Who exactly will be in your study, and why should a reader believe they stand for anyone else? Every empirical thesis answers this question, either deliberately in the methodology chapter or accidentally in the viva. The sampling method, not the sample size, decides whether you may generalize from 300 respondents to 20,000 employees, or whether your 15 interviewees were the right 15 people to ask. This guide sets out the probability and non-probability sampling methods a master’s student actually uses, matches them to research questions and analyses, works through a stratified sample and a purposive sample with real numbers, and ends with a paragraph template that examiners recognize as a justified sampling decision.
This article is part of our guide How to Write a Master’s Thesis by the Mastermind PhD team and relates to our Research Proposal support service. Sources last checked: November 2026.
What is a sampling method? A sampling method is the procedure used to select the units (people, organizations, documents, records) that will be studied from a larger population. Probability sampling methods give every unit in a sampling frame a known, non-zero chance of selection, which allows statistical generalization to the population. Non-probability sampling methods select units by convenience, judgment or referral; they cannot support statistical generalization, but they can be the right choice for qualitative depth, hard-to-reach groups or exploratory work.
Why do sampling methods matter more than sample size?
A large sample drawn badly is still biased. If 400 students answer a link posted in one Facebook group, you have precise information about people who use that group and nothing defensible about the student body. A probability sample of 150 drawn from the registrar’s list is smaller, yet it allows a confidence interval and an honest statement of sampling error.
Four terms carry the whole discussion:
- Population: all the units your question is about (for example, the 2,000 secondary-school teachers employed by one district).
- Sampling frame: the actual list you select from (the district’s payroll list, which may omit new hires and include people on leave).
- Sample: the units you selected and attempted to reach.
- Achieved sample: the units that actually provided data, which is what your tables describe.
The gap between the frame and the population is coverage error; the gap between the sample and the achieved sample is non-response. Both must be reported.
What are the probability sampling methods?
| Method | How it works | Thesis example | Best when | Main drawback |
|---|---|---|---|---|
| Simple random | Every unit in the frame has an equal chance; selection by random numbers | 320 teachers drawn from a numbered payroll list of 2,000 using a random number generator | A complete frame exists and the population is fairly homogeneous | Needs the full list; small subgroups may be missed |
| Systematic | Random start, then every k-th unit (k = N / n) | Every 6th patient file from an ordered list of 1,920 files, starting at a random number between 1 and 6 | Lists are long and in no meaningful order | Hidden periodicity in the list can bias the sample |
| Stratified | Divide the frame into strata, then sample randomly within each | Teachers split by school type (primary, preparatory, secondary) with proportional samples | Subgroups differ and you want each represented precisely | Needs stratum membership for every unit in the frame |
| Cluster | Randomly select groups (clusters), then study all or a sample of units inside them | 12 schools drawn at random from 80, then all teachers in those 12 schools surveyed | No individual-level frame exists, or travel costs are high | Larger sampling error for the same n; needs a design effect adjustment |
| Multistage | Clusters, then units within clusters, often in several steps | 10 districts, then 4 schools per district, then 15 teachers per school | National or regional studies | Complex weighting and analysis |
Of the probability sampling methods, stratified sampling is the workhorse of master’s surveys because most research questions include a comparison between groups (gender, sector, years of experience), and stratification makes sure each group reaches a usable size.
📱 Discuss your proposal idea with a specialist — You are told the specialist’s field and degree before work starts, and a meeting is scheduled to review your department’s requirements.
What are the non-probability sampling methods?
| Method | How it works | Thesis example | Best when | Main drawback |
|---|---|---|---|---|
| Convenience | Whoever is available and willing | Students in the three classes the researcher could access | Pilot studies; instrument testing; access is severely limited | No basis for generalization; volunteer bias |
| Purposive (judgment) | Units chosen deliberately for the information they hold | 15 nurses with at least 5 years of ICU experience across three shift patterns | Qualitative studies seeking depth and variation | Depends on the researcher’s judgment; criteria must be documented |
| Snowball (referral) | Initial participants refer others | Freelance translators recruited through two professional contacts, then their referrals | Hidden or hard-to-reach populations | Sample reflects the networks of the first participants |
| Quota | Fill predefined quotas for key characteristics without random selection | 50 men and 50 women shoppers, 25 per age band, approached at mall exits | Quick descriptive studies when no frame exists | Looks representative but carries interviewer selection bias |
| Theoretical | Cases chosen to develop an emerging theory (grounded theory) | After early interviews suggest a role for family support, participants living alone are added | Grounded theory designs | Requires analysis during collection; hard on a fixed timeline |
Patton (2015) lists more than a dozen purposeful strategies; among these sampling methods, the ones most useful in a master’s thesis are maximum variation (deliberately different cases), typical case, criterion sampling (everyone who meets a condition) and expert sampling.
How do you choose between the two families of sampling methods?
| Your situation | Choose | Why |
|---|---|---|
| Research questions ask about levels, differences or relationships in a defined population and you will run inferential statistics | Probability sampling | Confidence intervals and p-values assume random selection |
| A complete list of the population exists and access is granted | Simple random or stratified | The frame makes it feasible |
| The population is spread across many sites with no individual list | Cluster or multistage | Site-level lists are available |
| Research questions ask how or why, and the data are interviews or documents | Purposive | You need the most informative cases, not the average case |
| The population is hidden or stigmatized (undocumented workers, people with a rare condition) | Snowball with a documented referral chain | No frame can exist |
| You are piloting an instrument | Convenience | Representativeness is not the goal yet |
| You want a quantitative survey but no frame is available | Quota, stated clearly as non-probability | Better than convenience, but the results stay descriptive |
Mixing up the two families of sampling methods leads to the forbidden combination: a non-probability sample followed by claims that “the findings generalize to all employees in the sector”. If you must use a convenience sample for a quantitative thesis, run the statistics, but frame the inference as applying to respondents like yours and state the limitation in the methodology and the discussion.
Worked example 1: a stratified sample of teachers
A student studies job satisfaction among the 2,000 teachers in one school district and wants to compare three school types. A sample size calculation (Krejcie & Morgan, 1970) gives 322 for N = 2,000; the student rounds to 320. The payroll frame shows 1,200 primary, 500 preparatory and 300 secondary teachers.
| Stratum | Population | Share | Proportional sample | Equal allocation |
|---|---|---|---|---|
| Primary | 1,200 | 60% | 192 | 107 |
| Preparatory | 500 | 25% | 80 | 107 |
| Secondary | 300 | 15% | 48 | 106 |
| Total | 2,000 | All | 320 | 320 |
Proportional allocation mirrors the population and is right for estimating the district-wide mean. Equal allocation is better if the main question is the comparison between school types, because the secondary stratum would otherwise have only 48 cases. The student states the choice and the reason, then selects within each stratum with a random number generator and documents the seed.
Worked example 2: a purposive sample of ICU nurses
A qualitative study asks how ICU nurses experience moral distress. Instead of “15 nurses who agreed”, the student writes a sampling matrix: at least 5 years of ICU experience (criterion), both public and private hospitals, three shift patterns, and a spread of ages. Recruitment continues until each cell has at least two participants and no new codes appear in two consecutive interviews. Final sample: 15 nurses, 9 from public and 6 from private hospitals, aged 28 to 54. The matrix, the inclusion criteria and the recruitment route (ward managers forwarded an invitation; no manager knew who took part) all go in the methodology chapter.
How do you report the sample in the methodology chapter?
Examiners expect six items, in this order: the target population and its size; the sampling frame and its known gaps; the method and why it was chosen over the alternative sampling methods; the sample size and how it was calculated or justified; the achieved sample and response rate; and how non-response or exclusions were handled.
A template for a quantitative study: “The target population comprised the 2,000 teachers employed by the district in 2025/26. The payroll list served as the sampling frame; 38 teachers on long-term leave were excluded. A stratified random sample of 320 (proportional to school type) was drawn using R’s sample() function with a recorded seed. Of these, 274 returned usable questionnaires (an 85.6% response rate); non-respondents did not differ from respondents in school type (chi-square test, p = .41).”
A template for a qualitative study: “Participants were selected purposively using maximum variation sampling on hospital sector, shift pattern and experience, with a minimum of five years in intensive care as the inclusion criterion. Fifteen nurses took part; recruitment stopped when the sampling matrix was filled and two consecutive interviews produced no new codes.”
Which mistakes with sampling methods do examiners flag?
- Calling a sample “random” when it means “haphazard”. Random selection requires a frame and a random mechanism.
- Posting a survey link on social media, then reporting confidence intervals as if the sample were probability-based.
- Not reporting the response rate, or reporting the achieved sample as if it were the drawn sample.
- Stratifying on a variable that is never used in the analysis, while the variable that is actually compared has 14 cases in one group.
- Using cluster sampling but analyzing the data as a simple random sample, which understates the standard errors.
- In qualitative work, defending the sample by its size (“12 is enough”) instead of by its composition and the information it yielded.
📱 Ask on WhatsApp now — 5 years of experience across all disciplines and 14 academic services. Two-stage payment (80/20); revisions are free until supervisor approval, within 30% of the chapter.
Frequently asked questions
Is convenience sampling acceptable in a master’s thesis?
It is accepted in many programs when access is genuinely limited, provided you label it correctly, describe who the respondents were, and restrict your conclusions to them. It is not acceptable to describe a convenience sample as random or to generalize its results to a population.
What response rate is good enough?
There is no universal threshold. Online academic surveys often achieve 20% to 40%; organizational surveys with management support can exceed 70%. What matters is comparing respondents with the population on known characteristics and discussing non-response bias honestly.
How do I draw a random sample in practice?
Number the frame, then use a random number generator: RAND() in Excel followed by sorting, the sample() function in R, or Data > Select Cases > Random sample in SPSS. Record the seed or save the file so the selection can be audited.
What is the difference between quota and stratified sampling?
Both set target numbers per subgroup. Stratified sampling selects randomly within each stratum from a frame; quota sampling fills the targets with whoever is available. Only stratified sampling supports statistical inference.
How many participants does snowball sampling need?
Report the number of seeds, the number of referral waves and the final count, typically 10 to 30 in a master’s thesis. Because the sample reflects the seeds’ networks, start with seeds that differ from one another and say so.
Sources
- Cochran, W. G. (1977). Sampling techniques (3rd ed.). Wiley.
- Levy, P. S., & Lemeshow, S. (2008). Sampling of populations: Methods and applications (4th ed.). Wiley.
- Bryman, A. (2016). Social research methods (5th ed.). Oxford University Press.
- Patton, M. Q. (2015). Qualitative research & evaluation methods (4th ed.). SAGE.
- Krejcie, R. V., & Morgan, D. W. (1970). Determining sample size for research activities. Educational and Psychological Measurement, 30(3), 607–610. https://doi.org/10.1177/001316447003000308
- Etikan, I., Musa, S. A., & Alkassim, R. S. (2016). Comparison of convenience sampling and purposive sampling. American Journal of Theoretical and Applied Statistics, 5(1), 1–4. https://doi.org/10.11648/j.ajtas.20160501.11
- Palinkas, L. A., Horwitz, S. M., Green, C. A., Wisdom, J. P., Duan, N., & Hoagwood, K. (2015). Purposeful sampling for qualitative data collection and analysis in mixed method implementation research. Administration and Policy in Mental Health and Mental Health Services Research, 42(5), 533–544. https://doi.org/10.1007/s10488-013-0528-y
Sources last checked: November 2026. This article is general educational guidance and does not replace your university regulations or your supervisor’s instructions. How we prepare our content.