Research methods are the techniques researchers use to collect and analyze data. The main types are experiments, quasi-experiments, surveys, observational studies such as cohort and case-control studies, interviews, focus groups, observation and ethnography, case studies, action research, content and document analysis, secondary data analysis, and systematic reviews. They are grouped as quantitative, qualitative or mixed, and as primary (you collect the data) or secondary (someone else did).
Below, one question answered by two methods, then the twelve methods in one table and each with what it is, an example and the thing that most often goes wrong. Every example study is invented. The last sections cover how to choose a method and how a method differs from a methodology and a design.
One question, two methods
Many questions have parts that different methods answer. Take this one, from an invented college: why do students withdraw from first-year online courses at Lakeside Community College, and which students are most likely to? Records can answer the second half; only the students can answer the first.
| Analysis of enrollment records | Interviews with students who withdrew | |
|---|---|---|
| Answers | Analysis of enrollment recordsWhich students withdraw, from which courses, and in which week of the term | Interviews with students who withdrewWhy students decided to leave and what might have kept them |
| Method | Analysis of enrollment recordsSecondary data analysis, with statistics such as logistic regression | Interviews with students who withdrewSemi-structured interviews, analyzed thematically |
| Data type | Analysis of enrollment recordsQuantitative: five years of records the college already holds | Interviews with students who withdrewQualitative: transcripts of what 15 students said |
| Sample | Analysis of enrollment recordsEvery enrolled student, so no sampling within the college | Interviews with students who withdrewChosen on purpose from those who left and agreed to talk |
| Can't show | Analysis of enrollment recordsWhy anyone left, or whether the pattern holds at other colleges | Interviews with students who withdrewHow common each reason is; those who agreed may differ from those who didn't |
The table separates three things that course lists often blur. The method is the technique: records analysis, interviews. The data type is what each technique collects: numbers or words. The design is the plan that puts the methods together to answer the whole question.
Research methods at a glance
| Method | Strictly a | Data | Best for questions like |
|---|---|---|---|
| MethodExperiment | Strictly aDesign | DataUsually quantitative | Best for questions likeDoes changing X change Y? |
| MethodQuasi-experiment | Strictly aDesign | DataUsually quantitative | Best for questions likeDid a policy or program change Y, when you couldn't assign it? |
| MethodSurvey | Strictly aCollection method; a cross-sectional survey is also a design | DataMostly quantitative | Best for questions likeHow common is X, and who reports it? |
| MethodCohort, case-control, cross-sectional | Strictly aDesigns | DataQuantitative | Best for questions likeIs an exposure associated with an outcome? |
| MethodInterviews | Strictly aCollection method | DataQualitative | Best for questions likeHow do people experience or explain X? |
| MethodFocus groups | Strictly aCollection method | DataQualitative | Best for questions likeHow does a group talk about X? |
| MethodObservation and ethnography | Strictly aCollection method; ethnography is also a design | DataQualitative, sometimes counted | Best for questions likeWhat do people actually do, day to day? |
| MethodCase study | Strictly aDesign | DataMixed | Best for questions likeHow and why did X happen in this setting? |
| MethodAction research | Strictly aDesign | DataMixed | Best for questions likeHow can we improve our own practice? |
| MethodContent and document analysis | Strictly aAnalysis method | DataEither | Best for questions likeHow is X represented in texts, media or records? |
| MethodSecondary data analysis | Strictly aAnalysis of existing data | DataMostly quantitative | Best for questions likeWhat can existing data say about a new question? |
| MethodSystematic review, meta-analysis | Strictly aReview design; meta-analysis is an analysis method | DataEither; meta-analysis is quantitative | Best for questions likeWhat does all the research on X show? |
How methods are classified
The same method can be described along several lines at once:
- By data: numbers, words and observations, or both, which is the subject of qualitative vs quantitative research.
- By source: data you collect yourself, or records and datasets made by others, as in primary vs secondary sources.
- By purpose: exploring a new topic, describing how things are, explaining why, or evaluating a program. The research question examples show each type.
- By time: one snapshot (cross-sectional) or the same people followed over time (longitudinal).
- By control: whether the researcher intervenes, as in an experiment, or only observes.
Experiments and quasi-experiments
Experiments
An experiment changes one thing on purpose and measures what happens, holding everything else as steady as possible. In a randomized controlled trial, chance decides who gets which condition: the National Library of Medicine's MeSH vocabulary defines it by a test and a control treatment, enrolled and followed at the same time, assigned “by a random process”. Field experiments and A/B tests apply the same logic outside the lab.
Example: 400 students are randomly assigned to read an article on paper or on screen, then take the same quiz. Watch for: settings so artificial that the result doesn't carry over, and questions you can't ethically randomize.
Quasi-experiments
A quasi-experiment compares groups the researcher didn't assign: a school that adopted a new timetable and one that didn't, or one city before and after a new law. Natural experiments, where an event splits people almost at random, belong here.
Example: emergency visits for asthma in two similar cities, one of which banned wood-burning stoves. Watch for: differences between the groups, other than the change, that could explain the result.
Surveys and observational studies
Surveys and questionnaires
A survey asks a sample the same questions so the answers can be counted and compared. Closed questions give numbers, and open ones give text for qualitative analysis. A Delphi study is a survey of experts run in rounds. RAND developed the method in the 1950s to forecast the effect of technology on warfare: panelists answer anonymously, see the group's response and revise, aiming for consensus.
Example: an online questionnaire to 1,500 freelance translators about their rates and hours. Watch for: who doesn't answer, leading wording, and people reporting what they think they do rather than what they do.
Cohort, case-control and cross-sectional studies
These observational designs measure without intervening, mostly in health and the social sciences. MeSH defines them this way: a cross-sectional study measures a population or sample “at one particular time”; a cohort study identifies groups within a defined population, exposed to a factor or not, and follows them over time; a case-control study starts from people who have an outcome and compares them with a control group who don't.
Example: following 3,000 nurses for ten years to see whether night shifts predict later sleep problems. Watch for: confounding. Without random assignment, an association isn't proof of cause.
Interviews, focus groups and fieldwork
Interviews
Interviews collect people's accounts in their own words. Structured interviews ask fixed questions in a fixed order, semi-structured ones follow a guide but pursue what comes up, and unstructured ones are close to a guided conversation.
Example: 20 semi-structured interviews with hospital pharmacists about how they handle drug shortages. Watch for: leading questions, and treating what people say as a record of what they did.
Focus groups
A focus group brings a small group together to discuss a topic, and the interaction is part of the data. MeSH describes it as a method of data collection and a qualitative research tool in which a small group discusses its views together.
Example: four groups of rural parents discussing what would make them use a telehealth clinic. Watch for: one voice steering the room, and topics too private to discuss in front of others.
Observation and ethnography
Observation records what people do rather than what they say, either from outside or as a participant. Structured observation counts behaviors on a checklist and can be quantitative. Ethnography is long-term fieldwork inside a group, combining participant observation, conversations and documents.
Example: a year working shifts in a restaurant kitchen to study how cooks learn the trade. Watch for: people changing their behavior because they are watched, and the consent rules for observing them.
Case studies and action research
Case studies
A case study examines one case, or a few, in depth and in context: a company, a classroom, a policy, an event. It usually combines several sources, such as interviews, documents and figures.
Example: how one library system redesigned its services after a budget cut, from council minutes, staff interviews and usage data. Watch for: choosing a case because it is convenient rather than informative, and claiming it represents more than it does.
Action research
In action research, practitioners study a problem in their own setting while trying to fix it, in cycles of planning, acting, observing and reflecting. It is common in education, nursing and community work.
Example: three teachers test a peer-feedback routine for two terms, adjusting it after each cycle. Watch for: your own stake in the result, so record each cycle as it happens, not afterward.
Methods that use existing material
Content and document analysis
Content analysis codes texts, images or broadcasts systematically: counting the codes makes it quantitative, interpreting them makes it qualitative. Documentary and archival research reads records such as letters, minutes and case files as evidence, and discourse analysis studies how the language itself frames a subject.
Example: coding 800 news stories about heat waves for whether they mention climate change. Watch for: what the archive kept and what it lost, and codes two people would apply differently.
Secondary data analysis
Secondary analysis answers a new question with data someone else collected: a census, a national survey, court records, a company's sales history.
Example: testing whether commuting time predicts volunteering, using a national household survey. Watch for: variables that measure something slightly different from what you need, and thin documentation on how the data were collected.
Systematic reviews and meta-analyses
A systematic review answers a question from existing studies by a set, reported procedure. The Cochrane Handbook describes it as an attempt to collate all the empirical evidence that fits criteria set in advance, to answer a specific question. A meta-analysis combines the results of independent studies statistically. The PRISMA 2020 statement, short for Preferred Reporting Items for Systematic reviews and Meta-Analyses, sets out what to report, with a checklist and a flow diagram.
Example: pooling 25 trials of reading programs for children with dyslexia to estimate an average effect. Watch for: unpublished null results missing from the pool, and studies too different to combine. A thesis chapter's narrative review is a different job, covered in the literature review template.
Method, methodology and design
These words get mixed up in proposals, and examiners notice:
| Term | What it means | Example |
|---|---|---|
| TermMethod | What it meansA technique for collecting or analyzing data | ExampleSemi-structured interviews; logistic regression; thematic analysis |
| TermDesign | What it meansThe plan that arranges methods to answer the question | ExampleA randomized trial; a cross-sectional survey; a multiple-case study |
| TermMethodology | What it meansThe argument for why this design and these methods fit | ExampleThe reasoning in a dissertation's methodology chapter |
| TermResearch philosophy | What it meansAssumptions about reality and knowledge underneath | ExamplePositivism, interpretivism, pragmatism |
The method section of an IMRaD paper reports the first two; a dissertation's methodology chapter also defends them, as the methodology examples show. The common slips:
- Half a method. “Interviews” names collection only. “Twenty semi-structured interviews, analyzed thematically” names both halves.
- Calling any comparison an experiment. Without assignment by the researcher, it is quasi-experimental or observational.
- “Case study” as a label for a small sample. A case study is defined by its bounded case and context, not by its size.
- Method first, question second. If you chose interviews before you had a question, check that the question really asks about experience.
How to choose a research method
- Name what the question asks for
An amount, an association, a cause, an experience, a process, a representation or a summary of evidence. Each points to different methods in the table below.
- Check access, time and ethics
Can you reach the people or records you need, within your deadline, with the approval your institution requires? If not, change the method or narrow the question.
- Pilot before you commit
Run three interviews, send the survey to ten people, or code twenty documents. Problems found now cost an afternoon instead of a semester.
| If your question asks | Methods that fit |
|---|---|
| If your question asksHow common is X? | Methods that fitSurvey, cross-sectional study, secondary data analysis |
| If your question asksDoes X cause Y? | Methods that fitExperiment; quasi-experiment when you can't randomize |
| If your question asksDoes exposure to X predict Y later? | Methods that fitCohort study |
| If your question asksWhat led to a rare outcome? | Methods that fitCase-control study |
| If your question asksHow do people experience X? | Methods that fitInterviews, focus groups |
| If your question asksHow does a practice work day to day? | Methods that fitObservation, ethnography |
| If your question asksHow and why did this happen here? | Methods that fitCase study |
| If your question asksHow is X represented? | Methods that fitContent, discourse or archival analysis |
| If your question asksWhat does the research already show? | Methods that fitSystematic review, meta-analysis |
| If your question asksHow can we improve what we do? | Methods that fitAction research |
| If your question asksWhat do experts agree on? | Methods that fitDelphi study |
Many questions need two rows, which is where mixed methods come in. Once the method is settled, the research proposal example shows how to present it to a committee.
With a draft methods section, you can attach it and your research questions to a chat with the Strokes Assistant and ask it to review whether each question has a method that can answer it.
Frequently asked questions
What is the difference between research methods and methodology?
Methods are the techniques you use to collect and analyze data, such as a survey or thematic analysis. Methodology is the reasoning behind choosing them: why this design and these methods can answer your question. A journal article reports its methods briefly, while a dissertation's methodology chapter argues for them.
What are primary and secondary research methods?
Primary research methods collect new data yourself, through experiments, surveys, interviews, focus groups or observation. Secondary research methods work with data or studies that already exist, such as census data, archives, published datasets, or the studies pooled in a systematic review. Many projects combine the two.
Is a literature review a research method?
A systematic review is: it follows a set, reported procedure to find, select and appraise studies, and a meta-analysis adds a statistical summary. The literature review chapter of a thesis usually isn't a method in that sense. It surveys what is known to set up your own study, and its job is to show the gap.
Which research method is best for a dissertation?
The one that can answer your research question with the access, time, skills and ethics approval you have. A question about how common something is points to a survey or existing data; a question about experience points to interviews. Ask your supervisor which methods your department expects to see justified.
