SAMPLING

Mohua Chatterjee
Department of Psychology
Bethune College

Prepared for UG III

 Some key words
• Population: Any identifiable and well specified group of individuals
• Finite population: Countable
• Infinite population: Unlimited size
• Parameter: Measure based upon the entire population
• Sample: A smaller representation of population
• Statistic: Measure based upon a sample
• Probability: Equivalent to relative frequency; expressed in terms of fraction
 Advantages of sampling
• Generalizability

• Accuracy is increased because:

▪ The task assigned to the sample can be easily supervised
▪ The volume of work is reduced
▪ The results can be analyzed with great care
▪ Chances of errors in calculation are reduced because of limited number of data
• Cost (time & money) is reduced

• Speed of work is increased

 Steps in sampling design
• To define the universe/population
• To decide the sampling unit (geographical, construction, social, individual)
• To prepare the source list (sampling frame) – should be representative of
population, comprehensive, correct, reliable and appropriate
• To decide the size of sample based on the following criteria:
• Representativeness, efficiency, reliability, cost effectiveness
• Size & variance of population
• Areas of interest
• To decide parameters of interest (types & proportions of subgroups in population)
• To decide the sampling technique as per population size, convenience &
accessibility of elements & budgetary constraint
 Characteristics of a good sample design

• Must result in a truly representative sample

• Must result in small sampling error

• Must be cost effective

• Systematic bias must be controlled

• The results must be generalized to the population

 Types of sample design

On representation basis On element selection basis

Probability Unrestricted
sampling sample

Non probability Restricted

sampling sample
Probability sampling methods Non probability sampling methods

• Simple random sampling • Quota sampling

• Stratified random sampling • Purposive/Judgmental sampling

• Area/Cluster sampling • Accidental/Incidental sampling

• Systematic sampling

• Snowball sampling

• Saturation & dense sampling

 What is Probability sampling ?
• Every item of the population has an equal and independent chance of
selection.
• The investigator must know the size of universe (not always feasible)
• The desired sample size must be clearly specified
Merits:
• Obtained samples are considered representative
• Conclusions are worth generalization
Demerit:
• Sampling error exists because a limited no of elements are chosen as samples.
 Methods of drawing random samples
Computer-determined
Fishbowl method Random number table
randomness

• Can’t be applied with • Easily accessible • Used in case of large

large populations • Requires no formal population
training • Easy to adopt when
• Cumbersome & tedious to
computer facility is
number each element in a available
large population
• Mixing huge number of
slips poses problem
Probability sampling techniques
 Simple random sampling
• Every individual has an equal and independent chance of being selected.
• Sampling with and without replacement
Merits:
• The sample selected is regarded as the true representative of the population
• The sample reflects all important characteristics and segments of the population.
• No classification error
• Easiest and simplest probability sampling technique
• Foundation of all other random sampling techniques
 Demerits of simple random sampling:

• Doesn’t ensure that the elements in small numbers will be included in sample

• The technique doesn’t fully exploit the investigator’s knowledge concerning the
segments of population

• Sampling error is greater than stratified random sampling because of

heterogeneity of the samples drawn through simple random sampling technique
 Stratified random sampling
• The population is divided into 2 or more strata (internally homogeneous, non-
overlapping sub populations) based upon 1 or more criteria.
• A simple random sample is taken from each stratum.
Types:
• Proportionate stratified random sampling
• Disproportionate stratified random sampling

Reasons for stratification:

• Increasing the precision in estimating population attributes
• Convenience: The sampling survey in each unit may be supervised by one person
 Proportionate stratified random sampling
• Samples are drawn randomly in a similar proportion from each stratum of the
population
Composition of a Proportion of each Composition of
population stratum sample
N1 = 3000 0.30 n1 = 300
N2 = 4000 0.40 n2 = 400

N3 = 2000 0.20 n3 = 200

N4 = 1000 0.10 n4 = 1oo

Nt = 10,000 1.00 nt = 1000

 Disproportionate stratified random sampling
• The substrata of the sample are not distributed according to their proportionate
weight in the population.
• The investigator gives equal weight to each substratum. Therefore, some strata are
over represented while some are under represented.
 Proportionate stratified random sampling
Merits:
• Elements in small numbers are also included in sample (increases representativeness)
• Minimizes sampling error (sample has all the characteristics of the parent population)
• Eliminates the necessity of weighing the elements according to their original distribution
in the population.
Demerits:
• The researcher has to know the composition & distribution of population before
sampling.
• Time consuming
• Probability of classification error (putting an element in a wrong stratum)
 Disproportionate stratified random sampling
Merits:

• Less time consuming than proportionate stratified random sampling

• The investigator can give weight to the groups of elements which are not
represented frequently in the population

Demerits:

• Overrepresenting one stratum may introduce bias in the sample

• Not applicable when the researcher has no idea about the population composition

• Probability of classification error

 Area (Cluster) sampling
• Geographical divisions of territory are made on a map and a certain number
of them is drawn randomly as sample.
• The investigator interviews all elements of the randomly drawn areas/clusters.
• Further subdivision & selection of samples can be done if any section contains
extremely heterogeneous elements (Multi-stage sampling).
Application:
• In survey and field researches
• When lists of specific individuals are inaccessible
• Public opinion polls
Multi-stage sampling (diagram)
 Area sampling
Merits:
• Easy to cover large geographical area
• Respondents can be substituted for others within the same section because clusters and
not individuals are sampled.
• Saves time and money
• Flexible – the investigator can employ different forms of sampling in several successive
stages in multi-stage sampling.
Demerits:
• High sampling error
• Unequal sizes of clusters introduces bias into the sample
• Difficult to ensure that individuals included in one cluster are independent of other
randomly drawn clusters
Non probability sampling techniques
 Quota sampling
• The investigator selects number of elements arbitrarily, according to convenience from
different strata of population (“poor man’s proportionate stratified sample”)
Merits:
• Most satisfactory when quick & crude results are desired
• Convenient & less costly
• Guarantees the inclusion of individuals from different strata of population
Demerits:
• Individuals selected may not represent population
• Lacks external validity & generalizability
• Amenable to classification error
• Controls one variable (e.g., sex) and not all variables of theoretical or practical
significance during stratification
 Purposive or Judgmental sampling
• The investigator based on his impression makes a judgement regarding the
concerned cases, which are thought to be typical of the population.
Merits:
• Less costly & readily accessible to investigator (due to non-random nature)
• Convenient
• Guarantees that the sample are relevant to the research design
Demerits:
• No assurance regarding representativeness of sample
• Poor generalizability
• Subjectivity in sampling
 Accidental sampling
• Investigator selects sample according to convenience and economy.
• Named as Convenience sampling (Heiman, 1955)
Merits:
• Most convenient
• Saves time, money & labour
Demerits:
• Poor generalizability
• Scope for investigator’s bias and prejudices
• High sampling error, low validity & reliability
 Systematic sampling
• To draw every nth person from a predetermined list of elements.
• Possesses the traits of both probability & nonprobability sampling.
• Probability sampling First element is randomly selected.
• Nonprobability sampling All other elements except nth one are excluded
 Systematic sampling
Merits:
• Quick method
• Easy to check whether every nth element has been selected
• Easy to use
Demerits:
• All elements except nth elements are ignored
• The sampling error increases if the list is arranged in a particular order (e.g.,
alphabetically)
• No assurance regarding representativeness of sample
• Poor generalizability
 Snowball sampling
• All the persons in a group identify their friends who in turn identify their friends until the
researcher observes that a constellation of friendship converges into some type of a
definite social pattern.
• Indirectly sociometric and is used to study social change and diffusion of information
among specific segments of society
• Is used to study ‘hidden’ population
Merit:
• Studies small informal social groups and their impacts on larger formal structure
Demerits:
• Cumbersome and difficult when N exceeds 100
• Samples are chosen based on subjective choices of the original respondents
 Saturation and Dense sampling
• Coleman (1959) gave the classification

• Saturation sampling: Drawing all elements having characteristics of

interest

• Dense sampling:The researcher selects 50% or more from the population

• The techniques are convenient when N does not exceed 1000