Choose your sampling technique by asking who must be studied, how you can identify them, and what you want to claim about them. Then write the actual selection steps. Naming “simple random sampling” without showing how anyone was selected leaves the most important part unexplained.
This guide covers selection. Use the separate sample-size guide for the number of participants and the methodology guide for the whole chapter. Agree the design with your supervisor before recruiting people.
Your methodology should describe the selection process you actually used. UniDraft helps organise your confirmed sampling decisions, real evidence and guidelines into an editable project document. Start your project document, keeping the selection record available for your final review. It does not recruit respondents or turn a convenience sample into a random one.
Start with a selection worksheet
Fill this in before choosing a method name:
| Decision | What to record |
|---|---|
| Research question | The exact question this group can help answer |
| Target population | Who qualifies, where and during which period |
| Sampling unit | A person, household, school, record or another unit |
| Available frame | The dated list or other basis from which units can be selected |
| Coverage problem | Eligible people missing from that frame; duplicates or outdated entries |
| Intended claim | A population estimate, a comparison or an account of selected experiences |
| Access and permission | How you may contact people and protect their information |
A class WhatsApp group is not automatically a list of every eligible student. Check who is absent, who has graduated and whether the same person appears twice. If you cannot obtain a suitable frame, discuss a realistic design rather than inventing one in Chapter Three.
Probability and non-probability sampling: the practical difference
In probability sampling, selection uses chance and eligible units have known, non-zero inclusion probabilities. Those probabilities need not be equal. In non-probability sampling, selection probabilities are not established in that way; researcher choices, access or volunteering may determine who participates. Statistics Canada explains these two families and their limitations.
| Method | Selection idea | What you must be able to explain |
|---|---|---|
| Simple random | Randomly select units from a suitable frame | Frame, random procedure and whether selection is without replacement |
| Systematic | Choose a random start, then use an interval | List order, interval and any repeating pattern |
| Stratified random | Divide into defined groups, then sample randomly within each | Group sizes, allocation and within-group selection |
| Cluster | Select groups, then study units within selected groups | Stages, probabilities and analysis that respects the design |
| Convenience | Recruit people readily accessible | Access route and who it leaves out |
| Purposive | Select cases that meet a reasoned criterion | Why those cases can answer the question |
| Quota | Fill specified category totals without random selection within them | Quotas and how people inside each category were recruited |
| Snowball | Ask participants or contacts to refer eligible people | Starting points, eligibility checks and network limitations |
These are selection summaries, not instructions to use every method. Having equal numbers of men and women, for example, does not by itself make recruitment random.
Worked example: a systematic selection plan
Fictional teaching example, not a real study or recommended sample size. Suppose an authorised, checked list contains 240 eligible students and an independently justified plan requires 40 selections.
- Assign each entry an ID from 1 to 240 and keep the list order fixed.
- Calculate the interval: 240 ÷ 40 = 6.
- Draw a random start from the integers 1 to 6. Suppose it is 4.
- Select IDs 4, 10, 16, 22, continuing every sixth entry through 238. That gives 40 selections.
- Retain the selection record securely; separate contact details from the analysis file.
Before using this design, inspect the ordering. If every sixth entry has the same characteristic because of how the list was assembled, the procedure can select an unbalanced group. The Statistics Canada probability-sampling guide explains this periodicity problem. The example above uses a whole-number interval; do not improvise rounding rules for other cases without methodological guidance.
Selection is not completed participation. If 40 people are selected and 31 respond, record those as separate counts. Do not quietly invite nine friends and describe all 40 as the original random sample. Follow your agreed contact/non-response plan, record deviations and explain how missing responses may limit the findings.
An online questionnaire does not choose your sample for you
Compare these two plans:
- Open-link plan: publish a questionnaire link in groups and accept eligible volunteers. Explain the recruitment channels, screening and self-selection limitation.
- Selected-invitation plan: select people through a documented sampling procedure, then invite only those selected through an appropriate approved process. Explain how eligibility, access and duplicate responses were handled.
Both can use Google Forms. The software name does not decide which sampling method was used. A large volunteer sample still has selection limitations; a formula cannot remove them.
Do not attach a conventional random-sampling margin of error to an open-link convenience survey merely because a calculator produced one. Population inference from non-probability data requires additional assumptions and methods that must be justified; seek statistical advice where those claims matter.
Write the method so someone can follow the decision
Use this as a planning worksheet, not a completed-method paragraph to paste:
The eligible population is [definition]. The selection frame is [source and date], with [known coverage limits]. The planned sample size is [number and justification]. Units will be selected by [actual procedure]. Recruitment and non-response will be handled by [agreed plan]. The findings will be interpreted within [scope and limitations].
After collection, replace planned statements with what actually happened, including the number invited, the number responding and the number analysed. Do not claim ethical approval or permission that was never granted. Record exclusions in your coding and cleaning log.
Prepare for these defence questions
- Why can this population answer your research question?
- Who was missing from your frame or recruitment channels?
- How did the first participant enter the sample? How did the next one?
- What happened when a selected person did not respond?
- Which wider claims would your method not support?
Keep your answers tied to your selection record, not a memorised definition. If the actual process differed from the proposal, explain the change and its consequences honestly.
Is studying everyone a census or a sample?
A census aims to collect information about every unit in the defined population; a sample survey collects information from a selection. Statistics Canada's data-type guide makes that distinction. A small eligible population may make an attempted census practical, but “small” alone does not settle access, consent, cost or the analysis plan.
Fictional teaching example: a programme has 38 eligible students. You invite all 38, 31 return questionnaires and 29 have usable responses under your pre-agreed rules. Report those three counts. Do not write “all 38 students participated” or describe the 29 as a random sample. Explain the attempted census, non-response and exclusions separately.
Reaching everyone would remove selection by sampling for that defined population, but not every possible error. Incomplete coverage, misunderstood questions, missing answers and processing mistakes can still affect a census, as the non-sampling-error guide explains. Nor does a census of one programme automatically justify conclusions about every Nigerian undergraduate.
Before deciding on total enumeration, record the eligible count and date, whether each person can be contacted appropriately, the follow-up plan, and how you will report incomplete participation. If your department uses “total population sampling” or a similar label, explain the actual procedure instead of relying on the label alone.
Choosing interview participants purposively
Purposive or purposeful selection chooses cases for their relevance to the question. It is not a licence to interview anyone convenient and call the group representative. Palinkas and colleagues' methodological paper discusses purposeful selection in qualitative and mixed-method implementation research; the appropriate strategy still depends on your own design.
Use this original planning worksheet:
| Decision | What your record should show |
|---|---|
| Relevant experience | What makes someone able to discuss the question |
| Eligibility | Clear inclusion/exclusion criteria, without collecting unnecessary personal details |
| Useful variation | Which differences matter to the question and why |
| Recruitment route | Who can circulate an invitation, and how participation remains voluntary |
| Access limits | Which perspectives that route may miss |
| Review of recruitment | What your approved design says about checking adequacy or changing recruitment |
Fictional planning example: for interviews about revising reports after supervisor feedback, eligibility might require having received written comments and attempted a revision during the defined period. Recruiting only class representatives could miss other experiences. Interviewing your friends may be convenient, but it does not establish that you deliberately included the perspectives your question needs.
Do not copy a universal interview count from another project. Explain the rationale your design uses and the participants you actually recruited. Do not claim “saturation” merely because the last two interviews sounded similar; whether and how that concept fits depends on the analytic approach. If using reflexive thematic analysis, follow its methodological assumptions rather than importing a different approach's stopping rule.
For snowball recruitment, referrals help find possible participants; they do not remove the eligibility check. Avoid collecting someone else's contact details without the appropriate permission. Record how the initial contacts and referral networks could shape the accounts you hear.
Stratified sampling is more than filling group quotas
If you need defined groups represented through a probability design, explain how groups are formed and how selection occurs within each. Filling category totals with whoever volunteers is not the same procedure as random selection within strata.
Fictional allocation exercise, not a recommended sample size: a frame contains groups of 120, 80 and 40 people, totalling 240. If a justified plan calls for 60 selections allocated proportionally, the group allocations are 30, 20 and 10: each group size divided by 240, multiplied by 60. Random selection would then take place within each group's frame using the agreed procedure.
Record the group counts, allocation and selection steps. Actual participation may differ from the planned allocation. Unequal allocations, clustering, non-response and any weighting need appropriate analysis; do not fix those issues by changing labels after collection. These examples help you explain a decision, not replace a statistician's design advice.
Where UniDraft fits
When reporting the people who actually took part, use the response-denominator and case-outcome worksheet. A planned sample, returned questionnaires and analysed cases are different counts; keep them connected to the selection record.
UniDraft can help organise your final-year project around the method, facts and supervisor instructions you confirm. You remain responsible for the selection procedure, permissions, data collection and analysis. It does not recruit respondents or run your statistics.
Sources and scope
- Statistics Canada: probability sampling, for random selection, systematic sampling and design distinctions.
- Statistics Canada: non-probability sampling, for selection approaches and caution about participation bias.
- Statistics Canada: types of data and non-sampling error, for census/sample distinctions and errors that remain.
- Palinkas et al. (2015): Purposeful Sampling for Qualitative Data Collection and Analysis in Mixed Method Implementation Research, for the purposeful-selection principles; not a prescription for every student interview study.
The worksheet, numerical illustration and defence prompts are original teaching aids. They do not replace a supervisor-approved research design.
