Almost every Indian dissertation that gets into trouble over sampling got into it at the second sentence, not at the sample size. The scholar wrote “The population of the study consisted of all teachers in [state]” and then drew a sample from three schools in one block. Nothing after that can be repaired by a bigger sample.
This guide is about the three decisions that come before the number: which population you are actually talking about, which list you drew from, and which technique you used. Six of them are written out in full across disciplines, and each is followed by the justification paragraph that goes into Chapter 3.
How many is a separate question with its own arithmetic, and it is handled in our guide to calculating sample size with G*Power, the Krejcie-Morgan table and Cochran’s formula. Clinical dental sampling, where OPD recruitment and retrospective records raise their own problems, is covered in our guide to population and sample size for a dentistry dissertation.
Target population, accessible population, sample
Three things, not two, and the middle one is the one Indian drafts leave out.
- Target population. Everyone your conclusions are meant to apply to. Defined by attributes, a boundary and a period, not by a number.
- Accessible population. The part of it you could actually reach within your permissions, your travel budget and your study period.
- Sample. The units drawn from the accessible population by a stated procedure.
Your generalisation reaches the accessible population, and beyond it only by argument. Writing all three, and naming the gap between the first two, is what converts an obvious weakness into evidence that you understood your own design.
The sampling frame: the line nobody writes
A sampling frame is the actual list you drew from. Not the population, the list: the school register, the DISE code list, the customer database extract, the OPD register, the roster of registered firms, the voters’ list for the ward.
State four things about it: what the list is, who maintains it, the date you obtained it, and what it is known to omit. The last one is the honest part. A school register omits children who never enrolled; a customer database omits closed accounts; an OPD register omits patients who never presented. Each omission shapes what your findings can say.
Where a frame genuinely does not exist — street vendors, informal workers, migrants — say so, and explain what you did instead. “No sampling frame was available, so a [technique] approach was used from [starting points]” is a defensible sentence. Silence is not.
Nine techniques, and the limitation each one imports
| Technique | What it requires | The limitation it imports |
|---|---|---|
| Simple random | A complete frame and a randomising mechanism | Can scatter your units across an area you cannot travel |
| Systematic random | A frame and a fixed interval from a random start | Fails if the list has a hidden periodicity matching your interval |
| Stratified random | A frame plus the stratum variable known for every unit | The stratum variable must be on the list, which is often the binding constraint |
| Cluster | A frame of clusters, then all units within selected clusters | Units inside a cluster resemble each other, so effective precision is lower than n suggests |
| Multistage | Frames at each stage | Compounds the error of every stage; each stage must be described |
| Purposive | A stated inclusion criterion | No basis for statistical generalisation; defensible in qualitative work, weak in a survey |
| Quota | Known population proportions for the quota variables | Selection inside each quota is unregulated, so bias enters at the last step |
| Snowball | Initial contacts who know others in the population | Samples a network, not a population; report the number of chains and their seeds |
| Convenience | Availability | The weakest defensible option; usable only when named as such and its consequences stated |
Choosing is not about which technique is best in the abstract. It is about which one your frame and your access permit, and then about writing down what that choice costs you.
Six written sections
Square brackets mark what only you supply. No specimen names a real institution, a real dataset figure or a real finding.
1. M.Ed dissertation, stratified random
The target population comprised all secondary school teachers working in government secondary schools in [district] during the academic year [year]. The accessible population comprised teachers in the [n] government secondary schools of [block] and [block], being the blocks for which the District Education Officer granted permission on [date].
The sampling frame was the list of government secondary schools and their sanctioned teaching posts maintained by the office of the District Education Officer, obtained on [date]. The frame does not record teachers appointed on contract after [date], who are therefore outside the study.
Schools were stratified by location as rural and urban, since the literature reports differences in assessment practice by school location. From each stratum, schools were selected by simple random sampling in proportion to the number of schools in that stratum, and all teachers of the selected schools who consented were included. This yielded [n] teachers, of whom [n] returned usable instruments.
The justification sentence: “Stratification was applied by location because location is the variable on which the literature reports differences in the outcome, and proportional allocation ensures that the sample reflects the distribution of schools in the accessible population.”
2. MBA dissertation, purposive with a screening item
The target population comprised retail banking customers in [city] holding a savings account with a scheduled commercial bank. No frame of such customers is available to a student researcher, since customer lists are confidential.
The accessible population therefore comprised customers approached at [n] branch locations of [category of banks] across [zones] of the city during [period], with the permission of the branch managers concerned, recorded at Appendix [n].
Purposive sampling was used, with the inclusion criterion that the respondent had held the account for at least [n] months and had conducted at least one transaction in the preceding [n] weeks. Respondents were classified as digitally active or branch-visiting using the screening item at Item [n] of the instrument. Of [n] approached, [n] met the criteria and [n] completed the instrument.
The justification sentence: “Purposive sampling was necessary because no sampling frame of customers is available outside the banks themselves; the inclusion criteria are stated so that the sampled population can be described precisely, and statistical generalisation beyond it is not claimed.”
That last clause is the whole move. A purposive sample that admits it cannot generalise is far stronger than a purposive sample that quietly claims it can. The conceptual framework the variables sit in is in our guide to building a conceptual framework for an MBA dissertation.
3. M.Sc Nursing thesis, consecutive sampling
The target population comprised caregivers of [patient group] in [city]. The accessible population comprised caregivers of patients admitted to or attending the [named department] of [type of hospital] during [period], for which institutional permission was obtained on [date].
Consecutive sampling was used: every caregiver meeting the inclusion criteria and presenting during the data collection period was approached until the required sample size of [n] was reached. Inclusion criteria were [list]; exclusion criteria were [list].
Of [n] caregivers approached, [n] declined and [n] were excluded under criterion [n], giving a final sample of [n]. Attrition at the [interval] follow-up was [n], and the analysis at that point is based on [n] participants.
The justification sentence: “Consecutive sampling was adopted because the accessible population presents over time rather than existing as a list, and enrolling every eligible presenter avoids the selection that arises when a researcher chooses among those available.”
The attrition sentence is compulsory in any study with a follow-up, and the sample size has to be inflated for it in advance, not explained away afterwards.
4. M.Tech dissertation, sites rather than people
The population for this study comprised [systems or sites] of the type described in Section [n] operating under [condition] in [region]. The accessible population comprised the [n] such [systems] for which [authority] permitted instrumentation during [period].
Of these, [n] sites were selected purposively to span the range of [governing parameter] observed in the accessible population, from [value] to [value], so that the characterisation is not confined to a single operating regime. Measurements were taken at each site at [interval] over [duration] using [instrument, make and model].
The justification sentence: “Purposive selection across the range of [parameter] was preferred to random selection because the objective is characterisation across operating conditions rather than estimation of a population mean, and random selection from a small accessible population risks clustering within one regime.”
Engineering examiners accept purposive selection readily, provided the purpose is stated as a coverage argument rather than as convenience.
5. M.A. Economics dissertation, a census of units
The study uses secondary data and does not sample. The units of analysis are all [n] [states or firms] for which [named series] is published continuously over [period], which constitutes a census of the defined universe rather than a sample of it.
[Number] units were excluded because the series is discontinuous for them following [reason, for example reorganisation in the relevant year], and the excluded units are named in Appendix [n]. The findings therefore describe [n] units and do not extend to the excluded ones.
The justification sentence: “Since the analysis covers every unit for which the series exists over the study period, no sampling was required; the exclusions arise from data availability and are enumerated rather than assumed away.”
Secondary-data dissertations often skip this section entirely, which is a mistake: defining the universe and naming the exclusions is exactly the same intellectual job. Which series are continuous over which periods is covered in our comparison of data sources for an M.A. Economics dissertation in India.
6. Social work dissertation, snowball sampling
The target population comprised [hard-to-reach group] in [city]. No sampling frame exists for this population, and no register is maintained by any agency contacted during the preparatory phase.
Snowball sampling was therefore used. [Number] initial participants were identified through [named type of organisation], and each was asked at the close of the interview to refer up to [n] others meeting the inclusion criteria. Recruitment proceeded through [n] chains, of lengths [list], until no new themes emerged in [n] consecutive interviews. The final sample was [n].
The justification sentence: “Snowball sampling was the only feasible approach given the absence of any frame; the number of chains and their independent starting points are reported so that the risk of sampling a single network can be assessed by the reader.”
Reporting the chains is what separates a defensible snowball sample from an indefensible one. How many interviews are enough before you stop is addressed in our guide to how many interviews are enough for a qualitative thesis, and the consent requirements for work with vulnerable groups are in our guide to ethics approval and informed consent for a social work thesis.
The four-sentence template
1. The target population comprised [units] in [boundary] during [period].
2. The accessible population comprised [narrower units], being those for which [permission or access condition] was obtained on [date].
3. The sampling frame was [named list], maintained by [body] and obtained on [date]; it does not include [known omission].
4. [Technique] was used because [reason tied to the frame or the objective], yielding [n] units, of which [n] provided usable data.
Four sentences, and Chapter 3 has a sampling section that an examiner can check. Where this section sits in the chapter, and what the sections around it must contain, is set out in our written-out methodology chapter.
Five things examiners flag
- A population stated as a number. “The population was 300 teachers” describes a sample. A population is described by attributes and a boundary.
- No accessible population. The target population is a state and the data came from two schools, with nothing in between.
- A technique named but not performed. “Random sampling was used” with no frame and no randomising mechanism. If you cannot describe the mechanism, the technique was convenience.
- Numbers that do not reconcile. Approached, eligible, consenting, returned, usable. Every drop needs a reason, and the final figure must match Chapter 4.
- Generalisation beyond the accessible population. Chapter 5 talks about the state; Chapter 3 described two blocks. Either qualify the conclusion or explain the basis for extending it.
The variables that these units will be measured on need their own definitions, with instrument, score and cut-off, and those are written out in our guide to operational definitions across six disciplines.
Writing the section before the fieldwork, not after
The sampling section is the one most often reconstructed from memory, which is why the numbers so rarely reconcile. Write the four sentences before you start collecting, record the drops as they happen, and the section writes itself. Tesify keeps the methodology and the results in one view as the document grows, so a sample figure that stops matching its table shows up while there is still time to explain it.
Draft your sampling section in Tesify
Frequently asked questions
What is the difference between the target and accessible population?
The target population is everyone your conclusions are meant to describe. The accessible population is the part you could actually reach given permissions, travel and time. Your generalisation reaches the second directly and the first only by argument.
What is a sampling frame?
The actual list you drew from: a school register, a customer extract, an OPD register, a roster of firms. Name it, say who maintains it, give the date you obtained it, and state what it is known to omit.
What if no sampling frame exists for my population?
Say so explicitly and describe what you did instead, such as purposive recruitment from named starting points or snowball sampling with the chains reported. An acknowledged absence is defensible; an unmentioned one is not.
Is convenience sampling acceptable in an Indian dissertation?
It is accepted at Master’s level in many departments provided it is named as convenience sampling and its consequences for generalisation are stated. What is not accepted is convenience sampling described as random.
How do I justify purposive sampling?
By tying it to the objective or to the absence of a frame, and by stating the inclusion criteria precisely so that the sampled population can be described. Then decline to claim statistical generalisation beyond it.
Do I need a sampling section if I use secondary data?
Yes. Define the universe of units, state the period, say whether you analysed all of them or a subset, and enumerate the exclusions with reasons. The intellectual job is identical.
What is consecutive sampling and is it random?
It enrols every eligible unit that presents during the study period until the target is reached. It is not random, but it removes the researcher’s discretion over who is approached, which is why clinical studies prefer it to convenience sampling.
How do I report the response rate?
As a chain of reconciling numbers: approached, eligible, consenting, returned, usable, and the rate computed from the last two. State a reason for each drop.
Does a low response rate invalidate my study?
Not automatically, but it changes what you can claim, because those who responded may differ from those who did not. State the rate, discuss the likely direction of any bias, and carry the point into the limitations.
Should the sample be proportional to the population?
Under proportional stratified sampling, yes, and that is the usual choice. Disproportionate allocation is legitimate when a small stratum needs enough units for its own analysis, but then weighting has to be discussed.
How many participants do I need?
That is a separate calculation from everything on this page, and it depends on your design, expected effect and precision. The three standard routes are set out in our guide to calculating sample size for a thesis.