Research methods are the structured procedures researchers use to investigate questions, collect evidence, analyse data, and defend conclusions. The methodological landscape divides into three primary families: quantitative methods (experiments, surveys, secondary data analysis, econometrics, psychometrics, using statistical inference), qualitative methods (interviews, focus groups, ethnography, case studies, document analysis, discourse analysis, using interpretive analysis), and mixed methods (combining quantitative and qualitative strands under explicit integration design). Beyond the three families, four evidence-synthesis approaches operate across methodological lines: the systematic review (structured literature synthesis against predefined inclusion criteria), the meta-analysis (statistical pooling of effect sizes across studies), the scoping review (mapping the breadth of a research area), and the narrative review (traditional literature review without systematic inclusion protocol). This hub covers each family in depth: the logic of inference, the standard designs, the analytic techniques, the reporting standards, and the discipline-specific conventions that shape how research is conducted across health sciences, social sciences, business, education, humanities, law, and STEM.
Written by Naomi Alvarez, Lead Writing Expert (STEM and Engineering). Reviewed by Rohan Mehta, Lead Writing Expert (Health Sciences). Last reviewed 2026-04-24.
The Logic of Methodological Choice
Choosing a research method starts from the research question. A descriptive question (what is happening) leads toward survey research, observational studies, and document analysis. An explanatory question (why is this happening, what causes what) leads toward experiments, quasi-experiments, and strong observational designs with appropriate controls. An interpretive question (what does this mean to the people experiencing it) leads toward qualitative interviews, ethnography, phenomenology, and narrative inquiry. A predictive question (what will happen, what works) leads toward forecasting methods, machine learning models, and meta-analytic effect size estimation. A critical question (whose interests are served, what power dynamics are operating) leads toward discourse analysis, critical ethnography, and participatory action research.
The match between question and method is not always clean, and one hallmark of mature research design is recognising when the original question can only be answered by a combination of methods. A study of whether a clinical intervention reduces readmissions answers best with a randomised controlled trial for the causal question combined with qualitative interviews with patients for the experiential question. A study of whether a new educational policy improves learning answers best with longitudinal quantitative analysis combined with ethnographic observation of classroom implementation. The mixed methods frameworks discussed below formalise these combinations.
Quantitative Methods
Experimental Designs
The randomised controlled trial (RCT) is the gold standard experimental design. Participants are randomly assigned to treatment and control conditions; the random assignment balances known and unknown confounders in expectation, and the difference in outcome between groups is attributable to the treatment. The CONSORT statement (first published 1996, most recently updated as CONSORT 2025) defines the reporting standard for RCTs: the pre-registered trial protocol, the participant flow diagram, the primary and secondary outcomes, the statistical analysis plan, the intention-to-treat analysis, and the risk of bias assessment.
Extensions of the RCT include the cluster randomised trial (randomisation at the group level when individual randomisation is infeasible, with the intra-class correlation coefficient accounted for in sample size), the cross-over trial (each participant serves as their own control, with washout periods between conditions), the factorial design (testing multiple treatments simultaneously, including the interaction between them), the stepped wedge design (sequential rollout of an intervention with outcome measurement at each step), and the adaptive trial (trial parameters adjust based on interim analyses, increasingly common in oncology and device trials).
The quasi-experiment sits one step below the RCT in the evidence hierarchy. Random assignment is not feasible, but the design uses controls to approximate the counterfactual: the interrupted time series (outcome measurement before and after an intervention at known time points), the regression discontinuity design (treatment determined by a cutoff on a continuous variable, allowing causal inference on participants near the threshold), the difference-in-differences design (comparing outcome change in a treated group against outcome change in a comparison group), the synthetic control method (Abadie and co-authors, constructing a weighted composite comparison unit), and the instrumental variables design (using a variable that affects the treatment but not the outcome directly).
Survey Research
Survey research collects data through structured questionnaires administered to a sample of a population. The core design decisions are sampling (probability sampling via simple random, stratified, cluster, or multistage designs; non-probability sampling via quota, convenience, or snowball), mode (face-to-face, telephone, postal, web, mobile, with mixed-mode increasingly standard), questionnaire design (closed versus open items, scale construction, question order, skip patterns, cognitive interviewing for pretest), non-response management (follow-up waves, incentive calibration, weighting adjustments for unit non-response, imputation for item non-response), and measurement validity (content validity, construct validity, criterion validity, test-retest reliability, internal consistency via Cronbach's alpha or McDonald's omega).
Standard references include Dillman's Tailored Design Method, Groves and colleagues on total survey error, and the American Association for Public Opinion Research (AAPOR) standard definitions for response rate calculation. Major long-running surveys that supply secondary data for researchers include the General Social Survey (US), the European Social Survey, the British Household Panel Survey and Understanding Society, the Panel Study of Income Dynamics, the National Longitudinal Study of Adolescent Health (Add Health), the Health and Retirement Study, the National Health and Nutrition Examination Survey (NHANES), the Behavioral Risk Factor Surveillance System, the Demographic and Health Surveys, and the World Values Survey.
Observational and Secondary Data Analysis
Observational studies without treatment assignment remain the dominant design in epidemiology, sociology, political science, and economics. The standard designs are the cohort study (following a defined population forward in time, prospective or retrospective, with exposure and outcome measured, standard in epidemiology through studies like the Framingham Heart Study, the Nurses' Health Study, the Whitehall Study), the case-control study (identifying cases with an outcome and matched controls without, comparing exposure history), the cross-sectional study (exposure and outcome measured simultaneously, with limited causal inference), and the ecological study (group-level exposures and outcomes, vulnerable to the ecological fallacy).
The STROBE statement (Strengthening the Reporting of Observational Studies in Epidemiology) defines the reporting standard for observational work. Related extensions include STROBE-ME for molecular epidemiology, STROBE-Vet for veterinary epidemiology, and RECORD for research using routinely collected data.
Causal inference in observational data has been transformed by the potential outcomes framework (Rubin, Imbens) and the directed acyclic graph framework (Pearl). Standard methods include propensity score matching, inverse probability weighting, doubly robust estimators, targeted maximum likelihood estimation (TMLE), g-methods (g-formula, g-estimation, marginal structural models, Robins), and sensitivity analysis (E-value, Rosenbaum bounds).
Econometric and Quantitative Social Science Methods
Econometrics extends the quantitative toolkit with methods designed for the types of data and questions that arise in economics and adjacent disciplines. Panel data methods (fixed effects, random effects, first-differences) exploit variation over time within units. Time series methods (ARIMA, VAR, cointegration, Granger causality, state space models) model temporal dependence. Limited dependent variable models (logit, probit, multinomial logit, ordered probit, tobit, Heckman selection) handle categorical and censored outcomes. Structural models (demand estimation, dynamic discrete choice, auction models) estimate parameters of economic theory. Quantile regression (Koenker, Bassett) estimates the effect of predictors across the outcome distribution. Machine learning for causal inference (causal forests, double/debiased machine learning per Chernozhukov and co-authors, heterogeneous treatment effect estimation) combines flexible prediction with causal identification.
Psychometric Methods
Psychometrics develops and evaluates measurement instruments for unobserved constructs (intelligence, depression, customer satisfaction, organisational commitment). The standard approaches are classical test theory (Cronbach's alpha, split-half reliability, standard error of measurement), factor analysis (exploratory factor analysis, confirmatory factor analysis, principal components analysis), item response theory (Rasch models, two-parameter and three-parameter logistic models, graded response model, nominal response model), structural equation modelling (with software including Mplus, lavaan, Amos, LISREL, EQS), generalizability theory (Cronbach, Gleser, Nanda, Rajaratnam, decomposing variance across facets), and differential item functioning analysis (Mantel-Haenszel, SIBTEST, logistic regression, item response theory DIF).
Qualitative Methods
Interviews and Focus Groups
Semi-structured interviews are the workhorse qualitative data collection method. The researcher prepares an interview guide with open-ended questions but follows the participant's lead to explore emerging themes. Interview design decisions include depth versus breadth (typical studies run 60 to 90 minute interviews with 12 to 30 participants, with theoretical saturation achieved when additional interviews stop producing new themes), recording and transcription (verbatim transcription with speaker identification, denaturalised or naturalised depending on analytic aim, increasingly supported by AI transcription tools with human review for accuracy), sampling strategy (purposive, maximum variation, extreme case, snowball, theoretical), and reflexivity practice (field notes, interviewer journal, memo writing during analysis).
Focus groups collect data through moderated group conversation. The method is well suited for exploring shared meanings, normative expectations, and socially constructed understandings. Standard focus groups run 60 to 90 minutes with 6 to 10 participants, with the moderator attending to group dynamics alongside content.
Ethnography and Participant Observation
Ethnography immerses the researcher in a setting for extended observation and engagement. The method has deep roots in anthropology (Malinowski, Geertz, Clifford) and sociology (Chicago School, Whyte's Street Corner Society, Goffman, Becker) and now operates across disciplines from education to healthcare to organisational studies. Core practices include participant observation (the researcher participates in the setting while observing), field notes (descriptive, reflective, methodological, analytic memos), emic and etic perspectives (insider and outsider meanings), thick description (Geertz, contextually rich accounts rather than thin behavioural reports), and rigour practices (prolonged engagement, triangulation, member checking, negative case analysis, reflexivity, audit trail).
Contemporary variants include autoethnography (researcher's own experience as data, Ellis and Bochner), institutional ethnography (Dorothy Smith's approach tracing ruling relations through textually mediated practice), digital ethnography (online communities, social media, virtual environments), and rapid ethnography (intensive short-term field engagement, increasingly common in healthcare improvement and product design research).
Case Study Research
The case study investigates a bounded phenomenon in its real-world context. Robert Yin's Case Study Research: Design and Methods (now in its sixth edition) is the standard methodological reference, with Robert Stake's The Art of Case Study Research offering a more interpretive complementary framing. Design decisions include single versus multiple case design, holistic versus embedded design, case selection rationale (critical, extreme, revelatory, longitudinal), data source triangulation, chain of evidence (linking raw data to findings through documented analytic steps), and the case study protocol (an explicit document that governs data collection and analysis across cases). The case study format hub covers assignment writing for case studies across disciplines.
Phenomenology
Phenomenology investigates the lived experience of a phenomenon as it appears to consciousness. The transcendental phenomenology of Husserl and the interpretive phenomenology of Heidegger, Gadamer, and Merleau-Ponty have produced two main methodological streams: descriptive phenomenology (Giorgi, Colaizzi, Moustakas, with the researcher bracketing preconceptions to describe the essence of the experience) and interpretive phenomenology or hermeneutic phenomenology (van Manen, Smith's Interpretative Phenomenological Analysis, with the researcher's own interpretive horizon acknowledged). The method is prominent in nursing research, counselling psychology, educational research, and health sciences.
Grounded Theory
Grounded theory generates theory inductively from data. The original Glaser and Strauss formulation (1967) and the subsequent branches (Strauss and Corbin, Charmaz's constructivist grounded theory, Clarke's situational analysis) share core practices: theoretical sampling (data collection guided by emerging theoretical categories), constant comparison (new data compared against existing codes), open, axial, and selective coding, memo writing throughout analysis, and theoretical saturation as the stopping rule.
Narrative Inquiry and Life History
Narrative inquiry treats story as the unit of analysis. The method has roots in Bruner's narrative psychology, Riessman's work on interview narratives, and Clandinin and Connelly's narrative inquiry in education. Core practices include narrative interviews (open prompts that invite extended storytelling), restorying (representing participants' accounts in chronological or thematic structure), three-dimensional narrative analysis (temporality, sociality, place per Clandinin), and critical reflection on the researcher's own narrative positioning.
Discourse and Content Analysis
Discourse analysis examines language as social action. The main traditions include conversation analysis (Sacks, Schegloff, Jefferson, with detailed transcription of interactional sequences), critical discourse analysis (Fairclough, Wodak, van Dijk, linking linguistic choices to power relations), and discursive psychology (Edwards, Potter, examining psychological categories as discursive constructions). Content analysis operates on larger text corpora with coded categories; the Krippendorff and Neuendorf textbooks define the method, and contemporary practice often combines human coding with computational text analysis (topic modelling via latent Dirichlet allocation, word embeddings via word2vec and GloVe, large language model classification).
Qualitative Analysis Software
Qualitative data analysis is supported by software including NVivo (QSR International), ATLAS.ti, MAXQDA, Dedoose, and Quirkos for coded-data analysis, and Transana or F4 for detailed transcript work. The software supports coding, memoing, retrieval, matrix analysis, and visualisation but does not conduct the analysis itself; the analytic judgment remains the researcher's work.
Mixed Methods Research
Design Typology
Mixed methods research integrates quantitative and qualitative strands within a single study, under an explicit integration design. The standard reference is Creswell and Plano Clark's Designing and Conducting Mixed Methods Research (now in its third edition), which distinguishes three core designs. The convergent parallel design runs quantitative and qualitative strands simultaneously with independent collection and analysis, then integrates at the interpretation stage to triangulate findings. The explanatory sequential design runs the quantitative strand first, with qualitative work following to explain unexpected quantitative findings. The exploratory sequential design runs the qualitative strand first, with quantitative work following to test hypotheses generated qualitatively.
More complex designs include the intervention mixed methods design (qualitative strands embedded before, during, and after a quantitative trial, increasingly standard in implementation science), the case study mixed methods design (multiple data sources integrated within a single case), the participatory mixed methods design (community stakeholders involved in each methodological choice), and the multilevel mixed methods design (different methods at different levels of analysis, such as quantitative at the system level and qualitative at the individual level).
Integration Practices
Integration is the operational hallmark of mixed methods work. Integration happens at the design level (the research questions are formulated to require both strands), at the methods level (sampling and procedures are coordinated across strands), at the analysis level (joint displays that arrange quantitative and qualitative findings side by side, meta-inferences that combine findings into conclusions neither strand could reach alone), and at the reporting level (integrated narratives rather than strands presented separately with no connection).
The Good Reporting of a Mixed Methods Study (GRAMMS) criteria and the Mixed Methods Appraisal Tool (MMAT) define reporting and quality appraisal standards.
Evidence Synthesis: Systematic Review and Meta-Analysis
Systematic Review
A systematic review answers a defined research question by identifying, appraising, and synthesising all the research evidence that meets explicit inclusion criteria. The workflow is codified in the PRISMA 2020 statement (Preferred Reporting Items for Systematic Reviews and Meta-Analyses), with the extended PRISMA family covering specialised contexts (PRISMA-NMA for network meta-analyses, PRISMA-IPD for individual participant data reviews, PRISMA-DTA for diagnostic test accuracy reviews, PRISMA-ScR for scoping reviews, PRISMA-Children for paediatric reviews, PRISMA-Equity for equity-focused reviews).
Standard workflow stages include protocol registration (PROSPERO for most health and social reviews, the Open Science Framework for broader fields), search strategy (structured queries across MEDLINE, Embase, PsycINFO, CINAHL, Cochrane Central Register of Controlled Trials, Web of Science, Scopus, ERIC, ProQuest, grey literature sources), screening (title and abstract screening, full-text screening, dual independent review with conflict resolution by a third reviewer or discussion), data extraction (standardised forms, dual extraction, disagreement resolution), risk of bias assessment (RoB 2 for randomised trials, ROBINS-I for non-randomised intervention studies, QUADAS-2 for diagnostic accuracy studies, Newcastle-Ottawa for cohort and case-control studies, CASP for qualitative studies, JBI critical appraisal tools for a range of designs), synthesis (narrative synthesis following the Synthesis Without Meta-analysis (SWiM) guideline, meta-analysis where appropriate), and certainty of evidence assessment (GRADE framework for intervention reviews, CERQual for qualitative reviews, GRADE-CERQual integration for mixed methods reviews).
Dominant software in the field includes Covidence, Rayyan, and DistillerSR for screening; EndNote, Zotero, and Mendeley for reference management; and RevMan (Cochrane), Stata metan, R metafor and meta packages, and Comprehensive Meta-Analysis for quantitative synthesis.
Meta-Analysis
Meta-analysis statistically pools effect sizes across studies to produce a summary estimate. Core design decisions include effect size metric (standardised mean difference via Cohen's d or Hedges' g for continuous outcomes, odds ratio or risk ratio for binary outcomes, correlation coefficient or Fisher's z for correlation, hazard ratio for survival outcomes), model choice (fixed effects assuming a single true effect versus random effects assuming a distribution of true effects, with random effects now the default for most real-world research synthesis), heterogeneity assessment (Q statistic, I-squared statistic, tau-squared, prediction intervals), subgroup analysis and meta-regression (exploring sources of heterogeneity with pre-specified moderators), sensitivity analysis (leave-one-out analysis, alternative model assumptions, quality-weighted analysis), and publication bias assessment (funnel plots, Egger's test, trim and fill, PET-PEESE, selection models, p-curve, multi-level meta-analysis).
Advanced variants include network meta-analysis (comparing multiple interventions through direct and indirect evidence, NICE Decision Support Unit guidance, BUGS and R netmeta software), individual participant data meta-analysis (pooling raw data across studies with harmonised analysis, increasingly favoured where feasible), prospective meta-analysis (synthesising studies agreed in advance, coordinated through collaborative networks), and living systematic review and meta-analysis (continuous updating as new evidence appears, now the standard approach in fast-moving fields like COVID-19 research).
Scoping and Narrative Reviews
The scoping review (Arksey and O'Malley framework, extended by Levac and colleagues, now codified by the JBI scoping review methodology and PRISMA-ScR reporting standard) maps the breadth and nature of evidence on a topic without the narrowly focused question a systematic review requires. Scoping reviews are well suited to characterising a research field, identifying gaps, and informing systematic review feasibility.
The narrative review (traditional literature review) synthesises literature without explicit inclusion criteria or systematic search. The format remains common in graduate coursework and editorial context but carries higher risk of bias than systematic approaches. The literature review format hub covers the writing conventions for both systematic and narrative literature reviews.
Action Research and Practice-Based Inquiry
Action research engages stakeholders in cyclical inquiry that produces both practical action and research knowledge. Kurt Lewin's original 1940s formulation has evolved into distinct traditions: practitioner action research (teachers, nurses, social workers inquiring into their own practice, McNiff, Whitehead), participatory action research (community researchers and academic researchers in collaboration, Fals Borda, Freire), cooperative inquiry (Reason and Heron, groups as co-researchers), appreciative inquiry (Cooperrider and colleagues, focused on strengths and possibilities), and improvement science (Plan-Do-Study-Act cycles, increasingly standard in healthcare and education, see the educational leadership pillar for the education dialect).
Research Ethics and Integrity
All research with human participants operates under institutional ethics review. The governing frameworks include the Declaration of Helsinki (World Medical Association, most recently revised 2024), the Belmont Report (US National Commission, 1979, establishing respect for persons, beneficence, justice), the Council for International Organizations of Medical Sciences (CIOMS) International Ethical Guidelines, the International Council for Harmonisation Good Clinical Practice (ICH-GCP) for clinical research, the UK Research Integrity Office (UKRIO) Code of Practice, the Australian Code for the Responsible Conduct of Research, and institutional review board (IRB) or research ethics committee (REC) approval at the researcher's own institution.
Research integrity commitments include pre-registration of hypotheses and analysis plans (OSF, ClinicalTrials.gov, AsPredicted, AEA RCT Registry), data and code availability (TOP Guidelines, FAIR data principles, FAIR4RS for software), conflict of interest disclosure, authorship criteria (ICMJE authorship standards for biomedical work, with analogues in other fields), and misconduct prevention (plagiarism, fabrication, falsification per US Office of Research Integrity definitions).
Discipline-Specific Method Landscapes
Health Sciences
Health science research is methodologically eclectic but leans quantitative: RCTs dominate intervention evidence, cohort and case-control studies dominate epidemiological evidence, psychometric development supports patient-reported outcome measurement, and qualitative methods support behavioural and experiential research. See the nursing coursework support, pharmacy, public health coursework support, dental, nutrition, occupational therapy, physical therapy, physician assistant, speech-language pathology, radiology, anatomy and physiology, and veterinary pillars for sub-discipline method conventions.
Social Sciences and Business
Social science and business research splits across quantitative and qualitative traditions. Psychology leans experimental and psychometric; sociology and anthropology lean qualitative and interpretive; economics and political science lean econometric; education research straddles quantitative and qualitative with increasing mixed methods work; management and marketing span quasi-experimental, survey, case study, and ethnographic approaches. See the psychology paper assistance, sociology essay examples, political science paper assistance, economics writing services, business writing services, marketing, finance, accounting, advanced communications coursework support, education, educational leadership, criminal justice, and social work pillars.
Humanities
Humanities research uses textual analysis, archival work, comparative analysis, close reading, historiographic analysis, and critical theory rather than the hypothesis-testing framework of the sciences. Digital humanities has introduced computational text analysis (topic modelling, stylometry, computational literary analysis via Stanford NLP, HathiTrust, CLARIN) alongside traditional interpretive methods. See the english literature writing services, history essay examples, philosophy homework help, theology, linguistics, and arts and media pillars.
STEM and Engineering
STEM research relies on experimental measurement, computational modelling, simulation, and observational data analysis. Engineering adds design-build-test iterative methods and human factors research. See the math, statistics coursework support, physics academic resources, chemistry writing services, biology essay examples, sciences, programming coursework support, data science, cybersecurity, engineering, aerospace engineering writing guide, biomedical engineering coursework support, chemical engineering coursework support, environmental engineering, industrial engineering coursework support, nuclear engineering, and petroleum engineering pillars.
Law
Legal research uses doctrinal analysis (case law synthesis, statutory interpretation, constitutional analysis), comparative legal analysis (across jurisdictions, across traditions), empirical legal studies (quantitative analysis of legal phenomena, increasingly influential), and socio-legal research (qualitative studies of legal institutions in practice). See the law pillar essay help.
How to Write a Research Methods Chapter
- Clarify the research question and the epistemological framing. State the ontological and epistemological commitments that justify the methodological choice. A realist framing motivates different methods than a constructivist framing.
- Justify the methodological choice. Explain why the chosen method answers the research question better than alternatives. Cite methodological literature that grounds the choice.
- Describe the design in detail. For quantitative work, specify the sampling frame, sample size calculation, procedure, measures, and analytic plan. For qualitative work, specify the approach (ethnography, grounded theory, phenomenology, case study, narrative), the sampling strategy, the data collection procedure, and the analytic approach.
- Describe the analytic procedure. For quantitative work, specify the statistical model and the assumptions being tested. For qualitative work, describe the coding approach and the analytic software (if any).
- Address rigour and validity. For quantitative work, discuss internal validity threats and how the design addresses them. For qualitative work, address credibility, dependability, confirmability, and transferability (Lincoln and Guba, Morrow, Tracy).
- Document ethics approval and research integrity practices. Name the IRB or REC, the approval reference, the consent procedure, pre-registration details, and the data and code availability plan.