Sampling in Research: A Comprehensive Guide

Introduction to Sampling
Sampling is simply the process of choosing a smaller group of people or things from a larger
population to study. Researchers rarely study everyone in a population because it would take too
much time, cost too much money, or simply be impossible. Instead, they select a sample that
represents the larger group.

For example, if you wanted to understand the study habits of all university students in your country,
you wouldn't interview every single student. That would be impossible. Instead, you would select a
smaller group of students who represent the larger population.

The decisions researchers make about who to include in their study, how to include them, and
whether to reveal or protect their identities are all part of sampling. These decisions affect the
quality of the data collected and what conclusions can be drawn from it.

Different Research Paradigms and Their Sampling Approaches

Postpositivist Paradigm
Researchers who work within the postpositivist paradigm prefer probability sampling. According to
Kathleen M. T. Collins (2010), probability sampling involves randomly selecting participants so that
everyone in the population has an equal chance of being chosen. The goal is to make statistical
generalizations meaning researchers want to apply their findings from the sample to the larger
population.

For instance, if you randomly select 500 students from all universities in a country, you can make
predictions about all university students based on what you find in your sample.

Constructivist Paradigm

Constructivist researchers prefer purposive or theoretical sampling. As Collins (2010) explains, these
researchers begin by identifying groups, settings, and individuals where the processes they want to
study are most likely to occur. They strategically choose cases that will provide deep information or
unique perspectives.

For example, if you're studying how teachers adapt to new technology, you might deliberately select
teachers who are known to be innovative with technology, rather than randomly selecting teachers.

One important aspect of constructivist sampling is clearly explaining the criteria used to select
participants and describing their characteristics in detail. This helps readers understand whether the
findings might apply to their own situations—a concept called transferability (Guba and Lincoln,
2005).

Transformative Paradigm

Researchers in the transformative paradigm may use either probability or purposive sampling,
depending on whether they use quantitative, qualitative, or mixed methods. However, they make
sampling decisions with special awareness of representing groups that have been traditionally
underrepresented in research.

According to Baez (2002), transformative research confronts injustice and violence. Researchers in
this paradigm worry about hidden power arrangements that might influence who is included or
excluded from research. They also consider how the research process itself might disenfranchise
people in three ways:
1. By uncovering hidden power arrangements that could harm participants

2. By actions of researchers who might be dishonest

3. By researchers failing to expose power arrangements once they discover them

Pragmatic Paradigm

Pragmatic researchers can use methods from any paradigm if they help answer their research
questions. Since pragmatists often use mixed methods, they emphasize selecting sampling designs
appropriate for both quantitative and qualitative approaches (Collins and Onwuegbuzie, 2013).

Indigenous Paradigm

In the Indigenous paradigm, building relationships and following cultural protocols is the first step in
sampling decisions. According to the Work Group for Community Health and Development (2013),
relationships are powerful foundations for change.

Researchers in this paradigm consult with community elders and tribal associations, forming
community advisory committees to guide decisions about approaching and engaging participants. As
Chilisa (2020) notes, this paradigm is open to ideas from other paradigms, but all sampling strategies
must be based on culturally respectful entry into communities.

Bird-Naytowhow et al. (2017) demonstrated this approach by beginning with consultation with
community elders and tribal associations, then forming a community advisory committee to guide
their study of youth.

Understanding Population and Sample

Defining Your Population

Research constructs like "racial minority" or "Deaf student" can be defined in two ways:

Conceptual definitions use other concepts to explain meaning. For example, "working mothers"
might be conceptually defined as women who have children and are employed.

Operational definitions specify exactly how the construct will be measured. For example, "working
mothers" might be operationally defined as women with children under 18 living at home who work
at least 20 hours per week outside the home.

Researchers typically begin with a conceptual idea of who they want to study and then create a
formal definition through reviewing literature.

The Experimentally Accessible Population

In the postpositivist paradigm, the operational definition of the population is called the
experimentally accessible population—the actual list of people who fit the conceptual definition.

For example, if you want to study first-grade students, the experimentally accessible population
might be all first-grade students in your school district whose names are in the district database.

Your sampling frame is the actual list you use to select your sample. Examples include:

· Student enrolment lists

· Lists of clients receiving services at a clinic

· Professional association membership directories

· Telephone directories

Researchers must ask whether these lists are complete and up-to-date, and who might have been
left off. For instance, lists of clients at a mental health clinic exclude people who need services but
haven't sought them. Telephone directories exclude people without phones, those with unlisted
numbers, and most mobile phone users.

Population Validity

When the accessible population accurately represents the target population, this establishes
population validity. If the lists are inaccurate, systematic error can occur because of differences
between the true population and the study population.

Nonresponse and Attrition

Researchers must report and explain nonresponse (people not participating) or attrition (people
dropping out). This affects both internal and external validity (or credibility and transferability) of
findings.

Dimensions of Diversity in Sampling

Researchers must consider what dimensions of diversity are important to their study. Questions to
consider include:

· What sampling strategies provide a fair picture of diversity within important target populations?

· What dimensions of diversity are important in different groups?

· How can researchers avoid assuming all members of a cultural group are the same?

· Why is considering diversity important in your specific research context?

Sampling for Focus Groups: Example from Deaf Community

The presentation mentions sampling for focus groups in the Deaf and hard-of-hearing community,
highlighting the importance of representing diversity within groups that might appear
homogeneous.

Probability-Based Sampling Strategies

Probability-based sampling is recommended because researchers can mathematically analyze
possible bias and likely error. Sampling error is the difference between the sample and the
population and can be estimated for random samples.

In random samples, every member of the population has a known, nonzero probability of being
included. "Random" means the selection of each unit is independent of others. Random selection
can be done through lottery procedures, random number tables, or computer-generated random
numbers.

When a sample is drawn to be approximately the same as the population on studied variables, it is
considered representative.

Simple Random Sampling

In simple random sampling, each member of the population has an equal and independent chance
of being selected. Researchers can assign numbers to everyone in the population and use random
number tables or computer programs to select participants

Advantages: Simple process and compatible with many statistical tests.

Disadvantages: A complete list of the population might not be available, or subpopulations of

interest might not be equally represented.

Random Digit Dialing (RDD) is a type of simple random sampling used in telephone surveys. It
involves generating random phone numbers, eliminating problems of out-of-date directories and
unlisted numbers.

Systematic Sampling

In systematic sampling, researchers take every nth name on the population list. For example, if you
have 1,000 people and need 100 participants, you would select every 10th person. You would
randomly pick a starting point and then select every 10th name.

Advantages: You don't need an exact list of all sampling units; you just need to know how many
people are in the accessible population.

Caution: If files are arranged in a specific pattern, this could result in biased samples. For instance, if
files are alphabetical and n results in selecting only people whose last names begin with certain
letters.

Stratified Sampling

Stratified sampling is used when there are subgroups (strata) of different sizes you wish to
investigate. For example, if studying gender differences in special education, you'd stratify by gender
because boys are more frequently represented in special education.

Proportional stratified sampling means the sampling fraction is the same for each stratum, resulting
in different sample sizes for each group. This provides greater precision and reduces sampling error,
especially when variance between groups is large. However, you need information on the stratifying
variable for every member of the accessible population.

Disproportional stratified sampling is used when there are big differences in subgroup sizes. It
requires different fractions for each subgroup and weighting in analysis to adjust for selection bias.
This reduces variability within smaller subgroups by including more observations. The disadvantage
is that weights must be used in analysis.

Cluster Sampling

Cluster sampling uses naturally occurring groups—like city blocks or classrooms. Researchers
randomly choose clusters and then study all households or individuals in those clusters.

This approach is useful when:

· A full listing of individuals isn't available but a listing of clusters is

· Site visits are needed (saves time and money)

Disadvantage: In calculating sampling error, the number used for sample size is the number of
clusters, not the number of individuals. The mean for each cluster replaces the individual sample
mean. This reduction in sample size results in larger standard error and less precision.

Multistage Sampling

Multistage sampling combines different sampling strategies. For example, researchers might use
cluster sampling to randomly select classrooms, then use simple random sampling to select students
within each classroom.

Caution: Statistics for multistage sampling become quite complex. Researchers need to know that
too few strata will yield unreliable results. Between roughly 30 and 50 strata work well for
multistage samples using regression analysis.

Non-Probability Sampling Strategies

Non-probability sampling uses non-random criteria like availability, geographical proximity, or expert
knowledge. In non-probability sampling, each unit in the target population does not have an equal
chance of being included.

Important warning: This type of sampling has a higher risk of research biases, particularly sampling
bias.

Convenience Sampling

Convenience sampling is primarily determined by convenience to the researcher, including:

· Ease of access

· Geographical proximity

· Existing contact with the population

Convenience samples are sometimes called "accidental samples" because participants might be
selected simply because they happen to be nearby.

Example: Standing outside a shopping mall and stopping people as they exit to ask about their
shopping patterns.

Quota Sampling

In quota sampling, researchers select a predetermined number or proportion of units (a quota). The
quota should comprise subgroups with specific characteristics selected in a non-random manner.

Proportional quota sampling is used when the population size is known. For example, if a company
has 600 drivers and 400 train-riders among 1,000 employees, a sample of 100 employees would
include 60 drivers and 40 train-riders.

Non-proportional quota sampling is used when population size is unknown. For example, you might
decide to include 50 people under 40 and 50 people over 40 in your sample.

Key difference from stratified sampling: In stratified sampling, you take a random sample from each
subgroup. In quota sampling, selection is non-random, usually through convenience sampling. Who
is included is left to the researcher's subjective judgment.
Example: Interviewing 20 homeowners and 20 tenants between 45-60 in a certain suburb by
standing at a busy shopping street and talking to people who appear to meet the criteria.

Self-Selection (Volunteer) Sampling

Self-selection sampling relies on participants who voluntarily agree to be part of the research. This is
common for studies needing people with specific criteria, especially in medical or psychological
research.

The process involves:

1. Publicizing your need for subjects

2. Checking suitability of each volunteer and either inviting or rejecting them

Example: Advertising for long-distance runners to participate in a mindfulness study by placing

posters near parks or stadiums.

Important: Not all applicants will be eligible. Some may not fully understand what the study involves
or may have disqualifying factors.

Snowball Sampling

Snowball sampling is used when the population is hard to reach or there's no existing database to
help find participants. This method is common for studying socially marginalized groups like drug
addicts, homeless people, or sex workers.

The process:

1. Find one person willing to participate

2. Ask them to introduce you to others they know

Example: Studying homeless people by attending a housing advocacy meeting, connecting with a
homeless woman who agrees to participate, and being introduced to others in the community.

Important: When studying vulnerable populations, researchers must follow ethical considerations
and guidelines.

Purposive Sampling

Purposive sampling is a broad term for several techniques that deliberately choose participants
because of qualities they possess. It's also called judgmental sampling because it relies on the
researcher's judgment.

Purposive sampling is common in qualitative and mixed methods research, especially when studying
specific issues with unique cases. The idea is to concentrate on people with particular characteristics
who will help answer research questions.

Unlike random samples that deliberately include diverse cross-sections of ages, backgrounds, and
cultures, purposive sampling focuses on specific characteristics. The sample isn't representative of
the population, but for certain qualitative and mixed methods designs, this isn't an issue.

External Validity and Transferability

External Validity
External validity refers to the ability to extend research findings beyond the specific individuals and
setting where the study occurred. In the postpositivist paradigm, external validity depends on the
sampling strategy design and execution.

Generalizability is linked to the target population—the group to which researchers want to

generalize findings.

Transferability in Constructivist Research

In the constructivist paradigm, every case or process is viewed as both:

· An exemplar of a general class of phenomena

· Particular and unique in its own way (Denzin and Lincoln, 2018b)

The researcher's task is to provide thick description about the case so readers can understand the
contextual variables operating in that setting. The burden of generalizability then falls on readers,
who can subjectively generalize from the case to their own experiences.

Guba and Lincoln (2005) label this type of generalizability as transferability.

Common Concerns Across Paradigms

Despite contrasting views on sampling across different paradigms, researchers share common
concerns:

1. Ethics: All sampling decisions must be made within ethical constraints

2. Feasibility: Sampling must be practical and possible

3. Reporting nonresponse: All approaches should report and explain nonresponse or attrition

Conclusion

Sampling is a crucial aspect of research that affects the quality and usefulness of findings. The
approach researchers take depends on their research paradigm, goals, and practical constraints.
Whether using probability or non-probability sampling, researchers must carefully consider who to
include, how to include them, and how to address ethical issues.

Understanding the various sampling strategies and their appropriate applications helps researchers
make informed decisions that strengthen their studies and contribute meaningful knowledge to
their fields.