islamabad, Lahore Campus

Calibration estimators for estimating population mean

using different distance measures
BS Thesis

By

Muhammad Bilal

CIIT/FA20-BST-007/LHR

Presented to

COMSATS University Islamabad, Lahore Campus

In partial fulfillment

Of the requirement for the degree of

SPRING, 2024

ii
IN THE NAME OF ALLAH, THE
MOST BENEFICENT AND
MERCIFUL

3
 Abstract

In order to increase the accuracy of population parameter estimations based on auxiliary data,
calibration is frequently utilized. To estimate the population variance, we suggested three
estimators: a calibration variance estimator, a calibration exponential ratio estimator, and a
calibration coefficient estimator. The suggested assessment was contrasted with those of Bhal
and Tuteja (1991), Koyuncu and Kahilar (2012), and Isaki (1981). Compared to current
estimators, the calibration variance estimators that are proposed perform better.

4
 Table of Contents
CHAPTER 1
1.1. Survey Sampling.............................................................................................................................3
1.2. Use of Auxiliary Information in Survey Sampling.........................................................................4
1.3. Calibration Estimation....................................................................................................................4
1.4. Stratified Random Sampling...........................................................................................................5
1.5. Objective of the Study....................................................................................................................6
CHAPTER 2
2.1. Literature review........................................................................................................................7
2.2. Some Existing Estimators.............................................................................................................10
2.2.1 Usual Variance Estimator:..........................................................................................................10
2.2.2 Isaki (1983) variance estimator...................................................................................................10
2.2.3. Bahl and Tuteja 1991 variance estimator..................................................................................10
2.2.4. Upadhaya and sind (1999) variance estimator..........................................................................10
2.2.5. Koyuncu and Kadilar (2012).....................................................................................................11
2.2.2. Jabeen et al (2022).....................................................................................................................11
CHAPTER 3
Proposed Estimator using calibration technique..................................................................................13
3.1. Some proposed Calibration Variance Estimator...........................................................................13
3.1.1 Proposed Calibration Variance Estimator...................................................................................13
3.1.2. Proposed Calibration Ratio Estimator.......................................................................................16
3.1.3. Proposed Calibration Exponential Estimator.............................................................................19
CHAPTER 4
Simulation Study..................................................................................................................................23
CHAPTER 5
5. Conclusion.......................................................................................................................................26
Reference.............................................................................................................................................27

5
LIST OF TABLES

Table 4.1 Parameters and distribution of Study and Auxiliary

variable……………………..23

Table 4.2 Properties of each

stratum………………………………………………………….................... 23

Table 4.3 Mean Square Error for proposed estimator from Gamma Distribution……...24

Table 4.4 Mean Square Error for proposed estimator from Normal Distribution……….24
Table 4.5 Mean Square Error for proposed estimator form Exponential Distribution…24

Table 4.6 Mean Square Error for proposed estimator from Log Normal distribution….24

Table 4.7 Comparison of Mean Square Error of

Estimator……………………………………………25

6
1.1. Survey Sampling
The act of selecting samples to study real life events is called research sampling. A sample is a
small part of the population. Randomized controlled trials and randomized controlled trials are
the two main types of experimental research. i.

Using random sampling, samples are drawn from a population based on certain probabilities.
specifically for items selected for sampling. For a person to be considered a random sample, it
must be randomly selected.

Many Researchers use non-probability sampling.in which samples are selected based on
subjective opinion of a researcher rather than by random sampling. This sampling method
depends heavily on the researcher’s expertise.

Probability sampling can take many forms, including Stratified random sampling, systematic
sampling, cluster sampling, and multi-stage sampling. Non-probability sampling includes
methods such as purposeful sampling, convenience sampling, quota sampling, and snowball
sampling.

1.2. Use of Auxiliary Information in Survey Sampling

For the first and only time, supplementary information is employed when the entirety of the data
for a research variable is unavailable. It comes from a variable that is related to the variable
being utilized, either directly or inversely. To improve estimation efficiency, the majority of
statisticians employ auxiliary data in the design and estimation phases. In order to enhance the
effectiveness of control chart operations, the idea of supplemental information is also employed
(Graunt 1662).

If auxiliary variable and study variable are directly related, the ratio estimator should be used. In
under to improve the competency of the proposed estimators, one of the approaches is to use
Calibration Estimators.

1.3. Calibration Estimation

Calibration is commonly used in survey sampling to improve the precision of estimators of
population parameters based on auxiliary data. Since the first calibration, many researchers have
studied the estimation of the calibration (Deville and Särndal 1992)were the first to use the term
"estimate" (1992). According to (Deville and Särndal 1992) calibrated weights have the smallest
distance from sampling design weights.

7
The calibration method has three main advantages in survey sampling. First of all. Calibration
gives consistent results. Seconds, it produces an important way of combining data sources
efficiently. Thirdly, calibration approaches provides computational advantages for determining
estimates. Calibration is used as an effective tool to improve the precision of estimation. This
works by minimizing the distance between original and calibrated weights. As compared to the
simple estimators this technique may provide more efficient estimators, Calibration methods can
produce more accurate results when estimating various metrics, such as population averages or
totals.

The calibration procedure measures the difference between a calibrated weight and a specified
distance. You can reduce measures. Calibration methods produce better estimates than baseline
estimates.

Several constraints have been used to propose calibration estimators by Arnab and Singh (2005),
Estevao and Sarndal 2000, Deville and Särndal 1992, and Kim and Park (2010). (Koyuncu and
Kadilar 2012), (Kim et al. 2007) provided a few estimators for calibration.
.

1.4. Stratified Random Sampling

In survey sampling sometimes the sample is divided into two groups of interests which are
homogenous in nature. These homogenous subgroups are called strata and this technique to
formulate the homogenous samples is called stratified sampling technique. The variability in the
subgroup is lower than the whole individuals so that using this technique, a statistician can get
greater precision than simple random sampling.

(Hansen and Hurwitz 1946) proposed regression and ratio estimators in survey sampling.
Furthermore, (Kim, Sungur et al. 2007) developed ratio type estimators in stratified sampling.
(Koyuncu 2012) developed calibration estimators stratified double sampling.

1.5. Objective of the Study

Following main objectives are involved in this study

8
  The objective of the study is to propose calibration variance estimators following
(Jabeen et al 2022) and (Koyuncu and Kadilar 2012).

 The study proposes three calibration variance estimators i.e. simple calibration
variance estimator, calibration variance ratio estimator and calibration variance
exponential ratio using chi square distance measure and different calibration
constraints.

 R language is used to compare the calculated estimator with existing estimators in

the study

2.1. Literature review

9
 3 The calibrating procedure's estimators were initially put forth by (Deville and Särndal
1992). Two-phase sampling was calibrated by Hidiroglou and Särndal (1998), once at the
population level and once at the level of the first-phase sample. (Hidiroglou and Särndal
1998) also addressed variance estimation and domain-specific estimation. (Skinner 1999)
examines the case in which the complete response and non-sampling error assumptions
are not met.

4 Variance can also be estimated using calibration, according to (Berg 1999). (Berg 1999)

determined the circumstances in which linear equations have no solution and estimated

the parameters of the equations using a calibration technique. The development of

software for variance and total estimation using the calibration technique is another

contribution.

(Estevao and SaÈrndal 2000) Unsaturated models were investigated, and it was
discovered among the numerous functional from a general regression estimator with
calibration estimators the lowest all of these estimators have a Taylor variance that
can be found.

In a quasi-randomization model, (Folsom and Singh 2000) determined that

calibration weighting can also be used to correct for known coverage errors and
nonresponses form units. (Folsom and Singh 2000) provide a number of important
results including a modification of the (Deville and Särndal 1992) approach to
estimate variances and random errors in their expanded framework.

The model-calibration estimator proposed by (Wu and Sitter 2001) can handle any
linear or nonlinear model and compare this pseudo empirical likelihood estimator,
proposed by (Chen and Sitter 1999)

Review the variance estimators for regression estimators (Fuller 2002), and briefly explore non-
response adjustment in regression estimation (Fuller 2002). Survey sampling frequently makes
use of population data for planning and estimating reasons. The supplemental data, often known
as supplementary information,
is usually sourced from official sources like censuses. We can use auxiliary
information in both designing a survey and developing an estimator or we can use
the information in both stages. Although the model specification may not be exact,

10
 it is desirable for the estimators to maintain good properties. As well as being a
method for estimating the variables the super population model can also be used to
link auxiliary variables to the study variable. When auxiliary variable population
quantiles are available. (Harms and Duchesne 2006) built a quantiles estimator
using the calibration technique and compared it to other Estimators of quantiles in
the literature that used auxiliary information in the estimation stage.

We construct weights for the parameter estimators based on complete univariate quantitative
auxiliary data using a nonparametric model-assisted method. In actuality, neither the weights nor
any survey parameter can be estimated using them, as they are independent of any specific study
variable. Even in cases when there are linear correlations between the study and auxiliary
variables, non-parametric models outperform linear ones because of the intricacy of the
linearized variable.
Calibration weighting was applied under suitable quasi-randomization models to rectify unit
non-response and converge error (Kott 2006). It was demonstrated that the nonlinear form of
calibration adjustment eliminated the jackknife version in variance estimation.

Calibration is commonly used to incorporate auxiliary data for more accurate population
parameter estimation (Kim, et al. 2007). Rueda et al. (2007) created a local polynomial kernel
regression model and estimated the distribution function using non-parametric regression. They
proved that the nonparametric calibration approach produces a trustworthy approximation of the
distribution function using a stimulation study.

The finite population parameter of the calibration estimator, which was proposed and is widely
used in statistics as more efficient estimators.by (Koyuncu and Kadilar 2012)

Using the auxiliary information, (Deville and Särndal 1992) developed estimation
methods to improve the study variable's estimation. Using the supplementary data,
(Deville and Särndal 1992) developed estimation methods to improve the estimation
of study variable Another technique proposed by (Mangat and Singh 1990) i.e. the
use of two randomized devices. This offered scheme is more efficient than the usual
(Warner 1965). Randomized Response Technique. After that (Mangat 1994)
modified under the conditions that are obtained under which the suggested method
is improved than those of (Warner 1965) or (Mangat and Singh 1990).

(Arnab and Mothupi 2015) , (Diana and Perri 2012) and (Tarray and Singh 2016)
provide estimators using Randomized Response technique to collect the data from
the respondent taking in view the privacy of the respondent. It has been shown that

11
 calibration, model based and model based calibration approaches provided the same
estimators under certain conditions but their variances are different.

2.2. Existing Estimators

Many helpful current estimators developed by different survey sample statisticians are included
in this section. The field of survey sampling, which was only touched upon in passing in the last
chapter, has seen a great deal of development. You will get an overview of the current formulas
for estimators in this part, along with information on how they operate.
2.2.1 Usual Variance Estimator:
^S2y =s 2y

2.2.2 Isaki (1983) variance estimator

2
2 sy
S R=S . x 2
sy

MSE=[ S2y + S 2x −2 R S x S y ]

12
2.2.3. Bahl and Tuteja 1991 variance estimator

[ ]
2 2
^S BT =s2y exp s x −s x
s2x + s 2x

[ ]
¿
λ
MSE ( S ) =θ S λ + 04 −λ¿2
2
BT
4
y
¿
40
4

2.2.4. Upadhaya and sind (1999) variance estimator

2
Sus=s y
2
[ S x + λ 04
S x + λ 04 ]
MSE ( S 2us )=θ S 4y [ λ ¿40+ g 2o λ¿04−2 g o λ ¿22 ]

Where
¿
λ 40=λ 40−1

¿
λ 04=λ 04−1

¿
λ 22= λ22−1

1 1
θ= −
n N

2.2.5. Koyuncu and Kadilar (2012)

Koyuncu and Kadilar (2012). In stratified random sampling, new calibration estimators for
estimating population mean using auxiliary information were proposed. The classic unbiased
estimator of the population mean is given by under this stratified random sampling scheme.

l
y 1=∑ W ¿h y h
h=1

¿
W h are the calibration weights that reduce the chi-square distance measure to the smallest
possible value.

Following the distance measured are

13
 l

∑ ( Ωh −W h )2
h
L1 =
Qh−W h

2.2.2. Jabeen et al (2022)

Jabeen et al (2022) proposed calibration estimators using Kim et al (2007) and Kumar et al
(2017) is given

l
y 2=∑ W h y¿hy
h

¿
There y hy is assumed to be a variable with a usula variance estimator exponential and ratio
estimators, i.e.
¿ ❑
y hy = y hy

x❑
[ ]
❑
¿ hx −X hx
❑
y = y exp ❑
hy hy ❑
xhx + X hx

❑
¿ ❑ X hx
y =y .
hy yh ❑
x hx

The calibration constraints, which define the relationship between study variables and auxiliary
variables, are listed below.

l l

∑Ω 2
s =∑ W h s 2∗¿
h hx hx ¿
h h

L L

∑ Ωh x h=∑ W h x ¿h
h h

L L

∑ Ω h= ∑ W h
h h

14
Proposed Estimator using calibration technique
We propose some new variance estimators to estimate the variation in the population using
calibration technique.

3.1. Some proposed Calibration Variance Estimator

3.1.1 Proposed Calibration Variance Estimator

We use the concept by (Jabeen et al 2022) and propose the following estimators.

l
t 1=∑ Ωh s 2∗¿
hy ¿
h

As we know that

Where

¿ 2
shy =shy

With following constraints where h=1,2,…kth stratum.

l l

∑ Ω h x h= ∑ W h X h (3.1.1.1)
h h

l l

∑ Ωh s2hx=∑ W h S 2hx (3.1.1.2)

h h

l l

∑ Ω h= ∑ W h (3.1.1.3)
h h

And write down distance measure here the following langrage's function is given according to
the proposed distance measure and calibration constraint by then

∑ ( W h−Ωh )2
( ) ( ) ( )
l l l l l l
Δ= h
Qh W h
−2 λ1 ∑ Ωh x h−∑ W h X h −2 λ2 ∑ Ωh s2hx −∑ W h S 2hx −2 λ3 ∑ Ω h− ∑ W h
h h h h h h

(3.1.1.4)

∂ Δ 2(Ωh−W h)
= −2 λ1 ( x h )−2 λ2 s 2hx −2 λ3 ( 1 ) (3.1.1.5)
∂ Ωh Qh W h

15
∂ Δ 2(Ωh−W h)−2 λ 1 ( x h Q h W h )−2 λ2 ( s hx Qh W h )−2 λ 3 (Qh W h)
2

=
∂ Ωh Qh W h

∂Δ
=0
∂ Ωh

2
(Ωh −W h )−λ 1 ( x h Q h W h )−λ 2 shx Q h W h−λ3 ( Qh W h ) =0

Ωh =W h + λ 1 x h Q h W h + λ2 ( s 2hx Qh W h ) + λ3 ( Qh W h ) (3.1.1.6)

Substituting Ωh

(∑ ) (∑ ) (∑ )
l l l l l
x h Q h W h = ∑ W h X h−∑ W h x h
2 2
λ1 x Q h W h + λ2
h x s Q h W h + λ3
h hx
h h h h h

( ) ( ) ( )
l l l l l
λ1 ∑ s 2hx x h Q h W h + λ 2 ∑ s 4hx Qh W h + λ3 ∑ s 2hx Qh W h =∑ W h S2hx −∑ W h s2hx
h h h h h

( ) ( )
l l
λ1 ∑ x h Qh W h + λ2 ∑ s 2hx Qh W h + λ3 ( Qh W h )=0
h h

[ ]
l l l

∑ x QhW h 2
∑s 2
x h Qh W h ∑ x h Qh W h

[][ ]
h hx
h h h l l
l l l λ1 ∑ W h x ¿h−∑ x h W h
∑ x h s2hx Qh W h ∑ s 4hx Qh W h ∑ s2hx Qh W h λ2 = h h 0
h h h
λ3 ¿
l l

∑ xhQ h W h ∑ s 2hx Qh W h Qh W h
h h

2
[
D=3 [ ∑ xh Q h W h .∑ s hx Q h W h . ∑ Q h W h ] −3 ∑ x h Q h W h ( ∑ s hx Q h W h ) −3 ( ∑ s hx x h Q h W h) ∑ Q h W h +6 [ ∑ shx Q h W h ∑
2 2 2
] [ 2 2
] 2

[
D= [ ∑ x h Qh W h . ∑ s 2hx Qh W h .∑ Q h W h ]− ∑ x 2h Qh W h ( ∑ s 2hx Qh W h ) − ( ∑ s2hx x h Qh W h ) ∑ Qh W h +2 [ ∑ s2hx Q h W h ∑ x Q
2
] [ 2
]

[ ][ ] [
l l l l

∑ W h x ¿h−∑ x h W h ∑ s2hx x h Qh W h ∑ x h Qh W h
l l
λ1
h h h h
l

∑ s 2hx Qh W h ¿ Qh W h ¿ λ 2 = ∑ W h x −∑ x h
¿
h
l l l h h
¿
∑W −∑ s W h ¿ ∑ s Q h W h
2∗¿ 2 4 h
s
h hx hx hx ¿ ¿ λ3
h h h

16
 [ ][ ( )( )]
l l l 2 l 2
A= ∑ W h x ¿h−∑ x h W h ∑ s 2hx Qh W h − ∑ s 2hx Qh W h −¿
h h h h

[ ][ ] [ ]
l l
l

∑ x QhW h 2
h
l

∑ W h x −∑ x h W h ∑ x h Q h W h
¿
h
l l

λ1
∑ W h X ❑h −∑ W h x h
h h
h h h h
0 ¿ Qh W h ¿ λ 2 = l l

∑ W h S 2hx −∑ s 2hx W h
l

∑ x h s2hx Qh W h ¿ ¿ λ3 h h
h
0

[ ] [( ) ( )(∑ )]
l l l l l
B=− ∑ W h X h−∑ xh W h ∑ x h s 2hx Qh W h ( Qh W h ) − ∑ s 2hx Qh W h x h Q h W h +¿
h h h h h

[ ][ ]
l l l l

∑ x QhW h 2
∑s 2
x h Qh W h ∑ W h X h−∑ x h W h l l

∑ W h X h− ∑ x h W h
h hx

[]
h h h h
l l l l λ1 h h
∑x 2
s Qh W h ∑s 4
Qh W h ∑ W h S −∑ s 2 2
Wh λ2 = l l

∑ W h S 2hx −∑ s 2hx W h
h hx hx hx hx
h h h h
l l
λ3 h h
∑ xhQ h W h ∑ s 2hx Qh W h 0 0
h h

[ ][( )( )( )( )] [
l l l l l l l l
C= ∑ W h X h−∑ x h W h ∑ x h s2hx Q h W h ∑ s 2hx Qh W h − ∑ s 4hx Qh W h ∑ x h Q h W h − ∑ W h S2hx −∑ s2hx W h
h h h h h h h h

A B C
λ 1= , λ2 = , λ 3 =
D D D

Now put the λ 1 , λ2 , λ3 values in equation

[[ ][( ) ( )] [ ][ (∑
l l l 2 l 2 l l l
Ωh =W h +Qh W h ∑ W h X −∑ x h W h ❑
h ∑ s Qh W h −
2
hx ∑ s Qh W h 2
hx − ∑ W h Shx + ∑ s hx W h
2 2 2
s hx x h Qh W
h h h h h h h

By putting int 1

l
t 1=∑ Ωh s 2∗¿
hy ¿
h

Where

17
 [[ ∑ ] [(∑ ) (∑ ) ] [∑ ] [(∑
l l l 2 l 2 l l l
W h x −∑ x h W h W h S +∑ s W h
¿ 2 2 2 2 2
Ωh =W h +Qh W h h s QhW h −
hx s Qh W h
hx − hx hx s hx x h Qh W h
h h h h h h h

3.1.2. Suggested Estimator for Calibration Ratio

Using the idea presented by Jabeen et al. (2022), we suggest the following estimators.

l
t 2=∑ W h S2∗¿
hy ¿
h

As we know that

Where

2
¿2 2 S xh
S =S .
hy yh 2
S xh

With following constraints where h=1,2,…kth stratum.

l l

∑ Ω h x h= ∑ W h X h (3.1.2.1)
h h

l l

∑ Ωh s2hx=∑ W h S 2hx (3.1.2.2)

h h

l l

∑ Ω h= ∑ W h (3.1.2.3)
h h

And write down distance measure here the following langrage's function is given according to
the proposed distance measure and calibration constraint by then

∑ ( W h−Ωh )2
( ) ( ) ( )
l l l l l l
Δ=
h
Qh W h
−2 λ1 ∑ Ωh x h−∑ W h X h −2 λ2 ∑ Ωh s2hx −∑ W h S 2hx −2 λ3 ∑ Ω h− ∑ W h
h h h h h h

(3.1.2.4)

∂ Δ 2(Ωh−W h)
= −2 λ1 ( x h )−2 λ2 s 2hx −2 λ3 (3.1.2.5)
∂ Ωh Qh W h

18
 ∂ Δ 2(Ωh−W h)−2 λ 1 ( x h Q h W h )−2 λ2 ( s hx Qh W h )−2 λ 3 (Qh W h)
2

=
∂ Ωh Qh W h

∂Δ
=0
∂ Ωh

2
(Ωh −W h )−λ 1 ( x h Q h W h )−λ 2 shx Q h W h−λ3 ( Qh W h ) =0

Ωh =W h + λ 1 x h Q h W h + λ2 ( s 2hx Qh W h ) + λ3 ( Qh W h ) (3.1.2.6)

Substituting Ωh

(∑ ) (∑ ) (∑ )
l l l l l
λ1 2
x Q h W h + λ2
h
2
x s Q h W h + λ3
h hx x h Q h W h = ∑ W h X h−∑ W h x h
h h h h h

( ) ( ) ( )
l l l l l
λ1 ∑ s 2hx x h Q h W h + λ 2 ∑ s 4hx Qh W h + λ3 ∑ s 2hx Qh W h =∑ W h S2hx −∑ W h s2hx
h h h h h

( ) ( )
l l
λ1 ∑ x h Qh W h + λ2 ∑ s 2hx Qh W h + λ3 ( Qh W h )=0
h h

[ ]
l l l

∑ x 2h Q h W h ∑ s2hx x h Qh W h ∑ x h Qh W h

[][ ]
h h h l l
l l l λ1 ∑ W h x ¿h−∑ x h W h 0
∑ x h s2hx Qh W h ∑ s 4hx Qh W h ∑ s2hx Qh W h λ2 = h h
h h h
λ3 ¿
l l

∑ xhQ h W h ∑ s 2hx Qh W h Qh W h
h h

2
[
D=3 [ ∑ xh Q h W h .∑ s hx Q h W h . ∑ Q h W h ] −3 ∑ x h Q h W h ( ∑ s hx Q h W h ) −3 ( ∑ s hx x h Q h W h) ∑ Q h W h +6 [ ∑ shx Q h W h ∑
2 2 2
] [ 2 2
] 2

[ ] [
D= [ ∑ x h Qh W h . ∑ s 2hx Qh W h .∑ Q h W h ]− ∑ x 2h Qh W h ( ∑ s 2hx Qh W h ) − ( ∑ s2hx x h Qh W h ) ∑ Qh W h +2 [ ∑ s2hx Q h W h ∑ x Q
2 2
]

[ ][ ] [
l l l l

∑ W h x ¿h−∑ x h W h ∑ s2hx x h Qh W h ∑ x h Qh W h
l l
λ1
h h h h
l

∑ s hx Qh W h ¿ Qh W h ¿ λ 2 =
2 ∑ W h x −∑ x h
¿
h
l l l h h
¿
∑W −∑ s W h ¿ ∑ s Q h W h
2∗¿ 2 4 h
s
h hx hx hx ¿ ¿ λ3
h h h

19
 [ ][ ( )( )]
l l l 2 l 2
A= ∑ W h x ¿h−∑ x h W h ∑ s 2hx Qh W h − ∑ s 2hx Qh W h −¿
h h h h

[ ][ ] [ ]
l l
l

∑ x QhW h 2
h
l

∑ W h x −∑ x h W h ∑ x h Q h W h
¿
h
l l

λ1
∑ W h X ❑h −∑ W h x h
h h
h h h h
0 ¿ Qh W h ¿ λ 2 = l l

∑ W h S 2hx −∑ s 2hx W h
l

∑ x h s2hx Qh W h ¿ ¿ λ3 h h
h
0

[ ] [( ) ( )(∑ )]
l l l l l
B=− ∑ W h X h−∑ xh W h ∑ x h s 2hx Qh W h ( Qh W h ) − ∑ s 2hx Qh W h x h Q h W h +¿
h h h h h

[ ][ ]
l l l l

∑ x QhW h 2
∑s 2
x h Qh W h ∑ W h X h−∑ x h W h l l

∑ W h X h− ∑ x h W h
h hx

[]
h h h h
l l l l λ1 h h
∑x 2
s Qh W h ∑s 4
Qh W h ∑ W h S −∑ s 2 2
Wh λ2 = l l

∑ W h S 2hx −∑ s 2hx W h
h hx hx hx hx
h h h h
l l
λ3 h h
∑ xhQ h W h ∑ s 2hx Qh W h 0 0
h h

[ ][( )( )( )( )] [
l l l l l l l l
C= ∑ W h X h−∑ x h W h ∑ x h s2hx Q h W h ∑ s 2hx Qh W h − ∑ s 4hx Qh W h ∑ x h Q h W h − ∑ W h S2hx −∑ s2hx W h
h h h h h h h h

A B C
λ 1= , λ2 = , λ 3 =
D D D

Now put the λ 1 , λ2 , λ3 values in equation

[[ ] [( ) ( )] [ ] [(∑
l l l 2 l 2 l l l
Ωh =W h +Qh W h ∑ W h x −∑ x h W h ¿
h ∑ s QhW h − 2
hx ∑ s Qh W h 2
hx − ∑ W h S hx + ∑ s hx W h
2 2 2
s hx x h Qh W h
h h h h h h h

By putting in t 2

l
t 2=∑ Ωh S 2∗¿
hy ¿
h

Where

20
 [[ ∑ ] [(∑ ) (∑ ) ] [∑ ] [(∑
l l l 2 l 2 l l l
W h x −∑ x h W h W h S +∑ s W h
¿ 2 2 2 2 2
Ωh =W h +Qh W h h s QhW h −
hx s Qh W h
hx − hx hx s hx x h Qh W h
h h h h h h h

And

2
¿2 2 S xh
Shy =S yh . 2
S xh

3.1.3. Computed Exponential Estimator for Calibration

We suggest the following estimators based on the idea presented by Jabeen et al. (2022).

l
χ =∑ W h S 2∗¿
hy ¿
h
As we know that

Where

[ ]
2 2
¿2 2 Shx −shx
S =s exp
hy hy
S2hx + s2hx

With following constraints where h=1,2,…kth stratum.

l l

∑ Ω h x h= ∑ W h X h (3.1.3.1)
h h

l l

∑Ω 2
h hxs =∑ W h S 2hx (3.1.3.2)
h h

l l

∑ Ω h= ∑ W h (3.1.3.3)
h h

And write down distance measure here the following langrage's function is given according to
the proposed distance measure and calibration constraint by then

∑ ( W h−Ωh )2
(∑ ) (∑ ) (∑ )
l l l l l l
Δ= h
−2 λ1 Ωh x h−∑ W h X h −2 λ2 Ω s −∑ W h S
2
h hx
2
hx −2 λ3 Ω h− ∑ W h
Qh W h h h h h h h

(3.1.3.4)

21
∂ Δ 2(Ωh−W h) 2
= −2 λ1 ( x h )−2 λ2 s hx −2 λ3 ( 1 ) (3.1.3.5)
∂ Ωh Qh W h

∂ Δ 2(Ωh−W h)−2 λ 1 ( x h Qh W h )−2 λ2 ( s hx Qh W h )−2 λ 3 (Qh W h)

2

=
∂ Ωh Qh W h

∂Δ
=0
∂ Ωh

2
(Ωh −W h )−λ 1 ( x h Q h W h )−λ 2 shx Q h W h−λ3 ( Qh W h ) =0

Ωh =W h + λ 1 x h Q h W h + λ2 ( s 2hx Qh W h ) + λ3 ( Qh W h ) (3.1.3.6)

Substituting Ωh

( ) ( ) ( )
l l l l l
λ1 ∑ x 2h Qh W h + λ2 ∑ x h s 2hx Qh W h + λ3 ∑ x h Q h W h = ∑ W h X h−∑ W h x h
h h h h h

( ) ( ) ( )
l l l l l
λ1 ∑ s 2hx x h Q h W h + λ 2 ∑ s 4hx Qh W h + λ3 ∑ s 2hx Qh W h =∑ W h S2hx −∑ W h s2hx
h h h h h

(∑ ) (∑ )
l l
λ1 x h Q h W h + λ2 s 2hx Qh W h + λ3 ( Qh W h )=0
h h

[ ]
l l l

∑ xhQ hW h 2
∑ shx x h Qh W h
2
∑ x h Qh W h

[][ ]
h h h l l
l l l λ1 ∑ W h x ¿h−∑ x h W h 0
∑ x h s2hx Qh W h ∑ s 4hx Qh W h ∑ s2hx Qh W h λ2 = h h
h h h
λ3 ¿
l l

∑ xhQ hW h ∑ s 2hx Qh W h Qh W h
h h

[ ] [
D=3 [ ∑ xh Qh W h .∑ s 2hx Qh W h . ∑ Qh W h ] −3 ∑ x 2h Q h W h ( ∑ s 2hx Qh W h ) −3 ( ∑ s 2hx x h Q h W h) ∑ Qh W h +6 [ ∑ s2hx Q h W h ∑
2 2
]
2
[ 2 2
] [
D= [ ∑ x h Q h W h . ∑ s hx Q h W h .∑ Q h W h ]− ∑ x h Q h W h ( ∑ s hx Q h W h ) − ( ∑ shx x h Q h W h ) ∑ Q h W h +2 [ ∑ shx Q h W h ∑ x Q
2 2 2
] 2

22
[ ][ ] [
l l l l

∑ W h x ¿h−∑ x h W h ∑ s2hx x h Qh W h ∑ x h Qh W h
l l
λ1
h h h h
l

∑ s 2hx Qh W h ¿ Qh W h ¿ λ 2 = ∑ W h x −∑ x h
¿
h
l l l h h
¿
∑W −∑ s W h ¿ ∑ s Q h W h
2∗¿ 2 4 h
s
h hx hx hx ¿ ¿ λ3
h h h

[ ][ ( )( )]
l l l 2 l 2
A= ∑ W h x ¿h−∑ x h W h ∑ s 2hx Qh W h − ∑ s 2hx Qh W h −¿
h h h h

[ ][ ] [ ]
l l
l

∑ x 2h Q h W h
l

∑ W h x ¿h−∑ x h W h ∑ x h Qh W h
l l

λ1
∑ W h X −∑ W h x h ¿
h
h h
h h h h
0 ¿ Qh W h ¿ λ 2 = l l

∑ W h S 2hx −∑ s 2hx W h
l

∑ x h s2hx Qh W h ¿ ¿ λ3 h h
h
0

[ ] [( ) ( )(∑ )]
l l l l l
B=− ∑ W h X h−∑ xh W h ∑ x h s 2hx Qh W h ( Qh W h ) − ∑ s 2hx Qh W h x h Q h W h +¿
h h h h h

[ ][ ]
l l l l

∑ xhQ h W h 2
∑ shx x h Qh W h 2
∑ W h X h− ∑ x h W h l l

∑ W h X h− ∑ x h W h
[]
h h h h
l l l l λ1 h h
∑x 2
s Qh W h ∑s 4
Qh W h ∑ W h S −∑ s 2 2
Wh λ2 = l l

∑ W h S 2hx −∑ s 2hx W h
h hx hx hx hx
h h h h
l l
λ3 h h
∑ xhQ h W h ∑ s 2hx Qh W h 0 0
h h

[∑ ][(∑ )(∑ ) (∑ )(∑ )] [ ∑

l l l l l l l l
C= W h X h− ∑ x h W h 2
x s QhW h
h hx
2
s Qh W h −
hx
4
s Qh W h
hx xh Q h W h − W h S −∑ s2hx W h
2
hx
h h h h h h h h

A B C
λ 1= , λ2 = , λ 3 =
D D D

Now put the λ 1 , λ2 , λ3 values in equation

[ ][( ) ( )] [ ][ (
l l l 2 l 2 l l l
Ωh =W h +Qh W h ∑WhX ❑
h −∑ x h W h ∑s 2
hx Qh W h − ∑s 2
hx Qh W h − ∑ W h S2hx + ∑ s 2hx W h ∑ s 2hx x h Qh W
h h h h h h h

By putting in t 3

23
 l
t 3=∑ Ωh S2∗¿
hy ¿
h

Where

[[ ∑ ] [(∑ ) (∑ ) ] [∑ ] [(∑
l l l 2 l 2 l l l
W h x −∑ x h W h W h S +∑ s W h
¿ 2 2 2 2 2
Ωh =W h +Qh W h h s QhW h −
hx s Qh W h
hx − hx hx s hx x h Qh W h
h h h h h h h

And

[ ]
2 2
¿2 2 Shx −shx
S =s exp
hy hy
S2hx + s2hx

3.1.4

L
Y st =∑ w h Y h
h=1

Nh
W h= isthe stratum weight
N

2 ( √ φh− √ W h )
L 2

L2 = ∑
h=1 φhW h

( √ φh ) + ( √W h ) −2 ( √ φh W h )
L 2 2

¿∑ 2
h=1 φh W h

L
φh +W h −2 ( √ φh W h )
¿∑ 2
h=1 φhW h

( ) ( ) ( )
L L L L
φ h+ W h
∆=2
φh W h
−2 ( √ φh W h )−2 λ 1 ∑ φ h X h ∑ W h X h −2 λ ∑ φ h s2 h x −∑ w h s 2 h x −2 λ 3
h h h h

1
σ Δ 2−2 ( φh ) /2
= −2 λ X h−2 λ 2 ( s 2 h x )−2 λ 3
σ φh φh W h

We use the concept by (Jabeen et al 2022) and propose the following estimators.

24
 l
t 1=∑ Ωh s 2∗¿
hy ¿
h

As we know that

Where

¿ 2
shy =shy

With following constraints where h=1,2,…kth stratum.

l l

∑ Ω h x h= ∑ W h X h (3.1.1.1)
h h

l l

∑Ω 2
s =∑ W h S 2hx
h hx (3.1.1.2)
h h

l l

∑ Ω h= ∑ W h (3.1.1.3)
h h

And write down distance measure here the following langrage's function is given according to
the proposed distance measure and calibration constraint by then

∑ ( W h−Ωh )2
(∑ ) (∑ ) (∑ )
l l l l l l
Δ= h
−2 λ1 Ωh x h−∑ W h X h −2 λ2 Ω s −∑ W h S
2
h hx
2
hx −2 λ3 Ω h− ∑ W h
Qh W h h h h h h h

(3.1.1.4)

∂ Δ 2(Ωh−W h)
= −2 λ1 ( x h )−2 λ2 s 2hx −2 λ3 ( 1 ) (3.1.1.5)
∂ Ωh Qh W h

∂ Δ 2(Ωh−W h)−2 λ 1 ( x h Q h W h )−2 λ2 ( s hx Qh W h )−2 λ 3 (Qh W h)

2

=
∂ Ωh Qh W h

∂Δ
=0
∂ Ωh

2
(Ωh −W h )−λ 1 ( x h Q h W h )−λ 2 shx Q h W h−λ3 ( Qh W h ) =0

Ωh =W h + λ 1 x h Q h W h + λ2 ( s 2hx Qh W h ) + λ3 ( Qh W h ) (3.1.1.6)

25
Substituting Ωh

( ) ( ) ( )
l l l l l
λ1 ∑ x 2h Qh W h + λ2 ∑ x h s 2hx Qh W h + λ3 ∑ x h Q h W h = ∑ W h X h−∑ W h x h
h h h h h

( ) ( ) ( )
l l l l l
λ1 ∑ s 2hx x h Q h W h + λ 2 ∑ s 4hx Qh W h + λ3 ∑ s 2hx Qh W h =∑ W h S2hx −∑ W h s2hx
h h h h h

(∑ ) (∑ )
l l
λ1 x h Q h W h + λ2 s 2hx Q h W h + λ3 ( Qh W h )=0
h h

[ ]
l l l

∑ x 2h Q h W h ∑ s2hx x h Qh W h ∑ x h Qh W h

[][ ]
h h h l l
l l l λ1 ∑ W h x ¿h−∑ x h W h 0
∑ x h s2hx Qh W h ∑ s 4hx Qh W h ∑ s2hx Qh W h λ2 = h h
h h h
λ3 ¿
l l

∑ xhQ h W h ∑ s 2hx Qh W h Qh W h
h h

2
[
D=3 [ ∑ xh Q h W h .∑ s hx Q h W h . ∑ Q h W h ] −3 ∑ x h Q h W h ( ∑ s hx Q h W h ) −3 ( ∑ s hx x h Q h W h) ∑ Q h W h +6 [ ∑ shx Q h W h ∑
2 2 2
] [ 2 2
] 2

[
D= [ ∑ x h Qh W h . ∑ s 2hx Qh W h .∑ Q h W h ]− ∑ x 2h Qh W h ( ∑ s 2hx Qh W h ) − ( ∑ s2hx x h Qh W h ) ∑ Qh W h +2 [ ∑ s2hx Q h W h ∑ x Q
2
] [ 2
]

[ [] ] [
l l l l

∑ W h x ¿h−∑ x h W h ∑ s2hx x h Qh W h ∑ x h Qh W h
l l
λ1
h h h h
l

∑ s hx Qh W h ¿ Qh W h ¿ λ 2 =
2 ∑ W h x −∑ x h
¿
h
l l l h h
¿
∑W −∑ s W h ¿ ∑ s Q h W h
2∗¿ 2 4 h
s
h hx hx hx ¿ ¿ λ3
h h h

[∑ ][ (∑ ) (∑ )]
l l l 2 l 2
A= W h x −∑ x h W h
¿
h
2
s Qh W h −
hx
2
s Qh W h
hx −¿
h h h h

[ ][ ] [ ]
l l
l

∑ x 2h Q h W h
l

∑ W h x ¿h−∑ x h W h ∑ x h Qh W h
l l

λ1
∑ W h X h −∑ W h x h
❑

h h
h h h h
0 ¿ Qh W h ¿ λ 2 = l l

∑ W h S 2hx −∑ s 2hx W h
l

∑x 2
h hxs Qh W h ¿ ¿ λ3 h h
h
0

26
 [ ] [( ) ( )(∑ )]
l l l l l
B=− ∑ W h X h−∑ xh W h ∑ x h s 2hx Qh W h ( Qh W h ) − ∑ s 2hx Qh W h x h Q h W h +¿
h h h h h

[ ][ ]
l l l l

∑ xhQ h W h 2
∑ shx x h Qh W h
2
∑ W h X h−∑ x h W h l l

∑ W h X h− ∑ x h W h
[]
h h h h
l l l l λ1 h h
∑x 2
s Qh W h ∑s 4
Qh W h ∑ W h S −∑ s 2 2
Wh λ2 = l l

∑ W h S 2hx −∑ s 2hx W h
h hx hx hx hx
h h h h
l l
λ3 h h
∑ xhQ h W h ∑ s 2hx Qh W h 0 0
h h

[∑ ][(∑ )(∑ ) (∑ )(∑ )] [ ∑

l l l l l l l l
C= W h X h− ∑ x h W h 2
x s QhW h
h hx
2
s Qh W h −
hx
4
s Qh W h
hx xhQ hW h − W h S −∑ s2hx W h
2
hx
h h h h h h h h

A B C
λ 1= , λ2 = , λ 3 =
D D D

Now put the λ 1 , λ2 , λ3 values in equation

[ ][( ) ( )] [ ][ (∑
l l l 2 l 2 l l l
Ωh =W h +Qh W h ∑ W h X −∑ x h W h ❑
h ∑ s Qh W h − 2
hx ∑ s Qh W h 2
hx − ∑ W h Shx + ∑ s hx W h
2 2 2
s hx x h Qh W
h h h h h h h

By putting int 1

l
t 1=∑ Ωh s 2∗¿
hy ¿
h

Where

[ ] [( ) ( )] [ ] [(
l l l 2 l 2 l l l
Ωh =W h +Qh W h ∑ W h x −∑ x h W h ∑ s
¿
h
2
hx QhW h − ∑s 2
hx Qh W h − ∑ W h S 2hx + ∑ s 2hx W h ∑ s 2hx x h Qh W h
h h h h h h h

3.1.5

( )
L 2
1 Qh
L3 = ∑ −1
h=1 Qh wh

Nh
W h= isthe stratum weight
W

27
 ( )
L 2
1 Qhwh
L3 = ∑
h=1 Qh wh

( )
2
Qh−wh
¿
whQh

( ) ( )
σΔ Qh−wh d Qh−wh
=2
σ wh whQh wh whQh

d
whQh ( Qh−wh )
2 ( Qh−wh
whQh )
wh
2
w h Qh
2
−( Qh−wh )

2 ( Qh−wh
whQh ) whQh−Qh+ wh
w h Qh
2 2

L
Y st =∑ w h Y h
h=1

Nh
W h= isthe stratum weight
N

2 ( √ φh− √ W h )
L 2

L2 = ∑
h=1 φhW h

( √ φh ) + ( √W h ) −2 ( √ φh W h )
L 2 2

¿∑ 2
h=1 φh W h

L
φh +W h −2 ( √ φh W h )
¿∑ 2
h=1 φhW h

( ) ( ) ( )
L L L L
φ h+ W h
∆=2
φh W h
−2 ( √ φh W h )−2 λ 1 ∑ φ h X h ∑ W h X h −2 λ ∑ φ h s2 h x −∑ w h s 2 h x −2 λ 3
h h h h

1
σ Δ 2−2 ( φh ) /2
= −2 λ X h−2 λ 2 ( s 2 h x )−2 λ 3
σ φh φh W h

We use the concept by (Jabeen et al 2022) and propose the following estimators.

l
t 1=∑ Ωh s 2∗¿
hy ¿
h

28
As we know that

Where

¿ 2
shy =shy

With following constraints where h=1,2,…kth stratum.

l l

∑ Ω h x h= ∑ W h X h (3.1.1.1)
h h

l l

∑Ω 2
s =∑ W h S 2hx
h hx (3.1.1.2)
h h

l l

∑ Ω h= ∑ W h (3.1.1.3)
h h

And write down distance measure here the following langrage's function is given according to
the proposed distance measure and calibration constraint by then

∑ ( W h−Ωh )2
(∑ ) (∑ ) (∑ )
l l l l l l
Δ= h
−2 λ1 Ωh x h−∑ W h X h −2 λ2 Ω s −∑ W h S
2
h hx
2
hx −2 λ3 Ω h− ∑ W h
Qh W h h h h h h h

(3.1.1.4)

∂ Δ 2(Ωh−W h)
= −2 λ1 ( x h )−2 λ2 s 2hx −2 λ3 ( 1 ) (3.1.1.5)
∂ Ωh Qh W h

∂ Δ 2(Ωh−W h)−2 λ 1 ( x h Q h W h )−2 λ2 ( s hx Qh W h )−2 λ 3 (Qh W h)

2

=
∂ Ωh Qh W h

∂Δ
=0
∂ Ωh

2
(Ωh −W h )−λ 1 ( x h Q h W h )−λ 2 shx Q h W h−λ3 ( Qh W h ) =0

Ωh =W h + λ 1 x h Q h W h + λ2 ( s 2hx Qh W h ) + λ3 ( Qh W h ) (3.1.1.6)

Substituting Ωh

29
 ( ) ( ) ( )
l l l l l
λ1 ∑ x 2h Qh W h + λ2 ∑ x h s 2hx Qh W h + λ3 ∑ x h Q h W h = ∑ W h X h−∑ W h x h
h h h h h

(∑ ) (∑ ) (∑ )
l l l l l
λ1 2
s xhQ hW h + λ2
hx
4
s Q h W h + λ3
hx s Qh W h =∑ W h S −∑ W h s2hx
2
hx
2
hx
h h h h h

( ) (∑ )
l l
λ1 ∑ x h Q h W h + λ2 2
s hx Q h W h + λ3 ( Qh W h )=0
h h

[ ]
l l l

∑ x QhW h 2
∑s 2
x h Qh W h ∑ x h Qh W h

[][ ]
h hx
h h h l l
l l l λ1 ∑ W h x ¿h−∑ x h W h
∑ x h s2hx Qh W h ∑ s 4hx Qh W h ∑ s2hx Qh W h λ2 = h h 0
h h h
λ3 ¿
l l

∑ xhQ h W h ∑ s 2hx Qh W h Qh W h
h h

2
[
D=3 [ ∑ xh Q h W h .∑ s hx Q h W h . ∑ Q h W h ] −3 ∑ x h Q h W h ( ∑ s hx Q h W h ) −3 ( ∑ s hx x h Q h W h) ∑ Q h W h +6 [ ∑ shx Q h W h ∑
2 2 2
] [ 2 2
] 2

[
D= [ ∑ x h Qh W h . ∑ s 2hx Qh W h .∑ Q h W h ]− ∑ x 2h Qh W h ( ∑ s 2hx Qh W h ) − ( ∑ s2hx x h Qh W h ) ∑ Qh W h +2 [ ∑ s2hx Q h W h ∑ x Q
2
] [ 2
]

[ ][ ] [
l l l l

∑ W h x ¿h−∑ x h W h ∑ s2hx x h Qh W h ∑ x h Qh W h
l l
λ1
h h h h
l

∑ hx h h ¿ h h ¿ λ 2 =
s 2
Q W Q W
∑ W h x ¿h −∑ x h
l l l h h
¿
∑W −∑ s W h ¿ ∑ s Q h W h
2∗¿ 2 4 h
s
h hx hx hx ¿ ¿ λ3
h h h

[ ][ ( )( )]
l l l 2 l 2
A= ∑ W h x ¿h−∑ x h W h ∑ s 2hx Qh W h − ∑ s 2hx Qh W h −¿
h h h h

[ ][ ] [ ]
l l
l

∑ x QhW h 2
h
l

∑ W h x −∑ x h W h ∑ x h Q h W h
¿
h
l l

λ1
∑ W h X ❑h −∑ W h x h
h h
h h h h
0 ¿ Qh W h ¿ λ 2 = l l

∑ W h S 2hx −∑ s 2hx W h
l

∑ x h s2hx Qh W h ¿ ¿ λ3 h h
h
0

[ ] [( ) ( )(∑ )]
l l l l l
B=− ∑ W h X h−∑ xh W h ∑ x h s 2hx Qh W h ( Qh W h ) − ∑ s 2hx Qh W h x h Q h W h +¿
h h h h h

30
[ ][ ]
l l l l

∑ x QhW h 2
∑s 2
x h Qh W h ∑ W h X h− ∑ x h W h l l

∑ W h X h− ∑ x h W h
h hx

[]
h h h h
l l l l λ1 h h
∑ x s Qh W h 2
∑ s Qh W h4
∑ W h S −∑ s W h
2 2
λ2 = l l

∑ W h S 2hx −∑ s 2hx W h
h hx hx hx hx
h h h h
l l
λ3 h h
∑ xhQ h W h ∑ s hx Qh W h
2
0 0
h h

[ ][( )( )( )( )] [
l l l l l l l l
C= ∑ W h X h− ∑ x h W h ∑ x h s2hx Q h W h ∑ s 2hx Qh W h − ∑ s 4hx Qh W h ∑ xh Q h W h − ∑ W h S2hx −∑ s2hx W h
h h h h h h h h

A B C
λ 1= ,λ = ,λ =
D 2 D 3 D

Now put the λ 1 , λ2 , λ3 values in equation

[[ ∑ ][(∑ ) (∑ ) ] [∑ ][ (∑
l l l 2 l 2 l l l
Ωh =W h +Qh W h W h X −∑ x h W h
❑
h
2
s Qh W h −
hx
2
s Qh W h
hx − W h S +∑ s W h
2
hx
2
hx s 2hx x h Qh W
h h h h h h h

By putting int 1

l
t 1=∑ Ωh s 2∗¿
hy ¿
h

Where

[[ ∑ ] [(∑ ) (∑ ) ] [∑ ] [(∑
l l l 2 l 2 l l l
W h x −∑ x h W h W h S +∑ s W h
¿ 2 2 2 2 2
Ωh =W h +Qh W h h s QhW h −
hx s Qh W h
hx − hx hx s hx x h Qh W h
h h h h h h h

31
Simulation Study
As shown in table 4.1, we generate a unique simulated population where the values of y_ji^* and
x_ji^* come from different distributions. We use a number of the changes listed in table 4.2 to
obtain varying levels of link between the inquiry and helping variable. Every population is made
up of three strata, each with five units. By selecting n_i = 2, 3, and 4 units independently from
each stratum along these lines, we arrive at (5©2) (5©3) (5©4) = 500 samples. According to
Tracy et al. (2003), the correlation coefficients between the study variable and the auxiliary
variable for each stratum are calculated as follows: 0.5, 0.7, and 0.9. Our formula for calculating
mean square error was as follows:
( Nn )
∑ ( ^y i¿− ӯ ) 2
MSE ( ^y i) = k=1 i = 1,2,3, …..

( Nn )
Table 4.1 Study parameters, the distribution of the auxiliary variable
No. Study variable parameters and The distribution and
distribution parameters of the auxiliary
variable

1 ¿ 1 ¿1.5 −1 −y
¿
¿ 1 ¿0.3 −1 −x
¿

0f( y ji) = y ji exp ji

f( x ji ) = x ji exp ji

Γ (1.5) Γ (0.3)
¿2
2 1 ¿ 0.3−1 −y
¿ − x ji
f( x ) = 1 exp
¿
f( y ji) = y ji exp ji ¿ 2
Γ (0.3) ji
√2 Π
¿
3 1
− y ji
¿ 1 ¿0.3 −1 −x
¿
¿
f( y ) = exp 2 f( x ji ) = x ji exp ji

ji
√2 Π Γ (0.3)
¿2 ¿
− x ji
4 − y ji
1
f( y ) = 1 exp
¿ ¿
ji
2
f( x ) =
ji exp 2

√2 Π √2 Π

Table 4.2 Properties of each stratum

Strata Study variable Auxiliary variable

¿
1.Stratum y 1 i = 50 + y s1 x ¿
1i
x 1 i = 15 + √(1−ρ2xy 1 ) x ¿1 i + ρ xy 1 y
s1 y 1 i
2.Stratum y 2 i = 150 + y ¿2 i s2 x ¿
x 2 i = 100 + √ (1−ρ2xy 2 ) x ¿2 i + ρ xy 2 y
s2 y 2 i

32
 3.Stratum y 3 i = 100 + y ¿3 i s3 x ¿
x 3 i = 300 + √ (1−ρ2xy 3 ) x ¿3 i + ρ xy 3 y
s3 y 3 i

Table 4.3: The proposed estimator's mean square error derived from the Gamma
distribution

Estimator MSE
t1 767990.1
t2 669908.8
t3 901215.1

Table 4.4 The suggested estimator's mean square error based on the normal distribution
Estimator MSE
t1 765402.1
t2 667618.8
t3 367148.6

Table 4.5 The proposed estimator's mean square error derived from the exponential
distribution

Estimator MSE
t1 772548.1
t2 673911.3
t3 916224.7

Table 4.6 The proposed estimator's mean square error derived from the log Normal
Distribution

33
 Estimator MSE
t1 795742.6
t2 694150.1
t3 98024114

4.5 Comparison

Table 4.7: Difference of Mean Square Error of Estimators

of t1 of t2 of t3 of usual of of Isaki of Bhal and

variance Koyuncu (1981) Tuteja(1991)
estimator and Kadilar
(2012)
767990.1 669908.8 901216 249721461 328969173 8658473 9857128
765402.1 667618.8 91622407 254676687 335131390 8558324 4587155
772548.1 673911.3 91622407 246856254 325711652 8956234 7412584
795742.6 694150.1 98024114 251629447 331650409 7852131 7412589

4.6 Conclusion

We demonstrate that the suggested estimators we have developed perform better than those of
our competitors, including the basic variance estimator (Isaki, 1981) and the Bhal and Tuteja
(1991). Moreover, the calibration variance estimator that is straightforward performs better than
the calibration exponential ratio estimator and the recommended calibration variance ratio
estimator.

34
5. Conclusion
In this research, we proposed three new estimators by using the concept by (Jabeen, et al. 2022)

to estimate the variation of the population with the aid of calibration technique. We had three

constraints when developing the proposed estimators: the first is a variance estimator, the second

is a ratio estimator, and the third is an exponential estimator with different variances. To estimate

the accuracy of a finite population from each sampling, whether it be from probability sampling

or non-probability sampling, calibration estimation is the only technique that can be used. This is

why calibration estimation is a very important tool in daily life for the sufficient accuracy that

allows us to take one sample without having to repeat the procedure.

35
Reference
1. Arnab, R. and T. Mothupi (2015). "Randomized response techniques: A case
study of the risky behaviors' of students of a certain University." Model Assisted
Statistics and Applications 10(4): 421-430.
2. Berge, A. (1999). Extension of calibration estimators in survey sampling. Journal
of the American Statistical Association, 94, 635-644
3. Chen, J. and R. Sitter (1999). "A pseudo empirical likelihood approach to the
effective use of auxiliary information in complex surveys." Statistica Sinica: 385-
406.
4. Deville, J.-C. and C.-E. Särndal (1992). "Calibration estimators in survey
sampling." Journal of the American Statistical Association 87(418): 376-382.
5. Diana, G. and P. F. Perri (2012). "A calibration-based approach to sensitive data:
a simulation study." Journal of Applied Statistics 39(1): 53-65.
6. Estevao, V. M. and C.-E. SaÈrndal (2000). "A functional form approach to
calibration." Journal of Official Statistics 16(4): 379.
7. Folsom, R. E. and A. C. Singh (2000). The generalized exponential model for
sampling weight calibration for extreme values, nonresponse, and post
stratification. Proceedings of the American Statistical Association, Survey
Research Methods Section.
8. 8. Wilson, W. A. "Regression estimation for survey samples." (2002). 28(1) Survey
Methodology: 5–24.
9. J. Grunt (1662). Sir William Petty's Economic Writings, 1899, Cambridge University
Press, Vol. 2, 314–431.
10. Duchesne, P. and T. Hammers (2006). "On calibration estimation for quantiles."
Methodology of Surveys 32(1): 37.

In 1998, Hidiroglou and Särndal published a paper. "Use of auxiliary information for two-

36
 phase sampling." Methodology of Surveys 24: 11–20.
Aslam, M., Jabeen, R., and Zaka, A. (2022). Different calibration restrictions and their
effects on calibration estimators under the randomized response technique. 92(10), 1995–
2017, Journal of Statistical Computation and Simulation.

9. Kim, J. M., Sungur, E. A., & Heo, T. Y. (2007). Calibration approach estimators in stratified
sampling. Statistics & Probability Letters, 77(1), 99–103..
10. Kott, P. S. (2006). "Using calibration weighting to adjust for nonresponse and
coverage errors." Survey Methodology 32(2): 133
11. Koyuncu, N. (2012). Application of Calibration Method to Estimators in
Sampling Theory, Hacettepe University Department of Statistics, PhD. Thesis.
12. Mangat, N. and R. Singh (1990). "An alternative randomized response procedure."
Biometrika 77(2): 439-442.

13. Mangat, N. S. (1994). "An improved randomized response strategy." Journal of

the Royal Statistical Society: Series B (Methodological) 56(1): 93-95.
14. Rueda, M., Martínez, S., Martínez, H., & Arcos, A. (2007). Estimation of the distribution

function with calibration methods. Journal of Statistical Planning and Inference, 137(2),

435–448.

15. Skinner, C. J. (1999). "Calibration weighting and non-sampling errors." Research

in Official Statistics 2: 33-4
16. Tracy, D. S., Singh, Sarjinder., & Arnab, R. A. G. H. U. N. A. T. H. (2003). Note on calibration in
stratified and double sampling. Survey Methodology, 29(1), 99-104.

17. Tarray, T. A. and H. P. Singh (2016). "New Scrambling Randomized Response Models."

18. Warner, S. L. (1965). "Randomized response: A survey technique for eliminating

evasive answer bias." Journal of the American Statistical Association 60(309):
63-69.

19. Wu, C. and R. R. Sitter (2001). "A model-calibration approach to using complete
auxiliary information from survey data." Journal of the American Statistical
Association 96(453): 185-193.

20. Kim J. K., Park, M. (2010). Calibration estimation in survey sampling

37
 International Statistical Review 78(1):21–29

21. Tarray, Ahmad, T and Singh, H.P. (2016). New scrambling randomization
response models. Jordan Journal of Mathematics and Statistics. 9(1),1-15.
22. Rueda, M., Martínez, S., Martínez, H., & Arcos, A. (2007). Estimation of the distribution
function with calibration methods. Journal of Statistical Planning and Inference, 137(2),
435–448.

38