How to Choose a Sampling Method and Sample Size for Your Research

You usually cannot collect data from every person, organisation, document, or case relevant to your research.

If your study concerns 20,000 university students, for example, surveying all 20,000 may be unrealistic.

Instead, you select a sample from the larger population.

But choosing a sample involves more than deciding how many participants you want.

You need to answer two different questions:

How will I select the participants or cases?

How many do I need?

These are your sampling method and sample size.

A large sample selected poorly can still produce misleading evidence. A smaller sample may be appropriate for another research design if its selection and size are properly justified.

The goal is therefore not to find one “correct” sample size or automatically choose the easiest participants to reach.

Your sampling decisions should follow from your research question, population, methodology, analysis, and practical constraints.

Population vs sample

Your population is the broader group your research is concerned with.

Your sample is the subset from which you actually collect data.

Suppose your study investigates academic stress among undergraduate students at University X.

You might define:

Target population: All undergraduate students enrolled at University X.

Sample: The undergraduate students selected to participate in your study.

This distinction matters because your conclusions may be based on the sample but intended to say something about a broader population.

The stronger that intended generalisation is, the more carefully you need to think about how the sample represents that population.

Define your target population clearly

Before deciding how to sample, establish who or what belongs to the population.

“University students” may be far too broad.

You might instead define your population as:

Full-time undergraduate students enrolled at University X during the 2026 academic year.

Depending on the research question, you may need further boundaries involving:

• age;

• programme;

• year of study;

• geographical location;

• employment status;

• diagnosis or condition;

• organisational role;

• exposure to a particular experience; or

• another relevant characteristic.

Your population should follow from the study rather than being defined simply by who is convenient to recruit.

A focused research question helps establish whose evidence you actually need.

What is a sampling frame?

A sampling frame is the practical list or source from which a sample can be selected.

If your population is all enrolled undergraduate students at a university, an appropriate student register might provide the sampling frame.

But a sampling frame may not perfectly match the target population.

It could:

• omit eligible people;

• contain outdated records;

• include people who are no longer eligible;

• overrepresent some groups; or

• exclude people who are difficult to identify.

That difference matters because even a technically random selection can be biased if the sampling frame itself does not adequately represent the population.

Ask:

Who has a realistic chance of entering my sample, and who does not?

Probability and non-probability sampling

Sampling methods are often divided into two broad families.

Probability sampling

In probability sampling, selection uses a random mechanism and eligible population members have known or calculable selection probabilities under the design.

Common approaches include:

• simple random sampling;

• systematic sampling;

• stratified sampling; and

• cluster sampling.

Probability sampling is especially useful when the study seeks statistical inference to a defined population and an adequate sampling frame is available.

Non-probability sampling

In non-probability sampling, participants are not selected through a design in which every population member has a known selection probability.

Common approaches include:

• convenience sampling;

• purposive sampling;

• quota sampling; and

• snowball or chain-referral sampling.

Non-probability sampling is not automatically poor research.

It may be appropriate when the study seeks information-rich cases, investigates a difficult-to-reach population, uses qualitative methods, or operates in circumstances where probability sampling is impractical.

The important issue is whether the sampling method fits the research purpose and the claims you intend to make.

Simple random sampling

In simple random sampling, eligible population members are selected using a random process so that each has a defined chance of selection under the design.

Imagine a university has 5,000 eligible students and you have an accurate list containing all of them.

You could assign each student an identifier and randomly select the required number.

The strength of this approach is that selection does not depend on the researcher's personal choice of participants.

However, simple random sampling requires an adequate sampling frame and may still encounter problems such as non-response.

Random selection does not guarantee that everyone selected will participate.

Systematic sampling

Systematic sampling selects units at regular intervals from an ordered sampling frame, usually after choosing a random starting point.

For example, after determining an appropriate interval, you might select every tenth eligible record from a list.

This can be simpler to administer than drawing many independent random selections.

However, you need to consider how the list is ordered.

If there is a meaningful repeating pattern in the ordering that corresponds to your interval, systematic selection could introduce bias.

Stratified sampling

Stratified sampling can be useful when important subgroups within the population need adequate representation.

Suppose your university population contains students from four faculties.

Instead of drawing one simple random sample from the entire university, you might divide the population into faculty-based strata and sample within each stratum.

This can help ensure that important groups are represented and, with an appropriate design, may improve statistical precision.

But stratification should have a reason.

Do not create strata merely because demographic information is available.

Ask whether the subgroup is relevant to the research question, population structure, or intended analysis.

Cluster sampling

Sometimes the population is geographically or organisationally dispersed and obtaining a complete list of individuals is difficult.

You may instead be able to identify natural groups or clusters, such as:

• schools;

• classrooms;

• hospitals;

• villages; or

• branches of an organisation.

A cluster-sampling design selects clusters and then studies individuals within selected clusters according to the design.

This can reduce the practical cost of data collection.

However, people within the same cluster may be more similar to one another than people selected independently across the entire population. That clustering can affect statistical precision and therefore sample-size and analysis decisions.

Do not treat cluster sampling as simple random sampling with a different label.

Convenience sampling

Convenience sampling involves recruiting participants largely because they are readily accessible.

For example:

Surveying students who happen to be available in the library.

It is easy and inexpensive.

That does not make it representative.

Students using the library at a particular time may differ systematically from students who are elsewhere.

Convenience sampling may still be defensible for some exploratory, pilot, feasibility, or otherwise appropriately limited studies.

The important thing is to describe the limitation honestly.

Do not use a convenience sample and then write as though every member of the target population had an equivalent opportunity to participate.

Purposive sampling

Purposive sampling deliberately selects participants or cases because they have characteristics, experiences, or knowledge relevant to the research question.

Suppose you are conducting qualitative research on the experiences of nurses who implemented a particular electronic health-record system.

You do not necessarily need a random selection of everyone employed by the hospital.

You need participants who can meaningfully speak about that experience.

Possible inclusion criteria might require participants to:

• be registered nurses;

• work in the relevant departments;

• have used the system for a defined period; and

• have direct experience of its implementation.

Purposive sampling is common in qualitative research because the goal may be depth and relevance rather than statistical representation of a population.

Snowball sampling

Snowball or chain-referral sampling involves existing participants helping researchers identify or recruit other potentially eligible participants.

It can be useful when members of a population are difficult to identify or reach through conventional sampling frames.

However, participants' social networks can shape who enters the sample.

People connected to one another may share characteristics, experiences, or perspectives.

That limitation should be considered when interpreting the evidence.

Snowball sampling is not simply “ask participants to bring their friends.” It should have a defensible connection to the access problem and research population.

How do you choose the right sampling method?

Start with the study rather than with a favourite technique.

Ask these questions.

1. What is my research question?

Your sampling method must provide access to the people or cases capable of answering the question.

2. What is my methodology?

A large quantitative survey intended to estimate population characteristics has different sampling requirements from an in-depth qualitative interview study.

Acadelyra's research methodology guide explains why research design should follow the question and the evidence required.

3. What population do I want to make claims about?

If you intend to estimate characteristics of a large defined population, probability sampling may be particularly important where feasible.

If you are investigating detailed experiences among people with a specific characteristic, purposive sampling may make more sense.

4. Do I have a suitable sampling frame?

Probability sampling is difficult if you cannot identify the population from which selection will occur.

5. Are important subgroups involved?

If subgroup representation or comparison matters, stratification or another suitable design may need consideration.

6. How accessible is the population?

Some populations may be difficult to identify, contact, or recruit.

7. What can I realistically complete?

Time, money, access, ethics, geography, and participant availability all matter.

Feasibility should shape the design, but convenience should not silently replace methodological reasoning.

Sampling method should align with your aims and objectives

Suppose one of your research objectives is:

To compare academic stress between undergraduate nursing and engineering students.

If your sample contains 190 nursing students and only 4 engineering students, the planned comparison may be difficult to support meaningfully.

Likewise, if your objective concerns employees who have experienced organisational restructuring, recruiting employees regardless of whether they experienced the restructuring may produce evidence that does not match the objective.

Sampling is therefore not an isolated methodology paragraph.

It connects directly to what the study promises to investigate.

What is sample size?

Sample size is the number of participants, observations, cases, documents, or other units included in the study or analysis.

Students often ask:

“How many participants do I need?”

There is no universal answer.

The appropriate sample size depends on factors such as:

• research design;

• research question;

• intended analysis;

• population structure;

• expected effect size;

• required precision;

• variability;

• significance level;

• statistical power;

• sampling design;

• expected non-response or attrition; and

• practical and ethical constraints.

For qualitative research, different considerations usually apply.

That is why statements such as:

“30 participants are always enough”

or

“10% of the population is always the correct sample”

are not reliable general rules.

Quantitative sample-size decisions

In quantitative research, sample size should be connected to the study's primary analysis or estimation goal.

For a hypothesis-testing study, relevant considerations may include:

• the statistical test or model;

• expected effect size;

• desired statistical power;

• significance level;

• variability;

• allocation between groups; and

• the sampling design.

For an estimation study, you may instead focus on:

• desired precision;

• confidence level;

• expected prevalence or variability; and

• population size where relevant.

For more complex models, additional considerations may apply.

The key principle is:

Calculate or justify sample size for the research you are actually conducting.

Do not choose a formula from another dissertation simply because the equation looks familiar.

What is statistical power?

Statistical power is the probability that a statistical test will detect an effect of a specified size when such an effect exists under the assumptions of the calculation.

All else equal, very small samples may have insufficient power to detect effects that matter.

But simply increasing the sample indefinitely is not the goal either.

A sample-size calculation should be based on justified assumptions.

If you assume an unrealistically large expected effect, for example, the resulting required sample may appear conveniently small.

Where possible, assumptions should be informed by:

• previous research;

• pilot data;

• established measures;

• substantive knowledge; or

• a clearly justified minimum effect of interest.

If your study tests a research hypothesis, the sample-size reasoning should correspond to the hypothesis and planned analysis rather than being an unrelated calculation.

Sample size for surveys

Survey sample size is not determined simply by the number of questions in the questionnaire.

Depending on the study, you may need to consider:

• population size;

• desired margin of error or precision;

• confidence level;

• expected variability;

• subgroup analyses;

• sampling design;

• non-response; and

• planned statistical modelling.

If you need reliable estimates for several subgroups, the overall sample may need to be larger than a calculation designed only for the full population.

And if only a proportion of invited participants are likely to respond, you may need to invite more people than the final required number of completed responses.

Distinguish between:

number invited and final analytical sample.

They are not necessarily the same.

Qualitative sample size

Qualitative research requires different reasoning.

The goal is often not to estimate a population percentage or achieve statistical power.

Instead, researchers may consider factors such as:

• the research question;

• study design;

• participant diversity;

• specificity of the sample;

• complexity of the phenomenon;

• richness and depth of the data;

• analytical approach; and

• whether additional data continue to contribute meaningfully to the analysis.

You may encounter the concept of saturation, especially in some forms of qualitative research.

But saturation should not be treated as a magic sentence that automatically justifies any sample size.

Different qualitative traditions conceptualise adequacy differently, and researchers should explain what they mean and how it relates to their design.

Do not write:

“Ten participants are enough for qualitative research.”

without a study-specific methodological justification.

Mixed-methods studies may need two sampling arguments

A mixed-methods project may contain a quantitative component and a qualitative component.

Those components may require different sampling strategies and sample-size justifications.

For example:

Quantitative phase: A probability-based survey designed to examine relationships across a student population.

Qualitative phase: Purposively selected interview participants chosen to explore particular patterns identified in the survey.

The qualitative sample does not automatically need to be a random subset of the quantitative sample.

Its selection should follow the purpose of that phase and the overall mixed-methods design.

Representativeness is not just about sample size

A very large sample can still be systematically biased.

Imagine an online survey about student wellbeing receives 5,000 responses.

That sounds impressive.

But suppose students experiencing severe academic difficulties are much less likely to check university email or complete voluntary surveys.

Increasing the number of responses does not automatically eliminate that selection problem.

Representativeness depends on who can enter the sample, who actually participates, and how those processes relate to the research question.

Do not use sample size as a substitute for thinking about sampling bias.

Consider non-response and attrition

Not everyone you invite will necessarily participate.

And in longitudinal or intervention research, some participants may leave before the study ends.

Suppose your analysis requires 300 complete responses.

If you expect only 60% of invited participants to respond, inviting exactly 300 people is unlikely to produce 300 completed responses.

Likewise, a longitudinal study may need to account for expected attrition.

But anticipated response or attrition rates should be justified rather than invented merely to inflate recruitment numbers.

Where possible, use evidence from:

• similar previous studies;

• pilot work;

• institutional experience; or

• relevant published research.

Inclusion and exclusion criteria

Sampling also requires clear rules about who is eligible.

Inclusion criteria

These define characteristics participants must have.

For example:

• currently enrolled undergraduate student;

• aged 18 years or older;

• completed at least one semester;

• enrolled at University X.

Exclusion criteria

These define circumstances that make otherwise relevant individuals ineligible.

For example:

• exchange students enrolled for less than one semester;

• students currently on formal leave of absence.

Criteria should follow from the research question and design.

Do not create arbitrary exclusions merely to make recruitment easier or manipulate the eventual results.

Sampling bias

Sampling bias occurs when the process of selecting or recruiting the sample systematically favours some members or characteristics over others in a way that matters to the study.

Possible sources include:

• incomplete sampling frames;

• convenience recruitment;

• self-selection;

• non-response;

• excluding difficult-to-reach groups;

• recruitment through limited networks; or

• differential dropout.

You may not be able to eliminate every source of bias.

Your responsibility is to design the strongest feasible sampling strategy, recognise important limitations, and avoid making claims stronger than the sample supports.

How to justify your sampling method

A weak methodology statement says:

Convenience sampling was used because it was convenient.

That describes the method but provides almost no academic justification.

A stronger explanation might say:

Purposive sampling was used to recruit nurses with direct experience implementing the hospital's electronic health-record system because the study sought detailed accounts of that specific implementation experience.

The justification connects:

sampling method → population → research question → methodology.

For probability sampling, you might similarly explain why random or stratified selection is appropriate for the population-level inference or subgroup comparison the study intends to make.

How to justify your sample size

Do not simply state:

The sample size was 200.

Explain why.

A quantitative justification might identify:

• the primary outcome or analysis;

• calculation method;

• assumed effect or precision;

• significance level;

• desired power;

• expected non-response;

• design effect where applicable; and

• resulting recruitment target.

A qualitative justification might explain:

• why the selected participants are information-rich;

• anticipated diversity of perspectives;

• the analytical approach;

• the depth of data collection;

• how adequacy will be evaluated; and

• why the sample is feasible for the study.

The reader should be able to see that the number came from methodological reasoning rather than guesswork.

A worked sampling example

Suppose your research question is:

What relationship exists between weekly study time and academic performance among first-year undergraduate students at University X?

Step 1: Define the population

All first-year undergraduate students enrolled at University X during the study period.

Step 2: Identify the sampling frame

An up-to-date university enrolment register, assuming appropriate access and ethical approval.

Step 3: Consider the sampling method

If students come from several faculties and faculty differences matter, you might consider stratified random sampling to ensure appropriate representation.

Step 4: Identify the primary analysis

The study intends to examine the relationship between weekly study time and academic performance.

Step 5: Determine the sample size

The required sample should be justified using the planned analysis and appropriate statistical assumptions rather than choosing an arbitrary percentage of the student population.

The researcher should consider factors such as:

• expected effect size;

• desired statistical power;

• significance level;

• expected variability;

• the stratified design, if used; and

• anticipated non-response.

Step 6: Check feasibility

Suppose the calculation indicates that 350 completed responses are required.

The researcher then needs to consider whether:

• enough eligible students can be contacted;

• the expected response rate makes the target realistic;

• each important stratum can contribute enough participants; and

• the available time and resources allow the recruitment plan.

If not, the solution is not simply to reduce the sample until recruitment becomes convenient.

The researcher may need to reconsider the design, analysis, recruitment strategy, or scope of the study.

Step 7: Explain the decision

The methodology should then explain:

who constituted the target population, how the sampling frame was obtained, why the sampling method was appropriate, how sample size was determined, and how anticipated non-response was handled.

That gives the reader a logical chain:

Research question → population → sampling frame → sampling method → sample-size justification → recruitment.

Common sampling mistakes

Choosing participants only because they are easy to reach

Accessibility matters, but convenience alone is not a methodological justification.

Claiming random sampling without using a random process

Sending a questionnaire link to everyone in a WhatsApp group and allowing volunteers to respond is not simple random sampling.

Using a formula without understanding its assumptions

A sample-size equation is not automatically appropriate simply because another dissertation used it.

Treating a large sample as automatically representative

Thousands of responses can still be biased if important groups have little chance of entering the sample.

Ignoring non-response

The number invited and the number providing usable data may differ substantially.

Choosing sample size before deciding the analysis

Your analytical plan is often important to quantitative sample-size reasoning.

Applying quantitative rules to qualitative research

Statistical power calculations generally answer a different question from qualitative sample adequacy.

Claiming saturation without explaining it

If saturation is relevant to your qualitative approach, explain what you mean and how it will be assessed.

Forgetting subgroup requirements

A total sample may look adequate while containing too few participants for an important planned comparison.

Hiding limitations

A defensible sampling limitation is better than pretending your sample supports conclusions that it cannot.

Sampling and your conceptual framework

Your sampling decisions should also fit the concepts and relationships your study proposes to investigate.

Suppose your conceptual framework proposes that employment status moderates the relationship between study time and academic performance.

If almost everyone in your sample has the same employment status, the study may have little useful information for investigating that proposed moderation.

Likewise, if your framework includes several population groups, the sampling strategy needs to consider how those groups enter the study.

This is another example of why research design should be treated as one connected system rather than separate sections written independently.

A sampling checklist

Before finalising your sampling plan, ask:

• What exactly is my target population?

• What are my inclusion and exclusion criteria?

• Do I have an adequate sampling frame?

• Does my sampling method fit the research question and methodology?

• Am I using probability or non-probability sampling, and why?

• Can I justify why participants or cases have been selected this way?

• What claims do I intend to make from the sample?

• How was the sample size determined?

• Does the calculation or justification match my planned analysis?

• Have I considered important subgroups?

• Have I allowed appropriately for non-response or attrition?

• Could the recruitment process systematically exclude important participants?

• Is the sample feasible within the project's time and resources?

• Can I explain the important limitations honestly?

If you cannot explain why these participants and why this number, your sampling plan probably needs more work.

Final takeaway

Sampling is not simply the part of your methodology where you state how many questionnaires you distributed.

It determines whose evidence enters your study and what conclusions that evidence can reasonably support.

Start with the research question.

Define the target population carefully.

Decide what sampling frame is realistically available.

Choose a sampling method that matches your research purpose and intended claims.

Then justify sample size using reasoning appropriate to your design rather than relying on a universal number or copying a formula from another study.

For quantitative research, that may require formal calculations based on the planned analysis, precision, power, effect size, and sampling design.

For qualitative research, adequacy usually requires a different justification based on the question, participants, analytical approach, richness of the data, and methodological tradition.

And throughout the process, remember:

A bigger sample does not automatically fix a weak sampling strategy.

The better question is not simply:

“How many participants do I need?”

It is:

“Who needs to be represented in this study, how should they be selected, and what sample will provide evidence strong enough for the conclusions I intend to make?”

When you can answer those questions clearly, your sampling plan becomes much easier to defend.