SAMPLING

ARCH 5175
Assoc. Prof. Dr.-Ing. Sudipti Biswas
Sampling = The process of selecting a subset of
individuals from a larger population to represent the whole.
§ A sample is a subgroup of the population you are interested in.
§ Why do we need sampling?
• Saves time, cost, and resources compared to studying the entire
population.
• Ensures inferences can be made about the whole population.
§ The goal is to answer research questions about the
population, not just the sample.
§ Purpose:
• Save time & resources
• Increase feasibility of research
• Collect data from inaccessible populations
• Improve quality of data through focused study
• Facilitate quicker decision-making
• Population (Study Population, N):
The entire group relevant to your study (e.g., all students in a class, families in a
city, or electors in a region).
• Sample:
The smaller group selected from the population to collect data from.
• Sample Size (n):
The number of individuals in the sample.
• Sampling Design / Strategy:
The method used to select the sample.
• Sampling Unit / Element:
Each individual (student, family, elector) considered for selection.
• Sampling Frame:
The complete list of all units in the population. If this list does not exist, a proper
sampling frame is not possible.
• Sample Statistics:
The values (e.g., average age, income) calculated from your sample data.
• Population Parameters:
The estimated values for the whole population, derived from sample statistics.
§ Sample Size – Larger samples reduce sampling error and increase
accuracy.
§ Population Variation – Greater differences in the population require
larger samples for precision.
§ Sampling Procedure – Appropriate method selection (probability
vs. non-probability) ensures representativeness.
§ Participation Rate – Non-response bias can occur when
participation is low or selective.
§ Population Size – Small total populations may require a higher
sampling fraction.
§ Data Collection Quality – Errors in measurement or recording can
undermine sample accuracy.
§ Time & Resources – Constraints may limit sample size or diversity,
affecting results.
§ Accessibility of Population – Hard-to-reach groups may be
underrepresented.
§ Random/probability/mathematical sampling designs
§ Non-random/non-probability/theoretical sampling designs
§ ‘Mixed’ sampling design
Random sampling scheme is one in which every unit in the
population has a chance (greater than zero) of being
selected in the sample, and this probability can be
accurately determined
§ With Replacement (WR):
• A selected unit is returned to the pool and can be selected again.
• Each draw is independent.
• Used in certain statistical simulations.
§ Without Replacement (WOR):
• Once selected, a unit cannot be chosen again.
• More common in surveys.
Fishbowl / Lottery Method
§ Write each unit’s identifier (e.g., name, number) on slips of paper.
§ Mix them thoroughly in a container (“fishbowl”).
§ Draw required number of slips with or without replacement.
§ Best for small populations.
Random Number Table
§ Use a pre-printed table of random numbers.
§ Assign a number to each element in the population.
§ Select numbers according to table sequence until sample size is met.
§ Good for medium-sized populations without computer access.
Computer / Random Number Generator
§ Use software (Excel, R, SPSS, online randomizers).
§ Quickly generate truly random samples.
§ Ideal for large populations and reduces human error.
• Minimizes selection bias. • Requires a complete and accurate
sampling frame.
• Easy to understand and implement
(especially with SRS). • Can be costly and time-consuming
for large populations.
• Results are highly representative if
population list is complete. • Stratification may be necessary if
population is heterogeneous.
• Supports statistical generalization.
• With WR can produce duplicate
cases (not always desirable).
§ Simple Random Sampling (SRS)
§ Stratified Sampling
§ Cluster Sampling
§ Applicable when population is small, homogeneous & readily available
§ All subsets of the frame are given an equal probability.
§ Each element of the frame thus has an equal probability of selection.
§ It provides for greatest number of possible samples.
§ Selection purely by chance
§ Lottery method
§ Random number generator
§ Computer program

§ Can be with or without replacement.

§ Population is divided into homogeneous
subgroups (strata) based on characteristics
(e.g., gender, age), then random samples are
drawn from each.
§ Basis of stratification must be known before
sampling
§ Ensures representation from all key
subgroups.
§ proportionate stratified sampling: the
number of elements from each stratum in
relation to its proportion in the total population
is selected
§ disproportionate stratified sampling:
consideration is not given to the size of the
stratum.
• Population divided into clusters of homogeneous units (often by geography or
identifiable characteristics).
• Sampling units are groups, not individuals.
• All units in chosen clusters are studied.
• Useful when population is large and spread out.
• Two-stage process is most common
• Select clusters (e.g., areas, organizations) using SRS.
• Select elements within clusters using SRS.

§ Multi-stage cluster sampling extends the idea:

• You might first select regions, then districts within those regions, then villages, and finally
households within villages.
• Each stage can use different sampling methods (often SRS or systematic sampling).
§ A sampling method where not all members of the population have a known or
equal chance of being selected. Selection depends on researcher judgment,
convenience, or other non-random criteria.
§ Non-probability sampling designs are used when the number of elements in a
population is either unknown or cannot be individually identified or not feasible.
§ In such situations the selection of elements is dependent upon other
considerations.
§ Often used in exploratory research, qualitative studies, or when a specific subset of
the population is needed.
Quota sampling
Accidental/convenience sampling
Judgemental or purposive sampling
Expert sampling
Snowball sampling.
§ The main consideration is the researcher’s ease of access to the sample population.
§ Guided by some visible characteristic, such as gender or race, of the study
population that is of interest.
§ The sample is selected with convenience and continues until the quota is met.
• Pros:
• Ensures representation of certain groups.
• Least expensive way of selecting a sample

• Cons:
• Selection within quotas is not random, scope of bias
§ findings cannot be generalised
§ might not be truly representative
§ A type of non-probability sampling which
involves selecting whoever is easiest to reach.
§ Based upon convenience and stop collecting
data when you reach the sample size
§ Common among market research, newspaper
reporters.
§ This type of sampling is most useful for pilot
testing.

§ Pros: Quick, inexpensive.

§ Cons: Highly prone to bias; may not represent
the population.
• Selection based on the researcher’s judgment about who can provide the most
relevant and reliable information.
• Researcher directly approaches individuals most likely to have the needed
information and be willing to share it.
• Use Cases:
• Constructing a historical reality
• Describing a specific phenomenon
• Exploring topics with limited prior knowledge
• Research Context:
• Common in qualitative research
• In quantitative research, involves selecting a predetermined number of individuals
best positioned to answer the study’s questions.
§ Selecting respondents who are known experts in the field of interest.
§ Common in both qualitative and quantitative research, but more frequent in
qualitative studies.
§ Process:
§ Identify individuals with demonstrated/recognized expertise.
§ Seek their consent to participate.
§ Collect information individually or in groups (e.g., expert panels).
§ Snowball sampling is the process of selecting a sample using networks.
• Process
• Start by selecting a few individuals from the group or organisation
• Collect data from them, then ask them to identify others
• Referred individuals join the sample
• Continue until desired sample size or data saturation is reached

• Useful when little is known about the group being studied, hard-to-reach
populations.
§ Limitations:
• Sample depends heavily on initial participants
• Risk of bias if first participants have strong views or belong to a faction
• Becomes challenging to manage as the sample grows large
§ the sampling frame is first divided into a number of segments called intervals.
§ Then, from the first interval, using the SRS technique, one element is selected.
§ The selection of subsequent elements from other intervals is dependent upon the
order of the element selected in the first interval. If in the first interval it is the fifth
element, the fifth element of each subsequent interval will be chosen.
§ 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.
There are 50 students in a class, and
you want to select 10 students.

§ First, determine the width of the

interval (50/10 = 5).
§ This means that from every five you
need to select one element.
§ Using the SRS technique, from the
first interval (1–5 elements), select
one of the elements.
§ Suppose you selected the third
element.
§ From the rest of the intervals you
would select every third element.
§ Qualitative research
§ The main focus is to explore or describe a situation, issue, process or phenomenon, the
question of sample size is less important
§ Data is usually collected to a point where you are not getting new information or it is
negligible – the data saturation point. This stage determines the sample size.

§ Quantitative research
§ At what level of confidence do you want to test your results, findings or hypotheses?
§ With what degree of accuracy do you wish to estimate the population parameters?
§ What is the estimated level of variation (standard deviation), with respect to the main
variable you are studying, in the study population?
1. Population size:
§ Population size is how many people fit your demographic.
§ For example, you want to get information on doctors residing in North America. Your
population size is the total number of doctors in North America.
2. Confidence level:
§ The confidence level tells you how sure you can be that your data is accurate. It is
expressed as a percentage and aligned to the confidence interval.
§ For example, if your confidence level is 90%, your results will most likely be 90% accurate.

3. The margin of error (confidence interval):

§ There’s no way to be 100% accurate when it comes to surveys. Confidence intervals tell
you how far off from the population means you’re willing to allow your data to fall.
4. Standard deviation:
§ Standard deviation is the measure of the dispersion of a data set from its mean. The
higher the dispersion or variability, the greater the standard deviation and the greater
the magnitude of the deviation.
§ Define population size or number of people
§ Designate your margin of error
§ Determine your confidence level
§ Predict expected variance
§ Finalize your sample size

Confidence level corresponds to a Z-score, a constant value needed for this

equation. Here are the z-scores for the most common confidence levels:

§ 90% – Z Score = 1.645

§ 95% – Z Score = 1.96
§ 99% – Z Score = 2.576
For unknown population

assuming you chose a 90% confidence

level, .6 standard deviation, and a
margin of error (confidence interval) of
+/- 4%.

((1.64)2 x .6(.6)) / (.04)2

( 2.68x .0.36) / .0016
.9648 / .0016
=603
For known population

In an office of 500 people, with a 95%

confidence level and 5% margin of error:
Sampling Bias happens when the people or units you include in your sample
are not truly representative of the population, leading to results that are
systematically wrong.
§ Poor Sampling Frame – Your list misses certain parts of the population (e.g., using
a phone directory excludes those without phones).
§ Non-Random Selection – Choosing participants based on convenience or
availability rather than random chance.
§ Under coverage – Certain groups have little or no chance of being included.
§ Nonresponse Bias – People who don’t participate differ in important ways from
those who do.
§ Survivorship Bias – Only including those who "survived" a process, ignoring
those who dropped out.
• Use a proper sampling frame – Ensure your list covers the entire target
population.
• Choose the right sampling method – Prefer probability sampling (random,
stratified, cluster) over convenience sampling.
• Increase sample size – Larger samples reduce random error and improve
representativeness.
• Improve participation rates – Use reminders, incentives, and flexible survey
modes to reduce nonresponse bias.
• Weight the data – Adjust results statistically to account for underrepresented
groups.
• Pilot test – Identify potential bias early before full-scale data collection.