Sampling

Techniques
Presented by,
Dr Mirunalini Sundaravadivelu
Post Graduate in Public Health Dentistry
 Contents
Defining the Requirement of a Sampling Principles of
Definition
population good sample terminologies sampling

Factors affecting
Characteristics of a
the interferences Non-sampling Important things
Sampling errors representative
drawn from a errors about sampling
sample
sample

Strengths and
Various sampling Probability Non-probability weakness of
Mixed design
strategies sampling sampling various sampling
techniques

Sampling in
qualitative Referrences
research
 Procedure by which some members of the population are selected as
Definition representatives of the entire population
Defining the population

• A population is a entire group of individuals from which a sample is drawn

• First step is to define the population to be studied
• Definition of the population should be clear and complete, possible to ensure that all
elements within the population are represented
• Often called to as reference or target population
• Study population: The
population to which results of
the study will be inferred
• Sampling unit(BSU): Elementary
unit that will be sampled
• Sampling frame: List of all
sampling units in the population
• Sampling scheme: Method used
to select units from the
sampling frame
Need of sample
population,
• Obtain information from
large populations
• Ensure the efficiency of a
study
• Obtain more accurate
information
 Requirement of a good sample
• A good sample reproduces the characteristics of interest in the population as closely as possible
• A sample is representative in the sense that each sampled unit represents the characteristics of units
in the population
• The method by which sample is selected using a sampling frame is integral to the generalizability of a
survey
• The sample has to be representative of the larger population to obtain a composite profile of that
population(Bowling 2002)
• Good sample is also free from selection bias and measurement error
• Selection bias occurs when some part of the target population is not in the sampled population
• Measurement bias occurs when the measuring instrument has a tendency to differ from true value
in one direction
 Sampling terminology
• The class, families living in the city or electorates from which you select we select our sample are
called the population or study population, and are usually denoted by the letter N.
• The small group of students, families or electors from whom we collect the required information to
estimate the average age of the class, average income or the election outcome is called the sample.
• The number of students, families or electors from whom we obtain the required information is
called the sample size and is usually denoted by the letter n.
• The way we select students, families or electors is called the sampling design or sampling strategy.
• Each student, family or elector that becomes the basis for selecting our sample is called the
sampling unit or sampling element.
• A list identifying each student, family or elector in the study population is called the sampling
frame.
• If all elements in a sampling population cannot be individually identified, we cannot have a
sampling frame for that study population.
• Our findings based on the information obtained from our respondents (sample) are called sample
statistics.
• Our sample statistics become the basis of estimating the prevalence of the above characteristics in
the study population.
• Our main aim is to find answers to your research questions in the study population.
• From sample statistics we make an estimate of the answers to our research questions
in the study population.
• The estimates arrived at from sample statistics are called population parameters or the
population mean.
• It is not possible to include each member (sampling unit) of the population in an
experimental study or enquiry or examine all the millions of people of India to find the
prevalence of tuberculosis or to test the efficacy of a drug in all the patients suffering
from a particular disease.
• A daily life example is that of cooking rice.
• A housewife just picks up a few grains of rice from the cooking vessel and gets a fairly
good idea whether the entire lot of rice is fully cooked or it requires more cooking.
• Further, covering the entire population may be less accurate because a large number
of investigators are required to complete the huge task.
• Their uniformity and correctness may vary, collection will be costly, time consuming
and laborious.
• Because of all such difficulties, we prefer to use an appropriate sampling technique.
• In medical studies, the sampling data are collected from a population or
universe sufficiently large and representative of the population under study.
• Chosen by a standard sampling technique.
• The population or universe must clearly be defined before drawing a sample.
• For example, population may be an entire group of defined people such as
doctors, all members of household, women of 15 to 49 years of age and so on,
about whom information is required.
• It has to be further qualified which doctors, those in service or in general
practice, which women, married or unmarried, etc.
 Principles of sampling
• Principle 1 – in a majority of cases of
sampling there will be a difference
between the sample statistics and the true
population mean, which is attributable to
the selection of the units in the sample.
• Principle 2 – the greater the sample size,
the more accurate the estimate of the true
population mean.
• Principle 3 – the greater the difference in
the variable under study in a population
for a given sample size, the greater the
difference between the sample statistics
and the true population mean.
 Factors affecting the inferences drawn
from a sample
• The size of the sample – Findings based upon larger samples have
more certainty than those based on smaller ones. As a rule, the
larger the sample size, the more accurate the findings.
• The extent of variation in the sampling population – The greater
the variation in the study population with respect to the
characteristics under study, for a given sample size, the greater
the uncertainty. (In technical terms, the greater the standard
deviation, the higher the standard error for a given sample size in
your estimates.) If a population is homogeneous (uniform or
similar) with respect to the characteristics under study, a small
sample can provide a reasonably good estimate, but if it is
heterogeneous (dissimilar or diversified), we need to select a
larger sample to obtain the same level of accuracy. Of course, if
all the elements in a population are identical, then the selection
of even one will provide an absolutely accurate estimate. As a
rule, the higher the variation with respect to the characteristics
under study in the study population, the greater the uncertainty
for a given sample size
The two main objectives of sampling are:

• Estimation of population parameters (mean,

proportion, etc.) from the sample statistics.
• To test the hypothesis about the population from
which the sample or samples are drawn
 Sampling errors
• Selecting different random samples of similar size from the same population would give different
estimates – sample to sample variability
• Sampling error is the degree to which a single sample estimate might differ from the population
• Although sampling error cannot be eliminated entirely, the sampling technique chosen will influence the
extent of the error
• The margin of error given in surveys is an expression of sampling error, the error that results from taking
one sample instead of examining the whole population
• For most large sample surveys, the margin of error of an estimate is the half-width of the confidence
interval i.e, 1.96*standard error(SE) of estimate
• For instance, simple random sampling will give a closer estimate of the population than a convenience
sample of individuals who just happen to be at the right time of selection
 Non-sampling errors
• Selection bias and inaccuracy of responses are examples of non-sampling errors
• These errors are attributed to errors that occur rather than sample to sample variability
• These can be grouped into errors caused by,
Coverage problems with sample frame
Attrition in the sample
Problems in study measurements
• Selection bias can be reduced to a greater extent by using probability sampling methods
• Accurate responses can often be achieved through careful design and testing of survey instrument and
training of interviewers
• In many studies, when the response rate is very low, the non-sampling errors are of more concern than
sampling errors
There are two main characteristics of a representative sample

• Precision which implies the size of the sample

• Unbiased character.
 These are three important things surveyors should keep
in mind about their sample:
• Diversity
• Ensuring that all units in the survey are not similar to each other is a tall order, but it's important. To be a
truly representative sample, the entire group must represent the spectrum of diversity within the
population.
• Consistency
• It's also important that survey respondents have been tracked on a case-by-case basis before going ahead
with the survey. A good idea is to administer a consistency test for a sample, for example a pilot test, where
you compare the individual units of the sample with the whole to make sure it properly represents the
parent population's characteristics.
• Transparency
• Many factors determine the size and the structure of a population. Researchers need to discuss these
limitations and maintain transparency about the procedures followed while selecting the sample so the
survey results are viewed with the right perspective
The various sampling strategies in
quantitative research can be categorised as
follows
• Random/probability sampling designs;
• Non-random/non-probability sampling designs
selecting a predetermined sample size;
• ‘Mixed’ sampling design.
Probability sampling

To ensure true
Each member in the Randomness is an
randomness, the
population has a essential part of the
method of selection
known probability of protection against
must be independent
being selected selection bias
of human judgment
Probability or
random sampling
designs
• Simple random sampling
• Systematic sampling
• Stratified sampling
• Cluster sampling
• Multistage sampling
Simple random sampling

Principle: Equal chance for each sampling unit

Procedure: Number all units, Randomly draw units

Advantages: Simple, Sampling error easily measured, More representative in homogeneous

population

Disadvantages: Need complete list of units, Don’t always achieve best representation in
heterogeneous population
 Simple Random Sampling

• The method is applicable when the population is small.

• Homogeneous and readily available such as patients
coming to hospital or lying in the wards.
• It is used in experimental medicine or clinical trials like
testing the efficacy of a particular drug.
• Hence, this method is also sometimes called
‘unrestricted random sampling’.
• The sample may be drawn unit by unit, either by
numbering the units such as persons, families or
households of a particular population on the cards or
from the published tables of random numbers.
• To ensure randomness of selection, one may adopt
either lottery method or refer to table of random
numbers.
Methods of
drawing a random
sample
• Fishbowl draw/lottery
method
• Computer program
• Table of random numbers
Lottery method/Fishbowl draw
• Suppose, 10 patients are to be put on a trial out of the 100 available,
Not the serial number of patients on 100 cards and shuffle them
well.
• Draw out one and not the number.
• Replace the card drawn, reshuffle and draw the second card.
• Repeat the process till 10 numbers are drawn.
• Reject the cards that are drawn for second time.
• The 10 cards drawn thus will indicate the patient’s number to be put
on trial and the 10 patients selected in this manner form the random
sample.
• Similar procedure may be followed for selecting the control if need
be.
• This is sampling with replacement.
Computer
program
There are a number of
programs that can help us to
select a random sample.
Table of random number method
• The other common method of drawing the sample is by making use of
the published tables of random numbers.
• First give serial numbers to all the 100 patients in the last example at
random, starting from any patient. This reduces the bias at the very
start.
• Total number of patients or population from which desired number 10
is to be chosen is 100, a 3 digit number.
• Numbers, less than 100 may be chosen as they are, while those higher
than 100 may be divided by 100 and the remainders may be not as the
numbers chosen for the sample. Number higher than 100 could be
rejected too. Making use of the rows below.
• Thus, 10 numbers for the sample are chosen.
• We can start from any row or column or even diagonally.
 Random sampling can be selected using two
different systems:
Sampling without replacement;
Sampling with replacement.
• Suppose we want to select a sample of 20 students out of a total of 80.
• The first student is selected out of the total class, and so the probability of selection for
the first student is 1/80.
• When we select the second student there are only 79 left in the class and the
probability of selection for the second student is not 1/80 but 1/79.
• The probability of selecting the next student is 1/78.
• By the time you select the 20th student, the probability of his/her selection is 1/61.
• This type of sampling is called sampling without replacement.
• But this is contrary to our basic definition of randomisation; that is, each element has
an equal and independent chance of selection.
• In the second system, called sampling with replacement, the selected element is
replaced in the sampling population and if it is selected again, it is discarded and the
next one is selected.
• If the sampling population is fairly large, the probability of selecting the same element
twice is fairly remote.
Systematic sampling

Principle: A unit drawn every k units, Equal chance of being drawn for each unit

Procedure: Calculate sampling interval(k=N/n), Draw a random number(less than or

equal to k) for starting, Draw every k units from first unit

Advantages: Ensures representativity across list, easy to implement, time and labor
involved in the collection of sample is relatively small

Disadvantages: Dangerous if list has cycles

 • This method is popularly used in those cases when a
complete list of population from which sample is to be
drawn, is available.

Systematic
• It is more often applied to field studies when the
population is large, scattered and not homogeneous.

Sampling • Systematic procedure is followed to choose a sample by

taking every Kth house or patient where K refers to the
sample interval, which is calculated by the formula
• K = Total population/Sample size desired
• If the population is sufficiently large, homogeneous and
each unit is numbered, this method can yield accurate
results.
 • In some instances simple or systematic random sampling is
not appropriate and an alternative sampling method is
needed. Such situations arise when,
Population consists of a distinct subgroups
A sampling frame doesn't exist, it would be costly and time
consuming to obtain it
Complex Population to be covered is spread over a wide

sampling
geographical area and the travel costs and time involved in
covering the whole area are prohibitive

schemes • They include,

• Stratified sampling
• Cluster sampling
• Multistage sampling
• Either used alone or combination with each other
Stratified sampling
Principle: classify population into homogenous subgroups(strata), Draw sample in
each strata, combine results of strata

Advantages: More precise, more representative since associated with strata, all
subgroups represented, allowing separate conclusions about each of them,
estimates and corresponding confidence interval can be produced for each strata

Disadvantages: Sampling error is difficult to measure, loss of precision if small

numbers sampled in individual strata
 • This method is followed when the population is not
homogeneous.
• The population under study is first divided into
homogeneous groups or classes called strata and the
sample is drawn from each stratum at random in
proportion to its size.
Stratified • It is a method of sampling for giving representation to all

Sampling strata of society or population such as selecting sample

from defined areas, classes, ages, sexes, etc.
• This technique gives more representative sample than
simple random sampling in a given large population
• e.g. the entire population may be divided into five
socioeconomic groups or strata.
• Out of large groups 10% may be selected for study and
50% or even more from the smaller strata so that no
subsample is less than 30 in size.
 • Principle: Non-overlapping random sample of
groups(clusters) of units, then random sample of clusters
are obtained, in each selected cluster all the units are
Cluster examined

sampling • Advantages: simple- No loss of units required, Less

travel/resources required
• Disadvantages: imprecise if clusters homogenous(large
design effect), Sampling error difficult to measure
 • The sampling unit is not a subject, but a group(cluster) of
subjects
Cluster • It is assumed that:

sampling • The variability among clusters is minimal

• The variability within cluster is what is observed in the
general population
 • First stage: Probability proportional to size
• Select the number of clusters to be included
• Compute a cumulative list of the population in each unit
with a grand total
Two • Divide the grand total by the number of clusters and

stages of a obtain the sampling interval

• Choose a random number and identify the second cluster
cluster • Add the sampling interval and identify the second cluster

sample • By repeating the procedure, identify all the clusters

• Second stage:
• In each cluster select a random sample using a sampling
frame of subjects or households
Multistage
sampling
• Principle: Several chained samples, Several
statistical units
• Advantages: No complete list of population
required, Most feasible approach for large
populations, It introduces flexibility in
sampling, which is lacking in other techniques,
It enables the use of existing division and
subdivision which saves extra labor
• Disadvantages: Several sampling lists, Sampling
error difficult to measure
 Multiphase Sampling

• In this multistage, especially in a very large, we want some

estimates for at the national level, we need to do a sampling in
several chains samples and several statistical units are there.
• In this method, part of the information is collected from the
whole sample and part from the subsample.
• In a tuberculosis survey, physical examination or Mantoux test
may be done in all cases of the sample in the first phase; in the
second phase X-ray of the chest may be done in Mantoux
positive cases and in those with clinical symptoms, while
sputum may be examined in X-ray positive cases in the third
phase only.
• Number in the subsamples in 2nd and 3rd phase will become
successively smaller and smaller.
• Survey by such procedure will be less costly, less laborious and
more purposeful.
• Sometime two or more independent samples may be taken by
different survey teams to compare the results.
 Non-probability sampling
• Members are selected from the population in some non
random manner
• Don’t provide every unit in the population a known chance
of being included in the sample
• Also selection process is partially subjective
• The degree to which the sample differs from the population
remains unknown
• It is widely used in some areas but reduce our confidence
that the sample does indeed represent the population from
which it was drawn
• Major problem with these methods is the difficulty in
generalizing the study sample to the whole population
• Results of studies using non-random sampling should be
generalized with caution
• Researches should always avoid these sampling methods
Non-probability sampling

• Convenience (Accidental) sampling

• Judgement (Purposive) sampling
• Quota sampling
• Expert sampling
• Snowball sampling
• Modal instant sampling
• Heterogenicity sampling
 Quota sampling
• The main consideration directing quota sampling is the researcher’s
ease of access to the sample population.
• In addition to convenience, you are guided by some visible
characteristic, such as gender or race, of the study population that
is of interest to you.
• The sample is selected from a location convenient to you as a
researcher, and whenever a person with this visible relevant
characteristic is seen that person is asked to participate in the
study.
• The process continues until you have been able to contact the
required number of respondents (quota).
• Advantage: It is the least expensive way of selecting a sample; you
do not need any information, such as a sampling frame, the total
number of elements, their location, or other information about the
sampling population; and it guarantees the inclusion of the type of
people you need.
• Disadvantages: As the resulting sample is not a probability one, the
findings cannot be generalised to the total sampling population;
and the most accessible individuals might have characteristics that
are unique to them and hence might not be truly representative of
the total sampling population.
Convenience sampling
• Principle: researcher selects a sample by choosing those
who are convenient and easy to select. The subject are
either easiest to select or they are most likely to respond
in the study
• Whereas quota sampling attempts to include people
possessing an obvious/visible characteristic, accidental
sampling makes no such attempt. You stop collecting
data when you reach the required number of
respondents you decided to have in your sample.
• This method of sampling is common among market
research and newspaper reporters.
• It has more or less the same advantages and
disadvantages as quota sampling but, in addition, as you
are not guided by any obvious characteristics, some
people contacted may not have the required information
• Commonest method for selecting participants in focus
group discussions
 Judgement sampling
• Common non-probability method
• Principle: researcher selects a sample deliberately or
purposely on the basis of bis/her own judgement
• The primary consideration in purposive sampling is your
judgement as to who can provide the best information
to achieve the objectives of your study.
• Researcher only go to those people who in your opinion
are likely to have the required information and be willing
to share it with you.
• This type of sampling is extremely useful when you want
to construct a historical reality, describe a phenomenon
or develop something about which only a little is known.
• This sampling strategy is more common in qualitative
research, but when you use it in quantitative research
you select a predetermined number of people who, in
your judgement, are best positioned to provide you the
needed information for your study.
 • The only difference between judgemental sampling and
expert sampling is that in the case of the former it is
entirely your judgement as to the ability of the
respondents to contribute to the study.
• But in the case of expert sampling, your respondents must
be known experts in the field of interest to you.
• This is again used in both types of research but more so in
Expert qualitative research studies.

sampling
• When you use it in qualitative research, the number of
people you talk to is dependent upon the data saturation
point whereas in quantitative research you decide on the
number of experts to be contacted without considering
the saturation point.
• First identify persons with demonstrated or known
expertise in an area of interest to you, seek their consent
for participation, and then collect the information either
individually or collectively in the form of a group.
 Snowball sampling
• Snowball sampling is the process of selecting a sample using networks.
• To start with, a few individuals in a group or organisation are selected and the
required information is collected from them.
• They are then asked to identify other people in the group or organisation, and the
people selected by them become a part of the sample.
• Information is collected from them, and then these people are asked to identify
other members of the group and, in turn, those identified become the basis of
further data collection.
• This process is continued until the required number or a saturation point has been
reached, in terms of the information being sought.
• This sampling technique is useful if you know little about the group or
organisation you wish to study, as you need only to make contact with a few
individuals, who can then direct you to the other members of the group.
• This method of selecting a sample is useful for studying communication patterns,
decision making or diffusion of knowledge within a group.
• There are disadvantages to this technique - The choice of the entire sample rests
upon the choice of individuals at the first stage.
• If they belong to a particular faction or have strong biases, the study may be
biased.
• Also, it is difficult to use this technique when the sample becomes fairly large.
 • Frequent of cases is sample, in this type of sampling we sample the
most frequent cases.
• It can also be seen as the one with the highest happening of value
in a given distribution or the one with most characteristic incident.
Modal • In a lot of formal public informal public opinion polls, for example,
interviewing a typical voter.
instant • There are problems with these types of sampling. First off all how
are we going to know a model of case or typical case?
sampling • We can be able to say that a modal voter is could be any individual
that has average of age, level of educational background and
income in the population.
• But it will not be clear to use the average considering the skewed
distribution of income, for example, and, how would you know
that those three from the variables are only relevant event that will
classify as representative voter? What if religion and ethnicity
background is another factor?
 • Is can be used if we want to include all the opinions or
views and we are not going to consider about representing
of these views per head.
• This sampling is also called sampling of diversity and is
Heterogeneity almost opposite with the modal instant sampling, the
interest is to have a comprehensive variety of ideas, not to
sampling identify the modal instance or typical once.
• We determine that there is a ground of all possible ideas
applicable to some topic and that we want to sample the
population not the population of those people who have
the ideas.
• What the sample study need is ideas not people
 • Systematic sampling has been classified as a ‘mixed’
sampling design because it has the characteristics of both
random and non-random sampling designs.
• In systematic sampling the sampling frame is first divided
Systematic into a number of segments called intervals.
• Then, from the first interval, using the SRS technique, one
sampling element is selected.
• The selection of subsequent elements from other intervals
design: a is dependent upon the order of the element selected in
the first interval.
‘mixed’ • If in the first interval it is the fifth element, the fifth
element of each subsequent interval will be chosen.
design • Notice that from the first interval the choice of an element
is on a random basis, but the choice of the elements from
subsequent intervals is dependent upon the choice from
the first, and hence cannot be classified as a random
sample.
 • As the main aim in qualitative enquiries is to explore the
Sampling diversity, sample size and sampling strategy do not play a
significant role in the selection of a sample.
in • If selected carefully, diversity can be extensively and
accurately described on the basis of information obtained
qualitative even from one individual.
• All non-probability sampling designs – purposive,
research judgemental, expert, accidental and snowball – can also be
used in qualitative research with two differences
 • In quantitative studies we collect information from a
predetermined number of people but, in qualitative
research, we do not have a sample size in mind.
Qualitative • Data collection based upon a predetermined sample size
and the saturation point distinguishes their use in
vs quantitative and qualitative research.
Quantitative • In quantitative research we are guided by our desire to
select a random sample, whereas in qualitative research
we are guided by our judgement as to who is likely to
provide you with the ‘best’ information.
 • Principles and practice of biostatistics – Belavendra
Antonisamy
• Basic course in biomedical research – Lecture 11 –
Sampling
• Omair A. Sample size estimation and sampling techniques
for selecting a representative sample. J Health Spec
2014;2:142-7.
Referrences • Sampling Methods in Research Methodology; How to
Choose a Sampling Technique for Research. International
Journal of Academic Research in Management Volume 5,
Issue 2, 2016, ISSN: 2296-1747
• Epidemiology, Biostatistics and Preventive Medicine – 4th
Edition – James F.Jekel