To write the research methodology chapter, describe your design, sampling, instruments, analysis and ethics clearly enough for another researcher to repeat the study, and justify each choice against your research questions; the methodology chapter is where examiners test whether your findings can be trusted. It is also the chapter that generates the most viva questions. A weak methodology chapter describes what was done. A strong one explains why each choice was the right one for your questions, and admits what the design cannot show.
This guide covers what goes in the chapter, the main research designs and when each fits, how to present sampling, instruments and analysis, and how to justify every choice. It is written mainly for empirical theses in management, social sciences, education, health and engineering. The same principles apply to the methods section of a paper, in fewer words. Last reviewed September 2026.
What the methodology chapter has to do
It has two jobs. First, describe your procedures clearly enough that another researcher could repeat the study. Second, persuade the examiner that those procedures could actually answer your research questions.
The second job is where most chapters fall short. Many Indian theses contain long textbook definitions: what research is, what sampling is, the types of sampling, the types of research design. Examiners skip these pages. They want to know which design you used and why it suits your questions. Cut the definitions and spend the words on justification.
A common structure is:
- Research philosophy or approach (brief, and only where your field expects it)
- Research design
- Population, sampling frame, sampling method and sample size
- Data sources and instruments
- Pilot study, reliability and validity
- Data collection procedure
- Methods of analysis and tools
- Ethical considerations
- Limitations of the method
Engineering and science theses often call it “materials and methods” or “proposed methodology” and replace sampling and instruments with materials, equipment, experimental set-up, datasets and evaluation metrics. The logic is the same.
Choosing and justifying the research design
Your design follows from your research question, not the other way round. A question about “how much” or “how many” points to a descriptive survey. A question about “does X affect Y” points to a correlational, experimental or quasi-experimental design. A question about “how” or “why” people experience something points to qualitative work.
| Design | Use it when you want to | Typical data | Example |
|---|---|---|---|
| Descriptive (survey) | Describe the level or pattern of something in a population | Questionnaire, secondary data | Awareness of crop insurance among farmers in Thanjavur |
| Correlational / explanatory | Test relationships between variables, often with a model | Questionnaire with validated scales; panel data | Effect of work–life balance on nurses’ intention to quit |
| Experimental | Establish cause and effect by manipulating a variable with random assignment | Measurements from treatment and control groups | Effect of a new teaching method on test scores, with randomised classes |
| Quasi-experimental | Estimate an effect where random assignment is not possible | Pre/post measures, comparison groups | Learning outcomes in schools before and after a state programme |
| Qualitative (case study, grounded theory, phenomenology, ethnography) | Understand meanings, processes or experiences in depth | Interviews, focus groups, observation, documents | How first-generation women entrepreneurs obtain credit |
| Mixed methods | Combine breadth and depth, or use one to build or explain the other | Both of the above, in a planned sequence | Interviews to build a scale, then a survey to test it |
| Design and development (engineering) | Build and evaluate a system, algorithm or material | Simulation, experiments, benchmark datasets | A deep learning model for paddy disease detection, tested on field images |
For the full catalogue of designs, with the question each one answers, see types of research design.
Philosophy and approach
Management and social science examiners often expect a short statement of your research philosophy (positivist, interpretivist, pragmatist) and approach (deductive or inductive). The “research onion” from Saunders, Lewis and Thornhill’s textbook is widely used to structure this. Keep it to a page. It should explain why your design follows from how you view the problem, not list every philosophy that exists. In engineering and the natural sciences this section is usually left out.
Justify against the alternatives
The strongest justification names the obvious alternative and says why you did not use it. “A cross-sectional survey was chosen because the objective was to estimate prevalence across four districts. A longitudinal design would have allowed stronger claims about change over time but was not feasible within the registration period.” That shows the examiner you considered the choice, and it answers the viva question before it is asked.
Population, sampling and sample size
Be precise at each step:
- Population: who or what the findings are meant to apply to. “Women” is not a population. “Women members of self-help groups registered under the Tamil Nadu State Rural Livelihoods Mission in Villupuram and Cuddalore districts” is.
- Sampling frame: the actual list you drew from, and how complete it was.
- Sampling method: probability (simple random, stratified, cluster, systematic) or non-probability (convenience, purposive, snowball, quota); our sampling techniques guide covers how to choose and defend one. If you used a non-probability method, say so plainly and say what it means for generalisation. Do not describe a convenience sample as random.
- Sample size: the number, and how you arrived at it: a formula, a power analysis, a rule of thumb for your analysis method, or saturation in qualitative work. Our guide to sample size calculation covers Cochran’s formula, G*Power and the rules used for SEM.
- Response rate: how many you approached, how many responded, how many usable responses remained after screening.
For qualitative work, explain how participants were selected and why they could speak to your question, and describe how you decided you had enough interviews.
Instruments, reliability and validity
Describe every instrument: questionnaire, interview guide, test, observation schedule, laboratory equipment or dataset. For questionnaires, give the source of each scale, the number of items, the response format, and any changes you made. If you translated a scale into Tamil, Hindi or another language, say how (forward and back translation is the usual method) and who checked it. Put the full instrument in an appendix.
Pilot study
Report what the pilot tested, the sample size, and what you changed as a result. If the pilot led to no changes, say so and explain why.
Reliability and validity
Report evidence from your data, not just the original author’s. For multi-item scales this typically means internal consistency (Cronbach’s alpha, or composite reliability in SEM, with 0.70 as the commonly used threshold), and for SEM studies convergent and discriminant validity from the measurement model. Content validity is usually established by expert review of the questionnaire; say who the experts were in general terms and what they changed.
In qualitative research the equivalent concepts are credibility, dependability and transferability. Describe what you did: member checking, an audit trail, a second coder, thick description.
In experimental and engineering work, describe calibration, replication, controls, and the datasets and baselines used for comparison.
Methods of analysis, ethics and limitations
Methods of analysis
State which analysis answers which objective. A table like the one below is one of the clearest things you can put in the chapter, and examiners use it to check the whole thesis.
| Objective | Data | Analysis |
|---|---|---|
| 1. To assess the level of digital literacy among rural women SHG members | Survey, 20-item literacy scale | Descriptive statistics; comparison across districts (ANOVA) |
| 2. To examine the effect of digital literacy on use of mobile banking | Same survey, usage items | Logistic regression, controlling for age, education and income |
| 3. To understand barriers to adoption | Interviews with 25 members | Thematic analysis |
Name the software and version (SPSS 29, AMOS, SmartPLS 4, R, NVivo, ATLAS.ti, MATLAB, Python with named libraries). Explain why each test suits your data, including any assumptions you checked. If you are unsure which test fits, our guide to choosing a statistical test works through the decision. If your study tests hypotheses, list them here or refer back to where they are stated; our guide on writing a research hypothesis covers how to phrase them.
Ethics
Research with human participants needs approval from your institution’s ethics committee before data collection starts (see research ethics approval). In health research, the ICMR National Ethical Guidelines apply. Describe the approval (committee name, reference number, date), how informed consent was taken, how data were anonymised and stored, and any special protections for vulnerable participants. Committees and examiners notice when approval was obtained after data were collected.
Limitations
Every design has limits. State yours plainly: a non-probability sample, self-reported data, a single district, a cross-sectional design. Then say what you did to reduce each one, and how it affects what the findings can claim. A clear limitations section strengthens your position at the viva, because it shows you have already considered the points an examiner is likely to raise.
Common mistakes in the methodology chapter, and a final check
- Textbook background. Pages defining research, types of design and types of sampling, with no link to your study.
- Copying the synopsis. The synopsis described a plan in the future tense. The thesis must describe what you actually did, including what changed.
- Mismatched analysis. Using a parametric test on ordinal data without justification, or running regression on 60 responses with 15 predictors.
- Calling everything “random”. If you approached whoever was available, it was convenience sampling. Say so.
- Missing details. No dates of data collection, no response rate, no source for the scale.
Run through these points before handing the draft over:
- Each choice is justified by your research questions, not by what other scholars used or what software you know.
- Population, sampling frame, sampling method and sample size are all stated, with the calculation or rule behind the number.
- Every instrument has a source; adapted scales say what was changed and why.
- Reliability and validity evidence is reported for your data, not just quoted from the original author.
- Every objective maps to specific data and a specific analysis.
- Ethics approval, consent and data protection are described, with approval numbers where they exist.
- Limitations of the design are stated clearly, with what you did to reduce them.
- The chapter is written in the past tense and describes what you actually did, not what the synopsis proposed.
For where the methodology chapter sits in the thesis as a whole, see our guide to PhD thesis structure. If you want a methodologist in your field to review your design and analysis plan, that is what our research methodology consulting covers. The decisions stay yours, and you will need to defend them.
Sources
FAQ
Questions scholars ask
How long should the methodology chapter be?
Commonly 20 to 40 pages in a social science or management thesis, and often shorter in engineering, where methods may be spread across the technical chapters. Length should come from detail and justification, not from definitions.
Do I need a research philosophy section?
In management, education and many social sciences, examiners often expect a short one. In engineering, natural sciences and medicine it is rarely included. Look at recent theses from your department to see what your examiners are used to.
What if my method changed from what the synopsis proposed?
That is normal. Describe what you actually did and briefly explain why it changed. If the change was significant, make sure your research committee approved it and that the approval is recorded.
Is a convenience sample acceptable in a PhD?
It can be, especially where no sampling frame exists, but you must say it was convenience sampling and limit your claims to match. Examiners object to non-probability samples described as random, and to findings generalised to the whole population.
Should I write the methodology in past or future tense?
In the thesis, past tense: you are reporting what you did. In the synopsis, future tense, because the study has not yet been done.
