71

Sampling Distribution and Estimation (Unit 2)

Unit 2

Sampling Distribution and Estimation

2.1. SAMPLING
2.1.1. Introduction
Sampling is process in which the fixed numbers of observations are taken randomly from a larger population. A
techgue which is fundamental for
behavioural research is known as sampling and without using it, research work is
possibie. t is impracticable and impossible to study the whole population due to practical limitations of not
cost, time
other
tactors that are indispensable and operative in studying the whole population. For the sole reason and
of making the
research
fndings economicaland accurate, the concept of sampling has been introduced. For
about to purchase or not to purchase the fruits, a fruit merchant inspects randomly only example,
for takingg decision
a few of them rather than
examining each and every fruit.
The mportant objective of sampling is obtain maximum information of the population under study using minimum of money.
to
labour ame. A definite
and plan to obtain
a sample from the
sampling frame is sample design. The method or technique which is
adopted by the researcher in selecting the units of sampling from the
population is called sampling design.
According to Cocharn, "In every branch of science we lack the resources, to study more than a fragment of the
phenomena that might advance our knowledge". In this definition a 'fragment' is the sample and 'phenomena' is the
population.
According to Davis S. Fox, "in the social science, it is not possible to collect data from every respondent relevant to our
study but only from some fractional part is called sampling".
The basis for selecting a sample survey is the framework which is called
or roadmap sample
important aspects of the survey. For obtaining some type of relevant information using survey,design
and it affects other
researchers execute it for
some population, or universe. The
sampling frame must be defined in such a way that it represents the population of
interest out of which a sample is to be drawn at random. This sampling frame should be either
identical to the population or
may be a part of it. The sampling frame can be indirectly related to the population subject to some under
the population is preschool children and the frame is a listing of coverage (e.g..
preschools). The of the design sample may be simple or
complex which provides the fundamental plan and methodology for the selection of the sample.

2.1.2. Practical Sampling Concepts

Following are the basic concepts of sampling:

Practical Sampling Concepts

Target Population
Statistical Population
Sample
Census
Sample Frame
Sample Unit

1) Target Population: The enire group of unils, Delng he Tocus of study, is represented by the
consisting of all the persons in the counry, In a partucular geographical location, or a population or universe
group based on the purpose and scope or stuuy. INon-nunan enulies ike farms, houses or
special domestic or economic
business establishments also
comprise population.
For example, if a study is to be
conductea on the marks obtained by the students of Statistics then the 'universe' wil
be all the students of that subject in that class. u ne class is

'universe.
consisting of 50 students then the 50 students is the
 MBA First Semester (Statistics for Management) ic
12
2) Statistical Fopulation: To draw out statistical inferences from the set of entities of statistical population is based.
On
population. derive generalisations about the crows, to oniy the set of crowe
Oe For example, to
s t of study. However, if one chooses a population such as 'all crows', then the limitauons will be in

well as existing in the future. There will be

existingforasstudying
additional limitations
ot geography
where
statis u r c e se are limited
t i reso
the crows. The total of quantities that are measurable or Ser Or numbers is
p u a t i o n ' . Thus, when only a single character symbolised every element of such a set then 1t 1S aerined a

DonstaPulation'. e-g., income of individuals. Depending on the finite and infinite number of elements, statistical
CrowsC a n be infinite as well as finite. A random set may not be a statistical population. For example, a set of
crows at a
point of time will not a represent statistical population.
3p e A part ofthe population which is examined to estimate the characteristics ofthe population is called a sample.

For example,
A
poruon of rice is examined to assess the quality of rice in the bag where the portion of rice examined is the
sample and the bag of rice is the population.
u)
From a large consignment, proportion of defective articles is estimated only from a portion of the consignment.
The selected portion is called
sample.
4 Census: When in a research
every item or entity of the universe is included, that approach is called
metnod, detailed information regarding every population unit is collected. "census. InTthis
It is also called complete enumeration. his
method provides more accuracy in the results in comparison to the sampling method. For example, population survey
about various issues, such as., literacy, education, income, etc., include all the items of the population.
Difference between Census and
Sample
The differences between the census and sample
are as follows:
Basis Census
Meaning Sample
When the whole population is
targeted for|When a representative sample is prepared from
collecting the relevant information about the |the target population for collecting the
relevant
members of population, it is called Census. information, it is called Sampling.
Reliability of Data Data obtained through census may be reliable
and Data obtained through sampling is not fully
complete._ |reliable, it may have some error
Time Taken |As it covers the entire population, it is time-taking |As it covers only few representatives of
process. the|
population, it is not time-taking process.
Sampling Variance There is presence of virtually zero sampling|Since the data used in sampling is taken
variance as the data used is taken from from|
entire only few members of the population, sampling
population._ variance may be present.
Drawing of Data In census, all elements of the population are| In sampling, only key representative
processed to form data. elements|
of the population processed to form data.
Applicability Census may be applied to research including Sampling may be are applied to research including
limitedarea. a large area.
Homogeneity It may be useful for population having|It may be useful for
| heterogeneous elements. population having
homogeneous elements.
5 Sample Frame: The actual set of units and a list containing every member of the
drawn at random is called sample frame. For example,
population from which a sample is
ir wants to know the factors
one
natients being admitted in the hospital after an acute attack of asthma in a
that are responsible for the
know the names of all those
specified area, then it becomes necessary to
people of that particular area being admitted into the hospital for that reason.
A good sampling frame should be:
i) Relevant: Research topic must be directly linked with relevant things.
ii) Complete: Coverage of all relevant items.
iii) Precise: It should exclude all the irrelevant items
iv) Up-to-Date: Incorporating recent changes and additions, and deleting redundant items
6) Sample Unit: Every sample is made up or SEveral
sampling units. A sampling unit may or may nomemoers
or
components. These members or components are called
Dca Sampling
campling element. But in case of sampling units Consisng or severalelement. Generally sampling unit is referred as
cluster 0f units. The individual populauon uts wn come population units, each sampling unit is
called a
Therefore in this case a samplingis an element under the cluster of units, are called
inaiviaual samoling elements.
object on which the research is
of sampling unit. done. and it fc
a vammle. in case of inspecting the oT
quality
orange is
oranges,
only few oranges are selected
to examine
that
orange is aa sampling
colony unit. But
for research. in case
These of finding
houses the literacy
are sampling level
units. of a nartical
However c a Several
the individual to e
members ofand
houses each
each
are selected
selected
house ar
in

approached for examining the literacy level, thus these individual members
of sseveral houses are
sanmpling elements.
 73
Sampling Distribution and Estimation (Unit
2)
2.1.3. Characteristics of a
The various characteristics of a good sample Sample Design
Good
1) design of
rue Kepresentative: The true representative are as
follows: and matching its
the population properties is termed as good sanpic
wICicagETEgate of certain properties is the
population and sample is the sub-total of the univci
2) choice
ree trom
and itBias:
is unbiased
A good sample does not allows prejudices, pre-conceptions, and imaginations which affects its

3) ACCurare: A sample is called good when it vields accurate estimates or statistics and free from errors.

4)COmprenensive: A sample that is true representative of the population is also comprehensive in nature wnnich
b
goody definiteofpurpose
representative the
of investigation. A sample's characteristic may be comprehensive but it may not be a

population.
) APPToachable: The subjects of good sample are easily accessible where the tools of research are easily conducted and
easy collection of data is possible.
6) Good Size: The size of good sample is such that it vields an accurate result and the error due to probability can be
estimated.
7) Feasible: A good sample creates the research work more feasible.
8 o a l orientation: Any sample which is selected by the researcher should be able to satisfy the objectives of the
Tesearch. he sample should be taken in proper number. It should be customized to fit the environment under which
tneresearch is going to be conducted. If the sample is changed as per the requirement of the survey design then it can
come out with better results and
outcomes.

9)Practical: It means that the concepts of sample selection should be applied properly while conducting the research.
The researcher should be well experienced and wel instructed. The instructions which are passed to observer should
be clear, complete and correct in all terms so as to avoid errors and biasness on their part. The sample should be
selected on basis of thee sample design. The sampling units should be representative. The sample design should be
practical and feasible in nature.
10) Economical: It refers that the research should not incur huge costs, time or efforts. One of the objectives of any
research is to complete the research with minimum effort, time, money and resources. The researcher calculates per
unit cost for each respondent. The researcher should choose that sampling design which gives minimum per
respondent cost and maximum accuracy. The funding agency is highly concerned with per unit cost and the
corresponding accuracy.

2.1.4. Sampling Distribution

Sampling distribution is a type of probability distribution (for statistics). If any statistics is given then the repeated
sampling of the population determines the sampling distribution. The sampling distribution can be determined using
another formula (other than the "repeated sampling") which depends upon the following:
1) Distribution of the population,
2) The statistics which is taken in consideration, and
3) The size of the sample.

For example, let there be a large normal population. For the given samples, a person regularly calculates the sample mean
for a sample size taken from the population which will result in different means for the different sample sizes. And the
distribution of the sample means for any give sample size is termed as the sampling distribution of the sample mean. As the
population is normal hence the distribution will be normal.
In case the statistic is the sample mean the expected value of any statistic and the mean of sampling distribution are
equivalent:

The standard error is also called the standard deviation of any quantity in case of the sampling distribution of the statistics.
In case of the statistics as the sample mean, the standard error can be calculated by:

Vn

Where G is the standard deviation of the population distribution of that quantity and n is the size (number of items) in the

sample.
 for Management) Ar
Semester (Statistics
MBA First
74
and
where, the standard deviation of the population distribution of that quantity,
o=
n= size
of the sample. (4x) to achieve
half (1/2) the
A
VeTy important quadruple the sample size
error
implication formula is that one must u
on of this formula
mustone
factor, this may
have a factor in understanding cost.
dsurement where cost is a
benefit trade-offs. designing statistical studies
S1ze.
the sample
quadruple (i.e., (4x))
n a l r (1/2) the measurement errorthen
he/she should
bution.
Otner method is the sample median from the same population having a different sampling
For example,
Sampling Distribution
Population Sample Statistic

inf
inite, X-N(4, o*) Sample mean, X
-
Finite(Size N),X - N(4) Sample mean, X
-N
Infinite, X- BinomiaKp) SamplePr oportion, P P-Binomial (p)
Infinite Sample Difference

X,-N(4,.o)x, -N\4,o3)between Means, , ,-N4, n2

-X
2.1.4.1. Properties of Sampling Distribution
1)The mean of the population and the
sampling distribution's mean are equal.
According to the normal distribution the categorisation of the population mean in terms of standard deviation is:
1) Approx. 68% of all sample means into 1 standard deviation,
ii) 95% into 1.96 standard deviations, and
ii) 99% into the 3 standard deviation.
3) The standard error of the mean is the standard deviation of the sampling distribution.

2.1.4.2. Sampling Distribution of Mean Xx

The sampling distribution of the mean X, is the probability distribution for every possible value of sample mean X.

Expected Value of X
Hence, any statistic's expected value and the mean of sampling distribution are same. When the statistics is the sample mean,

Standard Deviation of X
Standard error of any quantity is the standard deviation of the sampling distribution of that
statistics.
The standard error in case of the statistics as the sample mean will be,

where, G = the standard deviation of the population distribution of that quantity, andd
n size of the sample.

2.1.4.3. Sampling Distribution of Sample Proportion p

r e a l life cases like success or failure, accept or
uhole nopulation be divided into reject, coin's head or tail, etc., there comes a
whole popula may
the binomial
wo mutuay Cusve Catcgories 1.e., either or situation, thesesituation when the
pracical exampies experiments. If the
taken from random sample (from the sampling
situations are the
binomial procedure
is conducted
of elements
n a
population) have the appropriately and x elements
ou
sample proportion specified characteristics, then n
sample proportion p is assumed to best statistics. This
De une

inferences for the population proportion parameter p. p can be used for the statistical

The formula for the sample proportion can be given as:

Elements ofsamplehavingchracteristic (x)
p Samplesize(n)
 15
Sampling Distribution and Estimation (Unit 2)
2.1.5. Sample Size for
The sample size for a research refers to the total number of elements of the population to be included l
for
con search study.
the research study.
Both qualitative and quantitative points are involved in specifying the size Or
u
The accuracy ot research depends on the sizc of the sample. It has been observed that a larger sanple gives O curad
The
cy
andesnate levels. Apart from it, the availability of money., resources and efforts also specify the size of a sampie.
sample size of any research study is
represented by "n
If the
populaion is highly heterogeneous, then the sample will be divided into various groups with an indication hat tne
Sampie S would be larger. If the population is homogenous then the sample will be small. Thus, at pre-assigned level
there is a nxation of the
sampling eror and the standard error of the statistic, thus having sample siZe as n is chosen.
2.1.5.1. Factors Influencing Sample Size
The following points should be taken care of while
deciding the sample size:
Factors Infuencing Sample Size

Size of the Universe

Availability of Resources
Level of Accuracy Required
Homogeneity or Heterogeneity of
the Universe
Nature of Study

Selection of Sampling Technique

Attitude of Respondents

Degree of Variability

1) Size of universe: It has been observed and statistically proven that the sample size should be large so that it can
represent the whole target population. If the universe size is large and heterogeneous then the researcher should take
large sample size and vice versa.
2) Avaiiability of Resources: The researcher needs a lot of resources to complete the task of research. These resources
bind the researcher. It means that the researcher has to complete the research within those allouted resources
Resources can be money, time, experts or any other variable. If the resources are easily available than the sample size
can be large otherwise smaller sample would be appropriate.
3) Level of Accuracy Required: For any research, the level of the accuracy affects sample size as bigger sampie
represents the whole population properly than a smaller sample. However, it is not always mandatory that a large
sample will always give more accurate results. The selection of the sampling technique and research tools also impact
the accuracy level of the research.
4) Homogeneity or Heterogeneity of the Universe: Homogenous universe is that universe in which all the elements are
similar, on the other hand heterogeneous universe is that universe where the elements are different from each other. A
homogeneous universe requires a smaller sample size, while researcher needs to tak bigger sample size in case of
heterogeneous universe because a small sample may not represent the whole population.
5) Nature of Study: The researcher may choose a small sample, if the research is intensive and fundamental
(continuous) in nature. But in those cases, where the research is extensive and applied (researches which are not going
to be repeated soon), then the researcher is bound to select a larger sample size. So, it is quite clear that the nature of

study also affects the sample sizze.

6) Selection of Sampling Technique: The sample size is also impacted by the kind of sampling technique selected by
the researcher for conducting the esearch. For example, when a researcher adopts the simple random sampling
technique then he must choose large sample size, but on the other hand in stratified sampling technique the researcher
needs a comparatively smaller sample size for accurate and effective results
7) Attitude of Respondents: The researcheris an experienced individual, so he will always chose a larger sample size in
which he thinks that the non-response rate will be higher. In simple words, if there are chances that the respondents
will not reply and return the questionnaire, then large sample size is chosen.

8) Degree of the Variability: Degree of varnabiltyspecifiesthe exlent to which the scores in any distribution are spread
in the population. Larger numbers always inCalcs grealer variabiliuty of scores in a population. Sometimes dispersion
is substituted for variability, and this letn 1s wideiy useu n statustics. If the population is heterogeneous, then the
researcher uses larger sample size so as t p c i e d preciSIon level. On the other hand, if the population is
more homogeneous then the smaller sample
size is selected by the researcher
 (Statistics for Managem
MBA First Semester agement) AU
76
Member
2.1.5.2. Determination and Selection of Samplediscussed in two estimation problems:
of optimal sample size has been
The method of deternnation
1) Determination sample size n
ot population mean.
when estimating the
2) Detemination of sample size n when estimating the population proportio

Determination of Sample Size (m) when Estimating the Population Mean

t s ciear that, for large samples, unbiased estimator of population mean (u) is sample mcan (A). The standard errorof

size:
1s vn The following pre-assigned value is required for estimating the sample
1) The desired oonfidence level.
2 The permissible sampling emor (E).
The population standand deviation (O)
So. E=-u and standard emor
of X .n
=

For e samples. Z= Or (X-u)=Z

o/Vn n
Suppose the desired confidence level is 95%. Then Z values defining 95% confidence level are t1.96. thus
1.96 x o
-4)= 1.96x or E=t1.96x- n=t
n Vn E

n96X orn

E
Let's consider an example to explain this procedure. If a researcher wants to caliculate the average income (in lac) of the
given population having an accuracy of 05 of in income. This means that the researcher can tolerate a loss of 50% in
income on either side of the true average income at 95% confidence level. Thus, it means that the researcher wants to be
95% confident about his decision.

The formula for confidence limits is: (X -u) =Z

Where.
= Population mean
=
Average income calculated from the sample
Z Value of z at 95% confidence level

= Standard eror of x

= Standard deviation
= Sample size

Ifthe researcher has decided to tolerate an error of 1/2 lac that is (X-u) =0.5
So 05-7 value of Z at 95% contidence level are t1.96

0.5=t1.96 1.96 x o
So Vn
n 0.5

The value of standard devlation (o) can be calculated by either:

)Assuming or guessing.
2) Consulting by the expert,
3) By experimenting to get the value, and
4) Obtaining o from other comparable studies.

Let 3 lac
 Sampling Distribution and Estimation 79
(Unit 2)
Now assum if the
population proportion is not more than 0.3 then p -
r 03 and q = 0.7. So.

Sample Size (n) = -96 0.31-03)0 sO67 6

896 4 896
(o.03 00009

Bxampie&
than 2
more
ABC hotel
management is interested in determining the percentage of the guests of
days. the hotel whe stay tor

ne reseauon manager wants to be 95% confident that the

valu what is the has been estimated
1.96 for the
most conservative sample size needed for thispercentage to be within t * ot the true
(Z =
given confidence level of 95)
problem
Solution: We have been
given the
Population is infinite following
E 0.03 (since the estimate
=
should be within 3% of the true value)
L96 (as per table of area under normal curve for the
given confidence level of 956).
As we want the most
conservative
sample size we shall take the value of p 0.5 and q
can determine the sample size for the = = 0.5. Using all this information. we
given problem as under
9 196).0.5)1-0.5) 0.9604
1067.11 1067
O.03 0.0009

Thus, the most conservative

sample size needed for the problem is =
1067.

2.1.6. Central Limit Theorem

With the increase in the
sample size, the mean's sampling distribution moves towards the normal distr1bution. It
influenced by the fonm of the population distribution. is not

According to the central limit theorem the sample means may also be calculated with the
help of Z scores.
Now Z

Where is the finite mean and a is the standard distribution.

The Z score equation for the sample means for the normal distribution can be given by.
Z OX
Or ZX-u

One can assume that the sampling distnbution is normally distnbuted for
any populations or processes in case of
size being n 2 30. If the population situatuons are adverse then a sampie
sample size of atleast 30 will be
required whle in case of
the non-pormal populations the smaller sample size will be
sufficient
As the sample means of sample size n230 are normally disubuted hence the norumal distibutioa for
from any population (either non-normaliy distributed or of unknow n sample data taken
shapes) can be calculated. This is the
of central limit
theorem
main advantage
In figure 2.1 column four different Populalion distibuuoAs are illustraled,
every column shows the shape of the
distribution for the sample size of theu FeSpecuvE sallple cans. With the increase in sanmple size a, the
distribution approaches the normal curve sample means

For any randomly selected sample (having

i observ alloms InUn the
approaches the normal distribution if the bainple n 1850, it ndicales population,
the sampling distnbution ot the mean
that with the increased sample size the chances of
getting normal distribution increases. IThe population niean and the sanmple mean are same
the population is greater than the standard enor of the mean while the standard deviation of
(fo the above discussed case).
From population, randomly select any sample hav ing n observations. If the size of the
a
the mean sampling distribution will apprOac th towards the nomal distribution. The larger sample is larger i.e., n2 30 then
more will be chances of sampling distnbutuon the size of the sample (n) ne
approa hine towards the normal disturibution.
 or
(Statistics for Management) AUc
Management
80 MBA First Semester

n 30
Population n= 5

Distribution I n 2

Exponential

Unifom

U-Shaped
M 1
Normal
A A
Figure 2.1: Population Distributions
Application of Central Limit Theorem
1) Vanous assumptions for the relationship between the sample statistics and the
using the Central Limit Theorem. population parameters can be made
The probability for
getting different sample means can be calculated by this theorem.
3) As the sample means of
sample size n230 are normally distributed hence the normal distribution for
non-normally distributed or of unknown shapes) can be calculated. This sample
from any population (either data taken
central imit theorem. is the main use of

Example 9: In a shipment of manufactured tools there were 2% defective tools. Calculate the
Lools in shipment of 400 tools, when: probabilities of defective
1) 3% or more tools are defective and
2) 2% or less tools are defeective.

Solution:
=p =0.02 (0.02X0.98) 0.14 = 0.007
H, and a, = 400 20
Using the correction for discrete variables, 1/2N = 1/800 =
0.00125, one has
(0.03 0.00125) in standard units 0.03-0.00125 -0.02
1.25
0.007
Required Probability (area under normal curve to right of z 1.25)
=
= = 0. 1056
In case of not using the corrections, the answer will be 0.0764.

Example 10: Consider facory having 000

a

nof22000 standad deviation ofworking

and he employeces, their monthly savings are normally distributed
s0.
dctribtion with replacement and without replacenent, llCalculate the mean and the standard deviation of having
2 sanples àre oblained (each the sampling
1)Withreplacement, sample having 4 emplovees).
mploye:
2) Without replacement.
Solutíon: Given N = 1000, = 2000, o = 50, n = 4
1) Sampling with replacement

H 2000
2) Sampling without replacement:
The mean for the sampling distribution mean is,
=2000
 (Statistics for Manageme
tistics for Management) AU
Semester

MBA First

4
satisfy (1) is
given by
that will
value of n
=
n
We know that P{lzl s 196) = 095, Therefore, the least

= 1.96

i.c., n= (: a s)
1.., n = 6 x40

1.., n= 245 86

Therefore. least size of the sample is 246

2.1.7. Sampling Process material, in what

manner, and by whom,

will be taken what times, on which 1s a

npon o which measurements at

incorporate that
data to he evaluated whicn
of the sampling plans should Following
a s samplingofplan". The designing stated in the goals.
as
the of interest
parameters and be answerable to all the questions
peentauve sample
are the stcps involved in developing a
Define the Universe: The universe
sampling plan
be restricted to
can gcographical limits, a particular type of of elements or
some
product or sOme

The collection
the the initial problem in procedures of sampling.
other constraints. Defining termed as the
universe is
inferences are to be made is
information desired by the researcher regarding what definition will result
ojects possessing must as ambiguous
target population is
target population or the universe. Precise definition of the thhe
in ineffective research and deceptive in meaning. Proper defining of
the target population requires translating
problem definition precisely as what should and should not be included in the sample.
be
constructed for some study to
2 Sample Frame: The frame is enlisted by somepopulation that already exists or it is
the frame of the populatiop 1s
conducted by the researcher. After the specification of the population to be studied.
which is built by the researcher for the
developed. Sampling frame consists of all thesampling units of the population
subjecti of his study or it consists of already existing list of the population.

Specifyingg the Sampling Units: A sampling the basis of the sampling frame which is the primary
unit is decided on

unit containing the sampling elements of the population like, city blocks,
households, a business organisation, etc. The

design of the project is the partial basis of the selection of the sampling unit. The units that are served as initial
sampling are called as units.
primary sampling The objective of the inquiry is the basis of the composition of one or
more units of the population.
Selection of Sample Design: The procedure of selecting units in the sample involves two primary methods of
sampiing namely. probability and non-probability methods that can be further sub-divided into specific methods of
seiection. The probability sample being one and the procedure involved in selecting the units has some specific chance
of its inclusion in the sample. An arbitrary method 1s adopted in selecting the units which is not depending upon the
chance for the non-probability sample. The purpose ot inquiry and the attitude or convenience of the investigators is
the basis of this method.
5Determination of Sample Size: The desired degree of accuracy in investigation is directly proportional to the size of
the sample which also depends upon the nature of the population and the method of selection. The ideal sample size in
marketing research investigationsdepends upon the type of series and the population size. There will be more degree
of heterogeneity and larger units will be drawn in the sampie due to larger size of the population. Hence the size of the
sample should be larger for being fully representative
Sedect the Sample: Execution of the actual
6 sampling Pprocess is
called as selection of the sample and the real selecüon
of the elcments of sample requinng
consideiabie anount or otice and [ield work especially when the personai
iterviews are involved. Detailed specific alion salnpng
or ne
design 1s the requirement tor the execution of the
n o orocess with respect w the implementatlon
samp or he s1ze ot
sampling technque, and sample size population, sämpling frame, sampling uni.

2.1.8. Advantages of Sampling

Following are the advantages of sampling
1) Saves Time, Money
and Effort: The subjects involved in
sainpling
late tabulate, present, analyse, and interprel and hence he researchersnall in number which gives him less
are

of the sample saves time, time

2 More Efective: Due to
the
sie being small than th of money and eftort.
collecting the information and investigalor works more effectivel the
population, the tiredness is
d in

Faster and Cheaper: Due tw small sample size the

data
reducc
3) interpretation takes less ume ahd n i v c 8 O m i n a l expenses collection, tabulation,
presentation, analysi and
 (85
Sampling Distribution and Estimation (Unit 2)
tabulation. presentation,
data involved in collection,
4) More ACcurate: Small are in sampling because small
are

analysis and interpretatio"

tion
5) GIves More Comprehensive Information: Theoroueh investigafion of the study is the resuit ot smai
chance to be included in h e
the membere of the population have equal
piete information as all
l
sample.

2.1.9. Disadvantages of Sampling

Following are the disadvantages of sampling worker
the research
be included in sampling by
) Biased Selection: Biased selection of the respondents may
large number ot tactors
because a
It
very dificult to select truly representative sample
2Dhcuy in Seection: is a

obstructing the method of selecting good samples

where the investigator
required in the sampling method
3)peialised Knowledge Needed: Specialised knowledge is

may commit serious mistakes due to its absence

with the researcher
uncooperative
of Cooperation: Due to scattered sample the subjects
Problem are
4)
is not suitabie
5) Less Accuracy: When the higher standard of accuracy is expected then the sampling
where census
dissimilar universe it impossible to derive a representative sample
6) Lmited Nature: Duc to small or is

study is the best possible substitute.

2.2. SAMPLING DESIGNS/SAMPLING TECHIQUES

2.2.1. Introduction the potential sampling
units from the
The plan, method. or which a researcher identifies and selects
technique through
the study is termed as "sample design'". Sample
a relevant sample for
sampling frame or the target population, to form of sample it is not possible to start
a

the principle of research process. Without proper designing

designing Is working any
the population. The designing
of sample may
the responses of relevant members of
survey. as it is the base which provides
used for it.
be simple or complex depending on the method
of
basis are responsible for different types
selection technique and the representation
The two factors namely. the element (random sampling) or non-
based on representation, may be either probability sampling
sample technique. The sample, On the basis of element selection the sample can be
restrictive or
probability sampling (non-random sampling). from the large population then it is termed as
unrestrictive sampling.When each sample element is selected individually
all the other types of sampling are termed as 'restrictive sampling',
unrestrictive sample' and
of two types, viz., probability sampling and non-probability sampling.
Thus, sample designs are basically
Types of Sampling Design

Probability Sanpling Non-Probability Samphing

2.2.2. Probability Sampling selected in the sample, known as

the universe are given equal chance of being
is

The method in which all units of that obuained through

is an assurance ot the results in terms of probability
are
Probability Sampling. There of the results lies in measuring the errors of estimation obtained
The Significance
probability or random sampling. intentional sample
which brings predominance of the random sampling designs over the
from a random sample
various probability sampling techniques:
design.Following are the
Types of Probability Sampling

Simple Random Sanpling

Systematic Sampling
Stratified Random Sanpling Cluster Sampling

Multi-Stage Sampling Arca Saunpiing

 (Statistics for Managemen:
stics for Management) AUc
Semester
MBA First
86

2.2.2.1. is equally
probable of pe:
etting
Simple Random Sampling where each
unit of the
population
n units are to be selected
This is the most famous and simple method
emethod
of samplingsampling w
population is
from which
lation
'N
'N'
1ro random sampling says tha
included in the sample. Let us consider of the being selected equally. Simple
that the size
and the choices are
adom tor a sample such that NC sample has the
nSample has the
probability
probability of
of beng included in
included
in the sample
sampic
) be
population to
nere is an eaual element of the
Cn

each other
nttosample
E aFac
2) c h possible combination has an equal chance of being chosen.
by the method of simple random
units is eliminated
pOSSibility of bias or personal prejudices in the sclection of the
sampling.
Methods of Simple Random Sampling
Some of the common methods of drawing simple random samples are: sample or SiZe n is
This method includes the following steps. Let N be the
size of population and a

enoa:
to be drawn:
i) The units of the population are numbered from 1 to N.
11) N identical pieces of papers, cards, capsules or balls are taken and are numbered from I to N
or
items are selected either by replacement
1) These are mixed carefully in a bag, bowl or some other container and n
without replacement in one stroke.
(ii) constitutes the desired
the items drawn in step
he number of units in the population bearing the numbers on

random sample.
increases to infinity this method is
ne applicaton of this method is simple and easy but as the size of the population
unfeasible.
2) By Using Random Numbers: The random number table that helps in selecting a sample has been constructed by
Some experts. The most popular of all the existing tables, Tippett's Tables is very popular and in use. Out of the gaven

population,random selection of thesamples is done

through these numbers. This table ranges from 0 to 9 having an
equal chance of appearing in any position of the table. The actual utilisation of the random numbers is for large
population where the random number table consist of several columns and rows out of which one isselected at random
and then the samples are continuously selected in sequence for the desired size of the samples. Without any bias it
provides a set of random numbers.

Advantages of Simple Random Sampling

Salient advantages of simple random sampling are as follows:
1) Freedom from Bias: The biggest advantage of simple random sampling is that it offers absence of human bias and
error of classification as every member of the population 1s provided with fair chance of selection.

2) Representativeness: Without any erors in the execution of simple random sampling, the results are fully typical to
the population of interest. Thus, theoretically a bad luck can only compromise the researcher's access to all the
important data about a given population not representing the sample.
3) Ease of Sampling and Analysis: Comparatively, Simple random sampling only requires a complete list of the
elements in the population whereas, the requirement or ouner methods of sampling is
Lnowledee of the population before the selection of subjects. The highly representativedeep research and advance
through simple random sampling simplifies the analysis of the results and data interpretation.sample of the population
Disadvantages of Simple Random Sampling
Disadvantages of simple random sampling are as follows:
1 Cost: Simple random sampling usage is limitea Dy
nOssible item in the universe to be
cost ractor. According to this method there is a same chance for
www.nhical chosen,
ne actual
items. For conducting personal nterviews by ne sample
chosen often consist
of widely
interviewers involves dispersed
enhances the cost of field operations. extensive travelling which
2) Availability of a Current Listing of Universe Elements: The
practical use of
accurate list of universe eler:ent w hich has serious limitations, simple random sampling requir
3) Statistical Eficiency: There is a possibility of a standard
in
getting
compared to the other sample design which is statisticallymore efficient
error
for thea same size sample. This is he when
associated with this sampling.
statistically
efficient sample d
difficulty
4) Administrative Dif+culties: There are aber of administrative difficulties
this method the selection of sample 15 conceptualy simple yet
problematic associated with simple
is the problem related to
maintaining supervisory control There is another random sampue
at times. g
in simple random sampling
adnministrative
during home personal
interview
 87
Sampling Distribution and Estimation (Unit 2)

2.2.2,2. Systematic Sampling at specified interva

the universe the other units are selected systematically
a
of one unit at random from
After tnehSeection the basis of any system the units or tne
of time. is method is applicable when the size of the population is finite and on

universe are arranged such as alphabetic arangement, numerical arrangement, or geographical arrangemenis

Advantages of Systematic Sampling

Systematic sampling has certain advantages, which are as follow:
it is simple and convenient
1) Pe and Conveníent: The results obtained by this system are generally satisfactory as

method.
2) Gives Similar Results: The results obtained by proportionate stratified random sampling are similar to the resu

given by systematie sampling when the population size is large.

on the
The sample obtained is representative sample as this method of selection is not dependent
3) naependent: a true
property of universe under study.
4) Litte Chance of Biasness: The sample is considered to be free from bias or least affected.
candidate contesting in a tie
5) Helps in Random Selection: The popularity of this method is in the selection of the
position and in random drawing of the prizes.

Disadvantages of Systematic Sampling

Following are the disadvantages of systematic sampling:
1) High Sampling Error: There are several clusters in the universe but in this method only one cluster of the population

isselected. Due to this reason, the sampling error is very high and it cannot be evaluated properly
but in
of Selecting Impracticable Units: Sometimes unrealistic units of the population are selected,
a
2) Possibility
practical situation it is very easy to determine the presence of periodicity or determination of its significance.
3) Biased: The drawer's bias affects the selection if all the chits are not identically folded and are not same in size, shappe
and colour.
because it is very difficult to
4) Not Suitable for Large Population: This method is not suitable for a large universe
Write down the names or particulars of a large number of units on the chits of paper.

2.2.2.3. Stratified Random Sampling

strata or parts of a universe instead
In the Stratified Random Sampling, the sample is selected from different homogeneous
of heterogeneous universe as a whole. The summary of this sampling procedure is as follows:
are mutually exclusive and include all items in the universe.
1) The sampled universe is divided (or stratified) into groups that
from each group or stratum.
2) A simple random sample is then chosen independently
from that of simple random sampling. In simple random
There is a difference in the process of stratified random sampling
random from the whole universe whereas in stratified sampling the design of thhe
sampling the sample items are chosen at
random sample from each stratum. The distribution of the sample among strata is
sample is the selection of the separate
based on chance in simple random sampling.
(i.e, strata)
Formally, divide the population into non-overlapping groups
N, N2
Such that
N+ N2. +.. . + N , =N

in each stratum, where f is the sampling fraction. For example, suppose that
Then randomly choosing a sample of f =n/N
in a universe of 1,10,O00 grocery stores is to be studied by the researcher. He will
the retail sale of a product such as sugar
based on store size, as stated below:
first sub-divide this universe into three strata,
Table 2.1
Store SizeStratum | Number of Stores Percentage of Stores
Large stores 22,000 20
Medium stores 33,000 30
Smallstores 55,000 50
Total 1,10,000 100

the three strata, the researcher guarantees for the desired sample
Then, by independent random sampling Within cach of their
allocation of stores within each size group, rather leaving representation to chance. For example, according to the
stratification scheme it was required to have toal samples of 120 stores, the researcher applies, simple random sampling
from each strata, that is expected to yiela about 24 lan a s (20%% of 120); about 36 medium stores (30% of 120); a
 Semester
(Statistics lor Manageme
ment).AUC
MBA First
88
of simple
in the strata.
n case
rand
about 60 small stores of the characteri
(50% of 1 20). Thus the sstrata
h u s all the
would get
representation

pOssible
that o n c stic of
sampling. the equal chances of occumences are given to
are
all sample points,
it may

small strata
may not at all get any C e s
representation
Advantages of Stratified Random Sampling
Advantages of stratified random sampling are as follows stratified in homogeneous groups
th
When it is
)More Representatives: This the is case of heterogencous
universe.
and any particularity regardine
heterogencous
different strata being is called proportionate
is reduced due to within the strata then it
heterogencity fractions are s a m e (f) then it is calle
t n e sampling
strata
(f) within the
stratified random sampling Wnen sampling fractions are different

disproportionate stratitied random sampling

ertainty: The uncertainty of random sanmpling against bias of intcentional sclection
is balanced by
this yPe
o saulipng
stratified random Sampling
Gater Precision: In comparison to simple random samoling the estimates provided by
are more procise duc to reduction in the variability in cach stratum.
result in administative
Convenience: Division of the universe into different strata of sub-groups
srauve
convenience.
Disadvantages of Stratified Random Sampling
Disadvantages of stratified random sampling are as follows:
1) Needs More Attention: Stratificd sampling results in proportionate or disproportionate sampling aesg
proportionate sampling. the sample size is proportional to the stratum size where, there is a higher precision
nomogeneous population. Disproportionate stratification provides changing sample size for eacn
stratum. Criteria used for allocating the strata points determine whether the precision of the design is excellent or
pitiful. It is best suited for strata with varying characteristics because the accuracy of one study is optimised only that

cannot be transferred to consecutive surveys.

2) Time Consuming: Obtaining the sample by this method involves several steps alongwith the record of the population
being studied. Sometimes due to non-availability of the list and to develop it, makes the work tougher as the strata
should be mutually exclusive resulting in increased sample size and increases the expenditure. The time of study is
also extended.
3) Complicated: Decisions on stratification are completed before the start of the study. The information collected
becomes invalid in choosing drawing conclusions if the choices are wrong. Data analysis is also complex due to the
consideration of number and size of strata population, size of total population and sample population.
4) Expensive: Use of the design requires large sample size which enhances the cost particularly in those cases where
there is a classification of the needed list to be purchased. There are other instances where the list is easily accessible
but the people are scattered geographically and made necessary arrangements to reaching them, which adds expenses

2.2.2.4. Cluster Sampling

According to this method there is further notüceable sub-division of the universe into clusters. Simple random sampling is
performed and clusters are drawn accordingly constituting a sample of all he units belonging to the selected clusters. For
example, if we have to conduct a survey inthe city of Mumbai, then the city may be divided into, say. 40 blocks and out of
these 40 blocks, 5 blocks can be picked up by random sampling and the people in these five blocks are interviewed to gie
their opinion on a particular issue. The clusters Chosen snould de simal in size, 1.e., more or less the same number ot
samnle units should be there in each cluster. This method is used in the collection of data about some common traits of the
nonnlation. Comparing it with random sampling cluster, sampling is certainly less precise and there is no doubt about it in
the reduction of the cost by intensifying surveys in selected clusters. Comparatively, within a cluster there is not much
information in 'n' observations that appears to De n " Tanouy uran observations. Due to its economie
cluster sampling is used. Cluster sample estimales are most reliable and effective per unit cost advantages
Difference between Stratified and Cluster Sampling
Table 2.2 shows the difference between stratified and cluster sampling:
Table 2.2: Difference between Stratified and Cluster Sampling
Stratified Sampling
One divide the population into a few sub-proups One divide the
Cluster Sampling
There are many elenents in each sub-group There
population into many sub-gvups:
i) are tew elenments in each
i) Selection of the sub groups depending upon
the criterion
ii) Accordine to some eriterion sub-group.
related to the variablesunder study. selected or by av.ailability in of ease Sub-groups
Homogeneity is secured within sub-proups. Heterogeneity is secured withun dat:a collection.
Securing heterogeneity between
sub-proups
within each sub-group
Secunng sub-groups.
homogeneiy between sub-groups.
Randomly choosing
the elements
Riandomly choosing the several
Lypicalstudy sub-groups to
 89
Sampling Distribution and Estimation (Unit 2)

Aadvantages of Cluster Sampling

Advantages of cluster sampling are as follows:
1) neaP; Quck and Easy: The technique of this sampling is cheap, quick and casy. The researcher alocates his limited
nc
samplescluster rather than sampilng
Oy
o
he randomly selected few clusters or areas while using
whole country by using simple random
sampling. Due to accessibility, the
Size: The size of the sample can te increased by the researcher by this technique.
2) LATEer sample
Suoycets O esearch can be increased, considering that the researcher has to only collect the sample from a given populanon.

sampling reduces because the sCope of

3) Convenient to Obtain: Obtaining of the custers is convenient and the cost of
study is limited only to custers instead of whole population.
4)Cost Eitective: The scope of custer sampling method is widely used in marketing research due to its complete cost
effectiveness and feasibility of implementation, espectally in the situations of area sampling

Disadvantages of Cluster Sampling

Given below are the disadvantages of cluster sampling of
) Least Representative: This techniquc of sampling is lcast representative of the population.
There is
a tendency
with cluster sample the results of the study are
smlar characteristics of the individuals within the cluster and
a
having
skewod due to the chance of over-representation or under-representation of the cluster
This is due to the
2) High Sampling Error: The percentage of error is high as it is a probability sampling technique. of the
cluster being lhmited in numbers which are included in the sampling and it leaves behind significant proportion
un-sampled population.
if the elements of a
3 L e s s EAicient: Cluster sampling may be statistically less efficient than simple random sampling
cluster are similar.
sampling single unit from the ciuster in
Sometimes not Appropriate: The sampling from the cluster is better than
a
4
the situation when the elements of a cluster are same.

2.2.2.5. Multi-Stage Sampling

is constituted by all the units
sampling is multi-stage sampling where in cluster sampling, sample
a
Modification of cluster
units is in two, three, four stages. Firstly. the
selected in a cluster but in multi-stage sampling the selection of the sample
into second units out of which another
universe 1s divided into first-stage sample units and then further
sub-divided stage
in the For example, in an urban
sample is selected and similarly third and fourth stage sampling performed
is same way.
then for each selected towns households will be taken as sub
of selection of and
survey first stage sampling will be
towns
will be selected of
samples third-stage sampling.
sample and depending upon the need, individuals
as

Advantages of Multi-Stage Sampling

Advantages of multi-stage sampling are as follows
1) Flexible: It is very flexible method of sampling.
as subsequent stages of samples are required only for units which
2) Saves Time: It saves a lot of time, energy and cost
and limited in number.
are selected in the prior stages

is easier than single stage designs due to the reason that

B) Administrative Efficiency: Comparatively, administering
units under multi-stage sampling.
frame of sampling is developed in partial
Areas: For the surveys of unde veloped areas it is totally useful as there is no
4) Helps in Survey of Undeveloped
frame for sub-dividing the materials into rational small sample units.
requirement of updated and accurale
number of units can be sampled tor a given cost due to
5) Sampling of Large Units: Under mulu-stage sampling large
consecutive clustering, which is not po0ssible in ost of thc simple designs

Disadvantages of Multi-Stage Sampling

Multi-stage sampling has the following disadvantages eluster
1) Large Number of Errors: Iis subject to à laiger ainiount of etOIS due to the process of d1viding the strata or

in different stages into divisons and sub-divisions

2) Greater Variability: There 1s a grealer amount of varlability in the estimates than any other method of sampling.
3) Less Emicient: Generally, it is less efli ient than an appropriate single stage random sampling

2.2.2.6. Area Sampling

The area sampling is a form of muli-stage samplng in which maps instead of lists or registers is used as 'sampling frame'.
Itis commonily used in those countries which do not have proper sampling frame like a population list. For geographic sub-
 Managemens
Semester
(Statistics for nt) AU¢
MBA First
90 geographical area i.
of units based on
and negative fee
The cluster The positive
sSampling' is the other name of 'area sampling'. sampling. into several sma
primary sampling units known as 'cluster designs' which are
famous as
area

for sampling
is
sub-divided
naller
of cluster
sampling are also applicable The total area
which 18 usea
Ior
area sampling.
to area sampling.
Dcable to sampg consider a city map, unique number for
area

For example, frame has a

18 sclected. and each s t r a t i f i c a t i o n is emplova
pe 'frame' in case of area sampling block
this samnlin
Ownas the geographical
considerations)
segments (more or less
subdivided into
Sample. F o r Dased on each block is
is taken by idei
identifying the blocks. At the end,
cqual
qual and
size) and these
u a l size) these 'segment samples' may be selected in the sampie.

dvantages of Area Sampling

Advantages of area sampling are as follows: because the scope of study is
of the clusters is convenient and the cost of sampling reduces
ining
limited only to clusters instead of whole population. or reduced
the cost per element is reduced in area sampling than in
stratified sampling because
omparatively,
listing of the elements.
sampling frames of the
n e
individual
area
sampling proves feasible in many situations due to unavailability of the
elenments of the population and hence nootherrandom sampling technique is possible to appiy.

Disadvantages of Area Sampling

Disadvantages of area sampling are as follows:
1)
Similar Elements: Statistically, sampling may be less efficient in comparison to simple random elements
area sampling due to
of the
similanity in
the elements of the area. Only in the extreme case when there is similarity in the a
clusters. a sampling of a single unit from that area is much better than sampling the more units from that area.

2Costly: In comparison with simple random sampling, area sampling has more costs and problems of statistical analysis.

2.2.2.7. Advantages of Probability Sampling

Following are the advantages of probability sampling:
) Unbiased Estimates: The sampling method which basically provides unbiased estimates are called random sampling,
having significant precision. If this level of objectivity is required by the investigator, then some alternative probability
sampling is important.
2 Relative Eficiency: Evaluation in quantitative tens of the relative efficiency of other sampling techniques under a given
sinuation by the researcher is permnitted by random sampling which is
3) Less Universe
normally impossible in non-probability sampling.
Knowledge Required: Comparatively, little knowledge of universe is required. Basically, two things
are required to be known only:
) A unique way of identifying each universe element, and
i) The total number of universe elements.
4) Fair: Every item in the population has an equal chance of being selected & analysed.
5) Easy: Easy data analysis and error calculation is allowed by this method of sampling

2.2.2.8. Disadvantages of Probability Sampling

The disadvantages of probability sampling are as follows:
1) Less Emcient: Other sampling methods are more statistically efficient in
comparison to this method,
2) Non-Uilisation of Additional Knowledge: No additional
knowledge is used in this method about
is organised. how the population
aComnlev and Time Consuming: In many cases, this method of
oriularly for marketing research in which the buaget constraints selection
and
is time
consuming and complicated.
probability methods of sampling. imitations of time
provide preference to non
AHich Level Skills: A very high level
of skill and experience 1s
required for using
More Time Requlred: For planning and executon of a sampling. probability
probability sample a lot of
ioh Costs: In
comparison to non-probabilnlySamping. probability time is required
sampling involves higher cost.
2.2.3. Non-Probability Sampling
Non-probability sampling is that type of sampling procedure which does not
that whether or not each item in the population has been een
included in thehave any ground for
probability sampling such as deliberate sampling. purpososive sampling and judgement 'There are estimating
nple.
different
the
names
proba
of no
s n. this type ofsampling
 )

the researcner deliberately selects items for the sample and the choice of researcher regarding the ite P
weightage. In other words, under non-probabilítý samoling the organiser of the inquiry purposively cnooses d a
O Consutute a sample on the basis that the small portion selected by him, out of a huge one 1S typical

representative of the whole universe. The various non- probability sampling designs are:

Types of Non-Probabillty Sampling

Convenience Sampling
Purposive Sampling
Panel Sampling Snowball Sampling

2.2.3.1. Convenience Sampling

On the basis of convenience and approachability. the choicc of the sampling units by the researcher, 15known s
convenience sampling'. Samples that are selected accidentally are called 'accidental samples'. Because of the seleeton
procedure (units are selected from theiractual place) it is called as 'sample of the man in the street' Due to their
tne
accessibility, samples units selected. For exampie, by adding the
are product in the nearby suitable shop8,
new
potential of the product is tested. This is accomplished by observing the purchasing and selling report of the prouct.

Advantages of Convenience Sampling

Advantages of convenience sampling are as following:
1) Economical: It is less costly and less time-consuming.
of the universe is ensured when the knowledge of the composition of
2) Proper Kepresentation: Proper representâtion
the universe is adequate for investigation and it is free from bias.
3) Avoid Irrelevant Items: It restricts the entry of unnecessary and irelevant items into the sample per chance.

4) Intensive Study: Intensive study of the selected items is ensured.

5) Accurate Results: Better results are obtained if the investigator has the capacity of keen observation and sound
judgement as well as un-biasedness.

Disadvantages of Convenience Sampling

Disadvantages of convenience sampling are as follows:
1) Personal Bias: The investigation plays and affects the selection due to enough scope of bias and prejudices.
2) No Equal Chance: Al the items of the universe to be included in the sample stand unequal chance.
3) No Degree of Accuracy: The degree of acuracy achieved is unknown in the investigation conducted by this method.
4) No Possibility of Sample Error: Possibility of calculating sample eror is negligible as it is based on mathematical
to non-random methods of sampling.
concept and it is not applicable
5) Unsuitable for Large Samples: This method is not appropriate for the large samples where the size of the universe
and the sample is significantly large.

2.2.3.2. Purposive Sampling

A non-probability sample which follows certain norms is called purposive sampling. Purposive sampling is basically of
two types:
1) Judgment Sampling: The study which is based on the parameters of population, where the units are selected by a
researcher or some other expert on his her judgement, is called 'judgement sampling'. This technique of sampling is
appropriate in the situation where the study of the population is difficult to locate or there are members who are
comparatively better than others for an interview in terms of knowledge or interest. For example, a sample of grocery
stores,considered to be "representative", is selected by a group of sales manager. Experimentally this was found to be
a failure because there is no unbiased way of evaluating the precision of sample results. In spite of these limitations,
when the size of the sample is extremely small this method is useful.

Judgment sampling has the following advantages:

1) Suitable for Small Sampling Units: When the universe includes small number of sampling units then various
important elements could not be selected by simple random selection, whereas judgement selection definitely include
these elements in the sample.
2) Studying Unknown Traits of Population: For the study of some unknown characteristics of a population ard for some
population whoso'taits are known, the population is stratified on the basis of these known properties. Sampling units are
selected on the basissof the judgment from each stratum. This method helps in obtaining a more representative sample.
 (Statistics for Managema
MBA First
Semester ment) Au
92
of business,
making and in of public polie
day-to-day problems sample designs. Th
olving executivsness
Problems: To solve
short of time
to wait
for
fail
probability
in
eisions, officials who are
blic
SOving their urgent problems judgement is the only prac
sampling
Disadvantages of judgement sampling are as follows: units to be sampled
are affec.
ected
the population
Non-Scientific: Due toersOnal
personal bias and prejudice
prejudice oi the investigator
of
Inerisk
invewhich the
and it is concluda is involved in
the Juage a
investi cd that the method is unscientific. The in the sample
which adjust to h
conchusions by incuding those items
euestined
preconcecived idea.
2) N to Calculate Sampling Error: To cvaluate thec reliability of sample results there does not exist any
If the individual nas Knowledge
method. This method can onlv be successful duc to quality of judgement.
then it leads to a representat ve sample,
n e population in the decisions making and has good judgement, taken into account that despPite
incorrect. This should be
th c aceision is based on the sample then it may be determine the size and probability of
sanmple being a good represcntative, no nethod is impartial to
c
sampling emor.

Quota Sampling
Quota sampling is the most commonly used non-probability sample designs, which is most comprehensively used in
consumer surveys. Principle of stratification is also used by this sampling method. In stratified random sampling the
TesearcheT begins by building strata. The common bases for stratification in consumer surveys are demographic, e-g., age,
gender. income and so on. Compound stratification is generally used, e.g., gender-wise age groups. For each stratum
Sampie sizes are fixed. Regarding stratified random sampling, the sampling within strata may be either proportional or
csproportional. Interviews are conducted with the designated quotas by the field workers, with the identification of
individual respondents being left to the field-workers.

Advantages of quota sampling are as follows:

) Economical: Travelling costs can be reduced by an interviewer because he does not require travelling all over a town
tor tracking pre-selected respondents. By applying numerous controls in a quota sample, it will be more expensive
despite having less bias in selection.
2)
Administratively Convenient: Being administratively convenient, with the of quota sampling. labour of selecting
a random sample can be avoided. The problem of call-backs and non-contacts is abolished altogether.
use

3) Minimum Memory Errors: For minimising memory errors in the situation when field work is be done
to quickly.
most appropriate and feasible technique is quota sampling.
4) Independent: It is not related with the existence of sampling frames. Due to unavailability of suitable sampling
frames, the only choice available is quota sampling.

Disadvantages of quota sampling are as follows:

1) Difficulty in Calculating Standard Errors: Random selection is not the criterion for
impossible to calculate estimates of standard errors for the results of he sample. quota sampling and hence it s

2 Diticulty in Obtaining Representative Sample: Depending on the mood and

selection of the convenience of the interviewers the
'representative sample is
impossible to obtain within the
quota.
3) Hamnpers Quality of Work: The quality oI work sulfers due to too much of
interviewers. independence of the incompler
4 Diicult to Supervise and Control: Under quota siamplng it is
totally difficult to
supervise and control field investigation
2.2.3.3. Panel Sampling
In nanel sampling a group of
encatedpartucipants are selccu a y dy random sampling method and
number of times during that the same group is askedto
edly for iterative studies. In this samplng,period of time. This
there is a sanmple is
semi-permanent where
in getting high response rate, even by mail facility of
selecting and
memb
contacting samples ha
Advantages of Panel Sampling
Advantages of panel sampling are as follows
1) Saves Cost and Time: In the collection of
inlormation it takes
lesser time and
Due cost.
21 Helps in Measuring Changes: to fixed sannple units, the
changes in
repeated reporting
3) Helps in Tracing Shift in Behaviour: Additional shifts of behaviour can be traced can be measurea.
 Sampling Distribution and Estimation (Unit 2) 93
Disadvantages of Panel Sampling
Disadvantages of panel sampling are as follows:
1) Not Representative: Under panel
sampling the sample is not fully representative.
2) Members become conditioned: The members of the trained for some speci ns
pancl may be
Sa
3) Diticult to Preserve Representative Character of P'ane!: Due to voluntary drop out of the professional members, it
1 u r e a to replace of those members who are difficult to retain thecir representative character for a long time.

2.2.3.4. Snowball Sampling

when the characteristic of the desired sample is limited then the special non-probability method is applicabie. In this
method. t 18 ditficult to locate the respondents because it will be very costly. Depending on the referrals of the initial
subyects snowball sampling generates additional subjects. Though this technique is biased and unable to represent a good
cross-section from the population but dramatically it reduces the scarch cost.

Advantages of Snowball Sampling

Following are the advantages of snowball sampling:
1) Identifving and Selecting Prospective Respondents: The reasonable method of identifying and selecting awaited
respondents from smal., hard-to-reach, uniqucly defined target populations is known as 'snowball sampling'
2) Useful in Qualitative Research: For qualitative research practices such as focus group interviews, this non-
probability sampling method is very useful.
3) Needs Little Planning: In comparison to other sampling techniques, this sampling techniques requires litle planning
and fewer workforce.
4) Less Costly: The primary advantage of this sampling method is the reduction of sample sizes and costs.

Disadvantages of Snowball Sampling

Snowball sampling has some disadvantages which are:
1) Biased: The overail research study involves biasness in snowball sampling. There are certain problems in this technique
because of some important differences between the people who are either known or unknown to social circles.
2) Limited Data Structure: Data structures are limited and unable to generalise the results to members of the target
population being larger, defined like all the other non-probability sampling techniques.
3) Limited Control: The subject obtained by the researcher is based on the previously observed subjects due to
researcher's little control over sampling methods.
4) Researcher has no Idea of Distribution: There is no guarantee for the representativeness of the sample. The
researcher is unaware of the true distribution of the population and sample.

2.2.3.5. Advantages of Non-Probability Sampling

The advantages of non-probability sampling are as follows:
1) True Universe Picture: Relevant sections of the universe may be selected in the proportions they appear in the
universe.

2) Economical: Geographical concentration can be achieved thus reducing costs.

3) Quick: Useful and quick method in certain circumstances.
4) Specific Cases Types: This can be the only technique available for some special cases such as illegal drug users.
5) Specific Members of Population: Helpful, if researchers are truly interested in specific members of a population, not
the entire population.
6) Pilot Study: Acts as an exploratory research like a pilx study, attempting to detenmine whether a problem exists or not.

2.2.3.6. Disadvantages of Non-Probability Sampling

The disadvantages of non-probability sampling are as follows:
inlormation ot the universe is needed
)Details Needed: Detailed initial
Errors: Errors in sample selection can easily occur
3) Subjective Nature: The subjectivity oI non- probability sampling prevents making inferences to the entire population.

4) Selection Bias: Validity and credibility are questionable due to selection bias.

5) Reliability: The reliability of the resulung estumates cannot be evaluated which results in the user not knowing how
much confidence can be placed in any interpretalions of the survey finding.
 94 Semester
(Statistics Management) AU
for Managemer
stics for
MBA First

Sampling
2.2.4. DifferenceDeTween Probability and
Non-Probability
Non-Probability Sampling_

| 1) Control
Basis between l Probability Sampling Sampling
error
cannot be
controlled in non-

in probability probability sampling.

Sampling error can Can be controlled

of higher
level selection

Chances of sampling.
existence
2) may be
not influenced by the| There
Selection dhcess 1s depends on biasnes
expertise of because it
Cperise of the researcher and quicker
Bias the specific technique. involvement of cost is very
low

3) Economy The involvement of time and costs may be high. The

alternative.
results 1s not very high
4) Reliability through reliability of The are not
testing or hypotheses the tests of significance
fo Y in obtaining Ibecause parametric
tests of significance
"rOus
more reliable results. applicable.
population it may De
more
5) Suitablity is heterogeneous then it is more For homogeneous
POpulation
reliable and representative. useful. be estimated are
to
at
oroad,
6) Usefulness
Complex, detailed estimates is required then it If parameters as market
shares or total
such
may be preferable. aggregated levels,
reasonably
sales, it may beused be scattered.
7) Degree of
pOpulation is high then the accuracy may be Accuracy in such situations may
Accuracy poor. of an
8) Sampling in the absence
inprobability sampling formal sampling frames are It is effective even

Frame required. elaborate sampling frame.

v E e n c e It the cheaper geographical spread of the population | More 'convenient, less time-consuming, non
1s high and likely to have lower then it may be very| sampling erors.
inconvenient.

2.3. ESTIMATION THEORY

Estimation Theory is the technique or method which is used to estimate the population parameters from sample studies. t
is very essential to estimate the parameters whenever a sample study has been organised. For example, a manutacturer 1s
interested to know the estimates of the future demand of his product. A businessman wants to estimate about the future
sales and profits, likewise an engineer is interested to be informed about the percentage of defective articles which are to be
produced in future by his machine and a tyre and bulb manufacturer will equally be interested to predict about their lengths
of life of their respective tyres and bulbs.

2.3.1. Estimator and Estimate

A sample statistic is used when one makes an estimate of a population parameter. This sample statistic is the 'estimator.

For example, the sample means, X = i s -

x
N
X is a point estimator of the population
mean ju. For
estimating the same parameter different statistics can be used. The
function of the N observed values, X1, X2,, An, Which 1s sampled from a random variable
estimator.
X, is called 'statistical

Criteria of Good Estimator

The following properties are possessed by a good estimator:
1D Unbiasedness:
If the real value of the parameter
COinciaeso wIth the value of its estimate then it is known as unbiased
estimator, Let us consider that is the mean
parameter the population. If the expected value of
enal to the population parameter then the estimator or a the estimator is
populaton parameter is said to be
unbiased if E()=0. unbiased, i.e., 0 is
For example, the sample mean (X) is an unbiased estimator. In case the random
value of X is u, the same value is to be estimated. sample is given where the expec
2) Consistency: When the size of any estimator is increased and the value .

parameter, then it is called consistent estimator. (statistic) of estimator comes

very close to the
3) For example, the estimator is called consistent ifa sample mean (X) comes
closer to the
parameter val alue of the mean
Campling Distribution and Estimation
(Unit 2) 95
4) Eiicency: There are several cases where unbiased and constant
one.For example, mean estimator of the parameter value can be more than
n Wold d a n are unbiased and consistent estimators of the parameter mean in a.norma
distribution. Mean would be
distribution of mean would beconsidered
less than
to be more efficient estimate than median because variance of the sampling
the variance of the sampling distributton or
5)SucEnyouncient estimator of the parameter is the statistic which contains all ne
the information of the paranel u
the sampe. 1he estimator, who utilises all the
Suncient eSalmato. This sufficient estimatorinfomation the about
for the parameter parameter,
would also provided oy
be the moStuu emelc aud
consistent estimator. It is not an unbiased
estimator.
2.3.2. Types of Estimation
Prof. R.A. Fisher developed "The Theory of
1) Point Estimation, and Estimation' in 1930, which is categorised into two classes
2) Interval Estimation.

2.3.3. Point Estimation

The specificvalue of the
sample statistic used to estimate population parameter is known as point estimate
a . ro
estimation manages the task of selecting a particular sample value which is an estimate for a populaion paranie
Pont estumaton of some population parameter is shown in figure 2.3. The population parameter of interest might be the
mean, varance, standard deviation, proportion or any other characteristic of the population. Collection of random sample
to estumare tne value of an unknown population parameter typically comprises of 'n' observations of the variable of
interest. The estimator of the population parameter is a function of these sample observations.

Population Typical sample

Size n
Parameter 0 Sampling Point estimate_
Estimation
(unknown) Values 0=f(1, X2.a)

Point estimate of 0
Flgure 2.3: Point Estimation of a Population Parameter

The point estimators for the population parameters u, t and o are given in table 2.3.
Table 2.3: Point Estimators of , xand o
Population Parameters Point Estimator Formula for Point Estimate

i=l

Median Middle value in sample (50%-ile)

25% trimmed mean Mean of middle 50% of values in sample
10% trimmed mean_ Mean of middle 80% of values in sample
P x/n where x number of
=
successes in n trial

s=-/n-1
i=l

2.3.3.1. Properties of Point Estimation

The information about the best estimator for a statistical point estimate is provided by the sampling distribution of the
estimator. Diferent sample statistics are used as point estimators of different population parameters. The general notations
used are:
of interest being estimated
6 population parameter (such as , G, p)
estimator ofe
6 sample statistic (such as ,s, P) orpoint
and 0 is read as 'theta hat".
Here, 0 (theta) is the Greek letter

norms for selecting an

estimator are:
The
1) Unbiasedness: Due to the error ofsampling.tne vaue ot a data determined from a given sample is expected to be
either above or below the actual value of popuiianon parameter of activity. If the mean of the sampling distribution of
sample means taken from a population 15 equal o e population mean, then the sample statistic is called an 'unbiased
 atistics Tor
(Statistics
TOt a s C t ) AUc
M a H g e n i

Semester

First
MBA

an
ased estimator
unbiased estimator of the oe

to be
6 is said
statistic
sample
estimator of the population parameter.
*eter.
Hence, u Hence, the

population parameter, provided E( 0) = ¬.

r, therefore, boh
Where, E( 8) = Expected value or mean of the sample statsur proportion
mean
and sample
We have E( X) distribution of sample
=
p and E p) =p., in a sampling parameters ju and P.
population
A and p are unbiased estimators of the coresponding 6) becomes ven
est1mator (
a point
(statistic) of
2) Consistency: When the increased and the value the
standard
error of
size of the sample is estimator'. For
example,
'consistent
COse to the population parameter, then it is called as Thus, the sample
increases.
size 'n'
tends to become smaller
as sample
Sampling distribution ofthe mean, ( 0 z ) = , is a consistent
Vn
the sample
proportion p
C a n A ) 1s a consistent estimator of the population mean (). Similarly,

estimator of the population proportion p because (o,)=

n it with anolici *
Cency: Efficiency is a relative term. Efficiency of an estimator means comparing the variance of
d;1.e
is less than
variance of ,
w o unbiased estimators 6, and 6,. If the
6.
Var(6,) < Var(®,) then the estimator 6, is called an efficient estimator of
which contains all the
information of the parameter in
4) Sutficient estimator of the parameter is the statistic the sample, iS termed
Suiciency: which is provided by
The estimator utilising all the information about the parameter, and the most
ne sample. would also be the most
efficient
as estimator. This sufficient estimator for the parameter
sulticient
consistent estimator. It is not an unbiased estimator.

2.3.3.2. Point Estimation of Population Mean (X)

because 1t combines
most common estimator of the population mean (0). It is very popular
The sample mean (X) is the
high efficiency with no biasness.
estimators.
show that if X is normal. the sample mean is more efficient than all other unbiased
In fact, one can

Point Estimation of Population Proportion (T)

2.3.3.3.
is the sample statistic P = X/n. where X is the number of
The recommended estimator of the population proportion
t is one of the two parameters in the binomial
"successes" found in the sample of 'n' observations. The parameter
in the large population. To estimate t, calculate the observed
distribution corresponding to the proportion of "successes
number of successes (x) in a sample on 'n tnals and estimate t as x/n. For example, to estúmate the proportion of
residents in a metropolitan area where public transport is chosen for their journey to work, a sample of 100 residents state
26/100 0.26.
transport. The best estimate of t is p
= =

that 26 took public

2.3.3.4. Point Estimation of Population Variance (G)

For the population variance o', the implied estimator is S. Given a sample, Xi, X2 Xn, the sample value of this random

variable is,

,- i=l
S
n-l

square deviation from the population

mean is o
The mean p. It therefore proves that =

estimate ECX -

o is obtainene
by averaging the squared deviations from the sample mean. f one takes many samples of sizen from aofpopulation,
ed for each sample would
average to O.
On tne other hand, if is to divide by
one
would be biased low and tend to underestimate o. In
estimating o, the unknown n. the estima0
one iIs computing squared deviations around the population mean u is replaced wit
sample mean X. That is, sample mean, not the
population mean.
smaller deviations aroundu
than the
The deviations around x are
it stimates
underestimates o. In order to co
o". Clearly, the mean deviation
if
than
the mean deviation aroundu, compensate for this biasness, arour
round x is smaller
one inflates the
sum or square
s
Sampling Distribution and Estimation (Unit 2)
97

by dividing bynumber smaller than n,

a most efficient
namcly, n 1. In addition to being unbiased, $° is also the
(minimam MsE) unbiased estimnator of a', providing X is normally distributcd. Though used most commonly, the
estitdtroro, the population standard deviation, is the square root of S? or S. It is a biased estimator of a. However, the
aeeiec O D1as 1s very small for reasonable sample sizes and S is much more convenient than altermative unbiasea
estimators.

EXdie 18: BeloW you are given the values btained from a random samplc of 4 observations taken from an infinite
population:
32 34 35 39
) Find a point estimate for
u. Is this an unbiased estimate of u? Explain.
ii) Find a point estimate for o. Is this an unbiased estimate
iii) Find a point estimate for o.
of d. Explain.
iv) What can be said about the sampling distribution of X ? Be surc to discuss the expected value, the standard deviation
and the shape of the sampling distribution of x ?

Solution:
i) Point Estimate of u = -32+34+35+39 =
35
4

It is an unbiased estimate ofu.

ii) Point Estimate of o =4 A

n-1

_(32-35) + (34-35)* +(35-35) +(39 +35) 9+l+0+16 _ 8.6667

4-1 3
It is an unbiased estimate of o

i) Point Estimate of o = vo* = V8.6667 = 2.9439

z-
Vn
iv) The sampling distribution has a mean equal to the population mean. The sampling distribution has a standard deviation
equal to the population standard deviation divided by the square root of the sample size. The shape of the sampling
distribution of x will be a normal curve.

2.3.4. Interval Estimation

The fixed interval of scores where the population's mean or some other parameter is expected to fall when that parameter
data is known as 'interval estimation'.
is to be estimated from the given sample
It is unexpected to exactly coincide the point estumate with the population parameter. Let us consider that durnng a survey

one finds that the average

income of a household is 3,00,000 per annum. This does not mean that every household has
Some households may have more than R3, 00,000 and sone may have less than this
income of 73,00,000 per annum.
will not coincide with the population parameter. The true value of the population parameter is
amount. Thus, point estimate
estimation which 1s an established interval consisting of a lower limit and an upper limit.
expected to fall into the interval
in the parlance of inferential statisties.
This interval is called "Confidence Interval

of Population Parameters
Difference between Point Estimate and Interval Estimate Interval Estimate of Population Parameters
Point Estimate of Population P'arameters An interval estimate is detined by two numbers. between
population paraneter is ä single
point estimate of
a
A which a population parameteris said to lie.
value of a stalistuc. For example a < X < b is an interval estimate of the
sample mean (X) is à point estimate
For example, the
Similarly, the sample population mean a. lt indicates that the populativn mean is
of population
the mean H. greater than a but less than b.
pont
estimate of the
proportion P is
LPopulation proportion P.
Confidence Interval
2.3.4.1. hat the tnue value of the population parameter will fall in an interval
and the prODaDIlity
If one selects the repeated samples
then this ntervial 1s caled "confidence interval'. Generally, 95% contidence level and,
is used in a certain percentage,
confidence level is used.
Sometimes 99%
 Semester
for Management) AUc
(Statistics for
Managemen
98 MBA First

samples,
the ilit
probaoy
that the t e
Suppose chooses 99% as the
one taking repeated the population
parameter wi
confidence level. one keeps on
level. If one
value will fall in this al COnfidence keep
interval is 99%. In other words, one isi 99% sure that the true value of
d i s t r i b u t i o n principle. Th
fall in this
interval. sl other one ral based
based on the sampling
on
vd. The actual selection of the confidence
actual interval is u
meaning of interval of
estimation is depicted confidene .
by the following figu
Point estimate
Confidence interval

Lower confidence Upper confidence

limit limit

Figure 2.4: Interval Estimation Diagram

hus. aparticular kind of interval estimate of a population parameter is known as Confidence Interval (C, Ine nteva
Apecea to nclude the given parameter instead of estimating it by a single value. The authenticity o a c
1ndicted by the confidence intervals
that how likely a parameter is contained in the interval 1s ucl
level or confidence coefficient.

ne contidence interval is enlarged by increasing the desired confidence level. For computing the margin of error an
nterval estimate of a population mean is formulated either by the population standard deviation ( ) or the sample standard
deviation.

This is formulated through

following cases:
1) Confidence Interval for
Population Mean (Small Sample)
2) Confidence Interval for Population Mean
(Large Sample)
i) Confidence Interval for Population Mean (o Known)
ii) Confidence Interval for Population Mean (o Unknown)

2.3.4.2. Confidence Interval for Population Mean (Small Sample)

The procedure of interval estimation of population mean when the population standard deviation is not known and the
sample sizeis small. is based on a probability distribution vwhich is known as the 't-distribution'. The t-distribution is very
similar to the normal distribution but this distribution has more area in the tails and less in the centre, in comparison to
normal distribution.

Degrees of freedom is the parameter which is the basis of t-distribution hence as there is an increase in the degrees of
freedom, t-distribution gradually approaches normal distribution where sample standard deviation (s) is the better estimate
of the standard deviation (G), of the population.

When the sample size is small (n s 30) with confidence coetficient (l -

a), the interval estimate of

given by:
a
population mean is

X
ttaior X-tai2Sus X tn +

Where, tan is the critical value of t-test statistie providing an area o72 in the
degrees of freedom, and
right tail of the t-distribution with (n -
1)

2x,-X
n-1

Critical values of t for the given degrees ol Ireedom can be obtained

from the table of
t-distrihistie
ation.
Example 19: The personnel denartment of
organisation
an would like to
estimate the
employees to determine the feasibilitythousand
of
providing a dental insurance plan A random family dental expenses of
following family dental expenses (in ) in the previous year: 11, 37, sample
25, 62, 51, 21,
of 10 emplovees
reveals the
of the
18, 43, 32, 20.
confidence interval average familydental
Setup a 99 per cent expenses for the
X and standard
employees of this organisa
isation.
Solution: The calculations for sample mean
deviation are shown in
table 2.4:
 Sampling Distribution and Estimation (Unit 2) 99

Table 2.4: Calculation for x and s

Variable (x)(X-X)(X-X
11 21 441
37 25
49
2 900
51 9 361
21 121
18 196
43 121
32 0
20 12 144
320 2358

From the data in table 2.4, the sample mean y 2 X = 320/10 = 732, and the sample standard deviation

x-X
n-1 2 6 2 =16.2. Using this information and ta2 = 1.833 at
df =9,
we have.

Xa = 32 + 1.833 32 =32 +9.3

V10
Hence. the mean expenses perfamily are likely to fall between 22.7 and 41.3, i.e., 22.7 SH < 41.3.

2.3.4.3. Confidence Interval for Population Mean (Large Sample)

Confidence Interval for Population

Mean (Large Sample)

Confidence Interval for

Confidence Interval for
Population Mean (o Known)
Population Mean (o Unknown)

2.3.4.4. Confidence Interval for Population Mean (o Known)

Let us consider that the population mean (j) is unknown and the true population standard deviation (G) is known. Then for
a large sample size (n 2 30), the sample mean X is the best point estimator for the population mean (). Since, sampling
distribution is nearly normal, then confidence interval of population mean ju is calculated as follows:

Xtz/2 O, or X t Zal2

or, X-Za/2SusX+ Zai2

n n
Where, Zan = Z-value, represents an area W2 in the right tail ol the standard normal probability distribution. (1 - a) = Level

of confidence (as shown in figure 2.5)

Confidence
ow2 = 0.025 level /2=0.025
-=95%

X Estumator
a12 0 - a/2 O-

Figure 2.5: Sampling Distribution of Mean x