Types of research design: which one fits your research question

What are the main types of research design?

Treat it as three decisions. First, the family: quantitative, qualitative or mixed methods. That is about the kind of evidence, numbers you can count and test or words and observations you interpret. Second, the specific design inside the family; a descriptive survey and a randomised experiment are both quantitative and are completely different pieces of work. Third, time: one round of data collection (cross-sectional) or several (longitudinal).

You will also see research labelled by purpose. Exploratory research looks at a poorly understood problem to find out what the important variables are. Descriptive research reports what exists. Explanatory research asks why, by testing relationships or causes. These labels sit on top of the designs rather than replacing them. I’ve read synopsis drafts that call the same study “exploratory” on page 2 and “explanatory” on page 9. Pick one and mean it.

Qualitative vs quantitative research: which does your question need?

Listen to the question word in your research question. It usually settles the matter.

Quantitative research answers how much, how many, how strongly related and what effect. It needs variables measured the same way for everyone, a sample big enough for statistical tests, and, usually, hypotheses stated in advance.

Qualitative research answers how and why: meanings, processes, experiences. Samples are small and chosen on purpose, and the aim is depth.

Neither is more scientific. The common mistake is choosing the family first (because the department always does surveys, or because the scholar fears statistics) and bending the question to fit. If your objectives are still fuzzy, fix them first with our guide on research objectives and questions.

Descriptive vs experimental: the quantitative designs compared

The five quantitative designs differ in how much control you have over the cause, and so in how strong a causal claim you can make.

Descriptive

You measure things as they are. Awareness surveys and prevalence studies are descriptive, and this is the commonest design in Indian commerce and social science theses. It is fine when the question is descriptive. It stops being fine when chapter 5 starts talking about “impact”.

Correlational

You measure several variables in the same people and test how they relate, with correlation, regression or structural equation modelling. A strong link between social media use and anxiety does not tell you which came first, or whether exam pressure drives both.

Causal-comparative (ex post facto)

The groups differed before you arrived: rural and urban students, men and women, first- and second-generation entrepreneurs. You compare them on an outcome. Since you didn’t create the groups, other differences between them may explain the result.

Quasi-experimental

There is an intervention, but you can’t decide at random who gets it. A state scheme reaches some districts first; a college tries a new teaching method in two sections only. Shadish, Cook and Campbell’s Experimental and Quasi-Experimental Designs for Generalized Causal Inference (2002) is the standard reference.

Experimental

You manipulate the cause and assign participants at random, which is what lets you credit the difference to the treatment. In health research this is the randomised controlled trial. In India a trial must be registered with the Clinical Trials Registry – India before the first participant is enrolled; CTRI has required prospective registration for all studies since 1 April 2018.

Descriptive vs experimental, then: a descriptive study tells you what is there, and an experiment tells you what happens when you change something. For the test that goes with each design, see choosing a statistical test.

What are the types of qualitative research design?

Creswell and Poth’s Qualitative Inquiry and Research Design (5th edition, 2023) groups qualitative work into five approaches, and most examiners recognise the set.

Case study. One bounded case or a few, studied in depth from several sources. “Self-help groups in India” is not a case. “Three SHG federations in one block of Nalanda district” is.

Phenomenology. The lived experience of something the participants share, such as caring for a parent with dementia. Few participants, long interviews.

Grounded theory. Building a theory of a process from the data, an approach that goes back to Glaser and Strauss’s The Discovery of Grounded Theory (1967). Scholars often use the label for any interview study. Don’t, unless you are actually building a theory.

Ethnography. Long immersion in a group’s daily life. Slow and demanding; it suits anthropology and sociology.

Narrative research. The life stories of one or a few people, read for sequence and meaning.

Plenty of good Indian theses simply use interviews analysed thematically, without claiming any of the five. Say so honestly. Our thematic analysis guide covers that analysis, and semi-structured interviews covers collecting the data.

What are the mixed methods research designs?

Mixed methods means collecting quantitative and qualitative data and deliberately integrating them. Two open-ended questions at the end of a questionnaire do not qualify.

Creswell and Plano Clark’s Designing and Conducting Mixed Methods Research (4th edition, SAGE, 2025) keeps the three core designs of its earlier edition.

Mixed methods is roughly two studies’ worth of work, and you must be able to defend both analyses. I’d only choose it when one kind of data really can’t answer the question. The exploratory sequential design is common in management theses where no scale exists for the Indian setting; our questionnaire design guide helps with the second phase.

Cross-sectional vs longitudinal design: when does time matter?

A cross-sectional study collects data once, or over a short period. Levin, writing on cross-sectional studies in Evidence-Based Dentistry (2006), notes they are usually used to estimate how common something is. Most PhD surveys are cross-sectional.

A longitudinal study measures more than once. A panel follows the same people; a repeated cross-section draws a fresh sample each round. Only longitudinal data shows change.

Be realistic about fieldwork time. A three-wave panel is possible, but people drop out between waves, so start with a bigger sample (see our sample size guide). Existing panel data from national surveys is often the practical route.

How do you choose a research design for a PhD?

Run through this before the synopsis goes to your DC or RAC.

• Write the main research question in one sentence and underline the question word: how many, is there a relationship, does it cause, how, why.
• Decide the claim the thesis must make (description, association, cause or understanding) and pick the simplest design that supports it.
• Ask whether you control who gets the intervention. If not, a true experiment is off the table.
• Ask whether time matters to the answer. If it does, plan repeated measurement or say why one time point is enough.
• Check access and time: can you reach the people, records or sites within your registration period?
• Look up a few recent theses on Shodhganga in your area to see which designs examiners there have accepted.

A worked example

An illustrative topic: “the impact of smart classrooms on learning in Tamil Nadu government schools”.

“Impact” is a causal claim, so a survey of teachers’ opinions can’t support it, though that is what many first drafts propose. Randomly assigning schools is out of the scholar’s hands. The realistic design is quasi-experimental: compare test scores in schools with and without smart classrooms, before and after installation, if the records exist.

If they don’t, change the question. “How are teachers using smart classrooms, and what gets in the way?” is a sound qualitative or mixed methods thesis. Keeping “impact” in the title while collecting opinions is what gets picked apart at the viva.

With the design settled, move on to sampling techniques. If you can’t yet choose, the problem itself may be unclear; see the statement of the problem.