Sampling techniques in research: which method to use and how to defend it

What are population, sampling frame and sampling unit?

Be clear on these three before you choose a method. Most viva questions on sampling come back to them only.

The population is the group your conclusions should cover, with limits on who, where and when. “Nurses” is not a population. “Staff nurses in government district hospitals in Karnataka, 2025–26” does.

The sampling frame is the list from which you actually select, and almost always it is different from the population. The district roll of teachers will not have contract staff who joined last month. A college register still carry students who stopped coming to class. So write down what your frame misses. Honestly say where your list is weak and examiners don’t mind. If your frame looks too perfect and gaps are not written, then they start doubting.

The sampling unit is the thing you select. It may be a person, household, enterprise, village or college. In a multistage design there is a unit at each stage. India’s Periodic Labour Force Survey shows how to describe this. MoSPI’s note on its sample design calls it a stratified multi-stage design. The first-stage units are Census 2011 villages and Urban Frame Survey blocks. Households are the ultimate units, and they are selected by simple random sampling without replacement.

Probability vs non-probability sampling: which one does your study need?

With probability sampling, chance picks the respondents, and each unit in the frame has a known chance of being selected. So you can put a margin of error on your estimate. With non-probability sampling, the choice comes from you, from your contacts or simply from who is available. Such results are useful, only you can’t stretch them to a whole population with a margin of error.

Look at your research question wording once. If it asks about the “level of”, “proportion of” or “extent of” something in a population, you need probability sampling, or a clear reason why you could not get one. When it is more about how or why something happens, purposive sampling probably suits you. Random selection there wastes fieldwork on people with little to share. Mostly your research design settles this first.

In disproportionate stratified sampling you over-sample a small group so that you can analyse it separately. Nothing wrong in that. But pooled figures then need weights, and many scholars are forgetting this. Cluster samples lose some precision because students of one college share the same teachers, fees and catchment area. So plan for more respondents.

Which sampling technique to use when there is no list? Three Indian examples

These are illustrations, not real studies.

MSME owners in one district

Take a commerce scholar who is studying how small manufacturers in one district adopt digital payments. Her guide tells her to “take 400 random MSMEs”. Random from which list? The public Udyam Registration portal gives national totals on its dashboard. You cannot draw a sample from totals. Also, lot of small units never registered.

Her options, in the order I would try. First, ask the District Industries Centre or an industrial estate association which list they keep. A partial list honestly described is better than nothing. Second, build her own frame in a sample of areas. She picks industrial areas or wards at random, lists every unit found there and samples from that list (two-stage area sampling). Third, if neither works, she can use quota sampling based on known shares by sector and size, and call it that. What she must not do is collect 400 responses through WhatsApp groups and write “random sampling”.

Patients in a hospital

A nursing scholar wants to study medication adherence among diabetic outpatients. New patients come every day, so no fixed list is there. She can do systematic sampling from the OPD register, means every k-th eligible patient after a random start. Or consecutive sampling, where she invites every eligible patient in a fixed period till the target is reached. Consecutive sampling is still non-probability. But it takes away her own choice of whom to approach. She has to write which days and hours she sat in the OPD. A Monday-morning clinic crowd is not the same as Friday evening.

Students across affiliated colleges

Suppose an education scholar wants to measure academic stress among undergraduates in the colleges affiliated to one state university. A list of 60,000 students is rarely available. But a list of colleges exists, from the university’s affiliation records or the AISHE directory of institutions run by the Ministry of Education. So the scholar can stratify the colleges by type, that is government, aided and self-financing. Within each stratum she picks colleges at random. Inside each college she picks sections at random and surveys every student present on that day. She must decide beforehand what happens if a principal refuses (the next randomly drawn college in the same stratum), and report how many refused.

Purposive, snowball and theoretical sampling in qualitative research

Qualitative samples are chosen for what people can tell you. Palinkas and colleagues (2015) wrote a widely cited paper on purposeful sampling. As per them, it means selecting information-rich cases, meaning people who know a lot about the phenomenon or who have lived through it. Write down who qualifies before going to the field. Where a variant fits, name it (maximum variation, typical case, criterion).

Snowball sampling is useful for groups with no list, such as moneylenders or caregivers of children with a rare condition. The problem with snowball samples is that everyone in the chain knows each other. To reduce this, begin with several contacts who are not connected, and keep a record of who referred whom.

Theoretical sampling comes from grounded theory. Glaser and Strauss explained it in The Discovery of Grounded Theory (1967). In this approach the analyst collects, codes and analyses the data together. The theory that is emerging then decides what to collect next and from where. So you cannot fix the full sample in your synopsis itself. Describe your starting sample and the logic you will use for later choices, and keep your memos as evidence. If you are doing thematic analysis and not grounded theory, don’t claim theoretical sampling. Your sample is purposive, and the analysis part is covered in our thematic analysis guide.

How to write the sampling section of your methodology chapter

Skip the textbook definitions. Your examiners know what stratified sampling is. They want what you did, clear enough for someone to repeat it. To see where this part goes in the chapter, check our research methodology guide.

1. 1

   Define the population

   Say who, where and when. For example, owners of registered micro and small manufacturing units in Coimbatore district in 2025. Just writing "MSMEs" is not enough.

2. 2

   Name the sampling frame

   The actual list you drew from, who maintains it, its date and what it misses. If there was no list, write that.

3. 3

   Name the method and each stage

   Technical name first, then how it worked, like how strata or clusters were formed, how random numbers were generated and what you did on refusal.

4. 4

   Give the numbers

   How many you approached, how many agreed, how many were usable after screening. Refer to your sample size calculation instead of repeating it.

5. 5

   State what the sample lets you claim

   One or two sentences on how far the findings can be generalised, and to whom.

A worked example

For the college study described earlier, the paragraph could look like this. The numbers are made up only for illustration.

The population comprised undergraduates enrolled in 2025–26 in arts and science colleges affiliated to the university. The frame for colleges was the university’s list of affiliated colleges (148, as on 1 June 2025). No list of students was available. A two-stage stratified cluster design was used. Colleges were stratified by management type, and four were selected from each stratum using computer-generated random numbers. Two colleges that declined were replaced by the next drawn college in the same stratum. In each college two second-year sections were chosen at random and all students present were invited. Of 1,020 students invited, 874 responded and 841 responses were retained after screening. Because strata of different sizes contributed equal numbers of colleges, pooled estimates were weighted by stratum size. The findings may therefore be generalised only to second-year undergraduates of arts and science colleges affiliated to this university.

See, no definitions anywhere. “Random” is used for the college draw and the sections, and nowhere else. That last sentence on whom the findings apply to, scholars mostly forget. If you used a questionnaire, pilot details come after this, and for that see questionnaire design.

Common examiner objections to sampling

All of these can be corrected easily before you submit to the university. Once the examiner’s report comes, the same correction drags on for months.

• Thesis says “randomly selected”. Actually the scholar just handed questionnaires to teachers sitting free in the staff room. Call it convenience sampling and examiner will not object.
• Only 20 people chosen purposively, and then the thesis makes claims about “entrepreneurs in India”.
• Scholar writes stratified sampling in methodology, but in results strata are mixed and weights are missing.
• Whole classrooms surveyed as clusters, then analysed as if every student got picked one by one. Then standard errors look smaller than they really are.
• Two pages defining every sampling type, and one line on what the scholar actually did.
• No response rate, or 90% for an online survey with no explanation.

Your analysis should match the kind of sample you have. Inferential tests take it that respondents were randomly selected. You can still run them on a convenience sample, as most Indian theses do. Only, describe the results as applying to your respondents. Our guide on choosing a statistical test assumes you are being honest in this way.