Research Methods

Introduction

Research Methods is a university-level aiMOOC about how researchers turn questions into defensible evidence. You will learn to move from a broad topic to a researchable question, choose an appropriate design, select participants or cases, define and measure concepts, collect and analyse data, evaluate the quality of evidence, address ethical responsibilities, and communicate findings transparently.

Research is not a single fixed recipe. Different disciplines ask different kinds of questions, work with different forms of evidence, and justify claims in different ways. A laboratory experiment, an ethnographic field study, a historical archive project, a survey, a case study, and a mixed-methods evaluation can all be rigorous when the methods fit the question and the reasoning is explicit. Throughout this course, you should therefore ask three recurring questions: What claim am I trying to support? What evidence would be relevant? What design would make the inference credible?

The diagram above presents one simplified cycle of observation, question, hypothesis, experiment, analysis, and conclusion. Real research is often less linear: literature review, theory, pilot work, data collection, analysis, and interpretation can feed back into one another. Treat the diagram as a model for disciplined inquiry rather than as a rule that every project must follow in exactly the same order.

Learning Goals

By the end of the aiMOOC, you should be able to explain the logic of research, formulate focused questions, distinguish major methodological approaches, justify a research design, select appropriate sampling strategies, operationalize concepts, evaluate reliability and validity, plan ethical data collection, analyse quantitative and qualitative evidence at an introductory university level, integrate evidence in mixed-methods research, and prepare a transparent research protocol.

From Topic to Research Question

A research project usually begins with a problem, puzzle, uncertainty, or gap in existing knowledge. A broad topic such as “student learning,” “urban heat,” or “workplace automation” is not yet a research question. A useful research question identifies what you want to understand, compare, explain, estimate, describe, interpret, or evaluate.

A strong research question is clear, researchable, appropriately scoped, and connected to evidence. It should be possible to explain what data would help answer it. Questions can be descriptive, comparative, relational, causal, interpretive, evaluative, or exploratory. For example, “How common is academic procrastination among first-year students?” is descriptive, while “How do first-year students describe the strategies they use to manage academic procrastination?” is interpretive.

A hypothesis is a specific proposition that can be examined with evidence. Hypotheses are common in confirmatory quantitative research, but they are not required in every project. Exploratory qualitative studies, historical inquiries, design research, and some forms of grounded theory may begin with guiding questions instead.

Literature Review and Conceptual Framework

A literature review does more than summarize previous publications. It maps what is known, identifies disagreements and gaps, clarifies concepts, compares methods, and helps you position your question. A good review records search terms, databases, inclusion decisions, and key sources so that the intellectual path to the research question is visible.

A conceptual framework explains the ideas and relationships that guide the study. It may draw on an established theory, a model, or a set of sensitizing concepts. In quantitative research, a framework can identify predictors, outcomes, mediators, moderators, and confounders. In qualitative research, it can guide attention while remaining open to unexpected meanings in the data.

Methodological Approaches

Quantitative Research

Quantitative research uses numerical data to describe distributions, estimate parameters, test hypotheses, compare groups, model relationships, or evaluate interventions. Common designs include experiments, quasi-experiments, surveys, cohort studies, case-control studies, and secondary analysis of existing datasets.

Quantitative reasoning depends on the quality of measurement and design. A precise statistical model cannot repair a badly defined construct, a biased sample, or a confounded comparison. Statistical uncertainty should be reported and interpreted in context rather than reduced to a single threshold.

Qualitative Research

Qualitative research investigates meanings, experiences, practices, interactions, texts, images, institutions, and social processes. Common approaches include ethnography, phenomenology, grounded theory, narrative inquiry, discourse analysis, case study research, and qualitative content analysis.

Qualitative researchers may collect data through interviews, focus groups, observations, documents, diaries, digital traces, or visual materials. Rigour involves careful sampling, transparent documentation, reflexivity, systematic analysis, attention to alternative interpretations, and a clear connection between evidence and claims.

An interview is not simply a conversation. Research interviews require a purpose, an ethically appropriate recruitment process, an interview guide or rationale, attentive listening, probing, accurate recording or note-taking, and an analysis plan. The relationship between interviewer and participant can shape what is said, so reflexivity is part of the method.

Mixed Methods

Mixed methods research intentionally combines quantitative and qualitative components in one program of inquiry. The key idea is integration: the two forms of evidence should inform each other rather than merely appear side by side. A study might use survey results to select interview participants, use interviews to explain an unexpected statistical pattern, or combine both strands in a joint interpretation.

Common mixed-methods structures include convergent designs, explanatory sequential designs, and exploratory sequential designs. The choice should follow the research question, timing, resources, and intended form of inference.

Research Design and Causal Reasoning

A research design is the overall logic connecting the question, evidence, analysis, and conclusion. It specifies what or whom you will study, when observations occur, what comparisons are made, and how alternative explanations will be addressed.

Experimental Designs

In a randomized controlled experiment, participants or units are assigned to conditions using a random mechanism. Random assignment aims to make treatment groups comparable on both observed and unobserved characteristics on average. This strengthens causal inference when implementation, measurement, attrition, and analysis are also handled carefully.

Not every question can or should be answered experimentally. Ethical constraints, feasibility, scale, rarity of outcomes, long time horizons, and the nature of the phenomenon may require observational or qualitative designs.

Observational and Quasi-Experimental Designs

Observational research measures exposures, characteristics, and outcomes without random assignment by the researcher. Examples include cross-sectional studies, longitudinal studies, cohort designs, case-control studies, and naturalistic observation.

A statistical association is not automatically a causal effect. A confounder is a variable related to both the exposure and outcome that can produce or distort an observed association. Quasi-experimental approaches such as interrupted time series, regression discontinuity, difference-in-differences, matching, or instrumental-variable designs can strengthen causal reasoning when their assumptions are plausible and clearly examined.

Case Studies and Comparative Designs

A Case study investigates a bounded case in depth, such as an organization, community, event, program, policy, or individual. Case selection should be justified: a case may be typical, extreme, critical, revelatory, or chosen for comparison. Comparative case designs can reveal similarities, differences, mechanisms, and contextual conditions that are difficult to see in a single case.

Sampling and Recruitment

A population is the larger set of units to which a study refers. A sample is the subset actually observed. Sampling decisions shape what you can infer.

In probability sampling, each unit has a known non-zero selection probability under the sampling design. Examples include simple random, systematic, stratified, and cluster sampling. Probability sampling supports design-based estimates of sampling uncertainty when the frame and implementation are appropriate.

In non-probability sampling, selection probabilities are not known. Convenience, purposive, quota, snowball, theoretical, and maximum-variation sampling can be useful, especially in qualitative research or research with hard-to-reach populations. The important task is to match the sampling strategy to the intended inference and to state its limitations.

Sampling Error, Bias, and Sample Size

Sampling error is the variation that arises because a sample rather than the whole population is observed. Sampling bias is systematic distortion caused by the way units enter or remain in the sample. Nonresponse, coverage problems, attrition, and self-selection can produce bias even in a large study.

Sample size planning should reflect the research goal. In quantitative studies, planning may consider expected effect sizes, desired precision, statistical power, design effects, and anticipated missing data. In qualitative studies, adequacy depends on the richness and relevance of information, heterogeneity of the sample, analytic approach, and the point at which additional data no longer materially change the analysis.

Sampling distributions help explain why estimates vary from sample to sample and why standard errors typically decrease as information increases. They are central to confidence intervals and many forms of statistical inference.

Concepts, Variables, and Measurement

A construct is an abstract concept such as trust, stress, social capital, political efficacy, or learning engagement. Operationalization turns a construct into observable indicators or procedures. The operational definition should make clear how the concept is represented in the study.

Variables can be categorical or numerical, and their measurement scale affects what summaries and analyses are meaningful. Researchers should distinguish between variables that describe outcomes, exposures or predictors, potential confounders, mediators, moderators, and control variables.

Reliability and Validity

Reliability concerns consistency of measurement. Examples include test-retest reliability, inter-rater reliability, and internal consistency. Reliability is necessary for many forms of measurement quality but does not guarantee that the intended construct is being measured.

Validity concerns whether evidence and reasoning support the intended interpretation or inference. Depending on the field, researchers discuss construct validity, internal validity, external validity, criterion validity, content validity, ecological validity, and other forms. Validity is not a permanent property of an instrument in isolation; it depends on how the measure is used and what claim is being made.

Pilot testing can reveal confusing items, technical failures, weak response categories, interviewer problems, timing issues, and unexpected participant interpretations before the main study begins.

Data Collection

Surveys and Questionnaires

A survey can gather standardized information from many respondents, but good survey design is demanding. Questions should use clear language, avoid double-barrelled wording, minimize unnecessary assumptions, and provide response options that fit the construct. Question order, mode of administration, device type, interviewer presence, and incentives can all affect responses.

Before launching a survey, test the instrument with people similar to the target population. Cognitive interviewing can help reveal how respondents interpret questions and choose answers.

Interviews and Focus Groups

Structured interviews use highly standardized questions. Semi-structured interviews combine prepared topics with flexible follow-up questions. Unstructured or in-depth interviews allow the participant's account to shape the direction more strongly.

Focus groups use interaction among participants as part of the data. They are useful for exploring shared norms, disagreements, language, and collective meaning-making. They are less suitable when confidentiality between participants cannot be reasonably protected or when the topic creates a high risk of coercion or harm.

Observation, Documents, and Digital Data

Observation can be participant or non-participant, structured or open-ended, overt or sometimes covert where ethically and legally permissible. Field notes should distinguish observation from interpretation as clearly as possible.

Documents and digital traces can be primary data. Researchers should ask who created the material, for what purpose, under what conditions, what is missing, and what legal or ethical expectations apply. Public accessibility does not automatically remove ethical responsibilities, especially when people could be identified or harmed.

Research Ethics and Integrity

Research ethics protects participants and supports trustworthy inquiry. For research involving humans, important principles include respect for persons, informed and voluntary consent, attention to risks and benefits, fair selection, privacy, confidentiality, and additional safeguards when people may be vulnerable to coercion or harm.

Informed consent is an ongoing process rather than only a signed form. Participants should receive understandable information about what participation involves, foreseeable risks, data use, withdrawal conditions, and relevant limits to confidentiality. Requirements differ across institutions, disciplines, countries, and types of research, so you must follow the rules that apply to your setting and obtain ethics review when required.

Research integrity also concerns fabrication, falsification, plagiarism, undisclosed conflicts of interest, inappropriate authorship, selective reporting, poor data stewardship, and misleading presentation. Transparent records of decisions, version control, secure data management, and clear authorship contributions reduce avoidable problems.

Data Protection and Confidentiality

Plan data protection before collection begins. Decide what personal data are necessary, how identifiers will be separated from research data, who will have access, how files will be secured, how long data will be retained, and what will happen at the end of the retention period.

Anonymization aims to prevent individuals from being identified from released data, but perfect anonymity can be difficult when datasets contain rich combinations of attributes. Pseudonymization replaces direct identifiers with codes but still permits re-identification through a separate key. Treat the distinction seriously.

Quantitative Data Analysis

Quantitative analysis usually begins with data checking and descriptive statistics. Examine missingness, impossible values, coding errors, distributions, and outliers before fitting complex models. Summaries such as counts, proportions, means, medians, standard deviations, quantiles, and visualizations should be chosen to suit the measurement scale and distribution.

Inferential statistics quantify uncertainty under a set of assumptions. Confidence intervals communicate a range of values compatible with the data and model. Hypothesis tests evaluate how surprising a statistic would be under a specified null model. A p-value is not the probability that the null hypothesis is true, and statistical significance is not the same as substantive importance.

Effect sizes and uncertainty are usually more informative than a threshold alone. Model assumptions, missing-data handling, multiple testing, sensitivity analyses, and researcher degrees of freedom should be reported.

A correlation describes the strength and direction of a statistical relationship under a chosen measure. It does not by itself establish causality. Before interpreting a coefficient, inspect the data visually and consider nonlinearity, outliers, subgroup patterns, measurement quality, and confounding.

Qualitative Data Analysis

Qualitative analysis is a systematic process of interpreting patterns of meaning. Depending on the methodology, you may code segments of text, compare cases, develop categories, identify themes, reconstruct narratives, analyse discourse, or build theory.

A transparent workflow can include familiarization with the material, memo writing, coding, comparison, category development, theme refinement, searching for disconfirming evidence, and linking interpretations back to the research question. Software can help organize data but does not perform the interpretive work for you.

Reflexivity means examining how your position, assumptions, relationships, decisions, and institutional context influence the research. A reflexive memo can record why codes changed, why a case was interpreted differently, how field relationships shaped access, or where your expectations were challenged.

Trustworthiness in Qualitative Research

Strategies for strengthening qualitative trustworthiness include prolonged engagement where appropriate, triangulation, negative-case analysis, thick description, transparent coding decisions, peer debriefing, audit trails, and careful use of participant feedback. No single technique guarantees quality; the strategy should fit the methodology and research claim.

Triangulation compares different sources, methods, investigators, theories, or data types to examine convergence and divergence. Agreement can strengthen an interpretation, while disagreement can reveal important complexity rather than simply being treated as an error.

Mixed-Methods Integration

Mixed methods should produce an integrated inference. Integration can occur during design, sampling, data collection, analysis, or interpretation. For example, quantitative results can identify unusual cases for follow-up interviews, while qualitative findings can help explain mechanisms behind an average treatment effect.

A joint display is a table, matrix, or visual structure that brings quantitative and qualitative findings together. Ask where the strands agree, where they differ, whether one explains the other, and whether the combination produces a new insight that neither strand could provide alone.

Open, Transparent, and Reproducible Research

Transparent research makes the path from question to conclusion inspectable. Useful practices include preregistering confirmatory hypotheses and analysis plans where appropriate, keeping version-controlled code, preserving metadata, documenting exclusions and transformations, sharing materials, reporting all planned outcomes, and making data available when ethical, legal, and consent conditions permit.

Open science is broader than data sharing. It includes practices such as open access, open data where appropriate, open code, open materials, open peer review in some settings, citizen science, and transparent workflows. Openness must be balanced with privacy, intellectual property, community agreements, security, and the rights of participants.

Reproducibility and replication are related but not identical. In many fields, reproducibility means obtaining the same computational results from the same data and analysis workflow, while replication means testing whether a finding recurs with new data or a new study. Terminology varies by discipline, so define the term you use.

Interpreting Evidence and Making Claims

A conclusion should not be stronger than the design allows. Descriptive data support descriptive claims. Associations support relational claims when measurement and analysis are sound. Causal claims require a credible identification strategy and plausible assumptions. Qualitative evidence can support rich claims about meaning, process, context, mechanisms, and variation when the analytic chain is transparent.

Generalization also takes different forms. Statistical generalization extends from a sample to a target population under sampling assumptions. Analytic generalization connects findings to theory or a conceptual explanation. Transferability asks whether detailed contextual knowledge allows readers to judge relevance to another setting.

Uncertainty is not a weakness to hide. Strong research distinguishes what is known, what is estimated, what is assumed, what remains ambiguous, and what further evidence would change the conclusion.

Reporting Research

A clear research report allows readers to understand what you asked, why it mattered, how you collected and analysed evidence, what you found, and how the evidence supports your interpretation. Report limitations as features of the inference, not as a ritual paragraph detached from the results.

Different designs have specialized reporting guidelines. Examples include CONSORT for randomized trials, STROBE for observational studies, PRISMA for systematic reviews, and COREQ for qualitative interviews and focus groups. Use the guideline that fits your design and discipline, and also follow institutional, journal, professional, and legal requirements.

Interactive Tasks

Quiz: Test Your Knowledge

Which feature most directly distinguishes a research question from a broad topic? (It specifies an inquiry that can be addressed with evidence) (!It contains as many concepts as possible) (!It guarantees a statistically significant result) (!It avoids all reference to prior research)

What is the central purpose of random assignment in an experiment? (To make treatment groups comparable on average) (!To guarantee a representative sample) (!To eliminate all measurement error) (!To ensure every participant receives treatment)

Which statement about correlation is correct? (It describes association but does not by itself establish causation) (!It proves that one variable causes the other) (!It can only be used with experimental data) (!It eliminates the influence of confounders)

What is operationalization? (The process of turning a construct into observable indicators or procedures) (!The process of selecting a journal for publication) (!The process of assigning participants to random groups) (!The process of removing all missing data)

Which sampling method gives units known selection probabilities under the design? (Probability sampling) (!Convenience sampling) (!Snowball sampling) (!Purposive sampling)

What does reliability primarily concern? (Consistency of measurement) (!The moral acceptability of the study) (!The size of the target population) (!The novelty of the research question)

What is a defining feature of mixed-methods research? (Integration of quantitative and qualitative evidence) (!Use of only numerical data) (!Avoidance of interpretation) (!Automatic use of random assignment)

Why is reflexivity important in qualitative research? (It examines how researcher positions and decisions shape the inquiry) (!It converts interviews into random samples) (!It guarantees participant anonymity) (!It replaces the need for systematic analysis)

Which statement best describes informed consent? (It is an ongoing process of voluntary and informed participation) (!It is only a signature collected after data analysis) (!It allows researchers to ignore confidentiality) (!It removes the need for risk assessment)

What is the best interpretation of a study limitation? (It identifies a boundary or uncertainty affecting the inference) (!It proves that the study has no value) (!It should be omitted when results are significant) (!It is the same as research misconduct)

Memory Game

Construct	Abstract concept represented through indicators or observations
Confounder	Variable related to both an exposure and an outcome that can distort an association
Reliability	Consistency of a measurement process
Reflexivity	Examination of how the researcher influences and is influenced by the inquiry
Triangulation	Comparison of multiple sources methods or perspectives to examine convergence and divergence
Preregistration	Time-stamped specification of planned questions hypotheses or analyses before examining relevant outcomes

Drag and Drop

Match the correct terms.	Topic
Random assignment	Allocation to conditions using a chance mechanism
Purposive sampling	Selection of information-rich cases because they fit the study purpose
Operational definition	Concrete rule for representing or measuring a concept
Thematic analysis	Systematic development and interpretation of patterns of meaning
Confidence interval	Interval estimate used to express statistical uncertainty

...

Crossword Puzzle

Validity	What term concerns whether evidence supports the intended interpretation or inference?
Sampling	What process selects cases or units from a wider population or field?
Consent	What voluntary agreement should follow understandable information about participation?
Triangulation	What strategy compares multiple sources methods or perspectives?
Reliability	What term describes consistency of measurement?
Reflexivity	What practice examines the researcher's role and influence in the inquiry?

LearningApps

Cloze Text

A focused study begins with a researchable

that can be connected to relevant evidence. A research

links that question to sampling, data collection, analysis, and inference. Turning an abstract construct into observable indicators is called

. A variable that can distort an exposure-outcome association because it is related to both is a

. Consistency of measurement is described as

. In qualitative inquiry, critical examination of the researcher's position and decisions is called

. Ethical participation depends on informed and voluntary

. Combining quantitative and qualitative strands so that they inform one another requires

. Transparent workflows can strengthen computational

. A conclusion should remain proportional to the strength of the

.

Open-Ended Tasks

Easy

Research Question Clinic: Choose a broad university-level topic, write three possible research questions, and explain which one is clearest, most researchable, and most appropriately scoped.
Measurement Map: Select an abstract construct such as trust, motivation, or belonging and create a one-page diagram showing at least three possible indicators and the limitations of each.
Sampling Snapshot: Find a published study in your field, identify its target population and sample, and write a short note explaining whether the sampling strategy fits the claim.
Ethics Scenario: Create a brief case in which a researcher faces a consent, privacy, or confidentiality problem and write a justified response.

Standard

Interview Mini-Study: Design a five-question semi-structured interview guide, conduct one practice interview with an informed volunteer, and write a reflexive memo about how your wording and follow-up questions influenced the conversation.
Survey Pilot: Draft a ten-item questionnaire on a non-sensitive topic, pilot it with several volunteers, collect feedback on ambiguity and response options, and revise the instrument.
Data Visualization Audit: Select a graph from an academic or public report, reconstruct the intended claim, and evaluate whether the axes, scales, labels, and graphical choices support a fair interpretation.
Methods Comparison Video: Produce a three-minute video comparing how a quantitative, qualitative, and mixed-methods study could investigate the same research problem.

Advanced

Mini Research Protocol: Write a complete protocol containing a research question, conceptual framework, design, sampling plan, measures or qualitative prompts, analysis strategy, ethics plan, and limitations.
Replication Blueprint: Choose a published empirical finding and design a replication or reproducibility plan that specifies what would be repeated, what would change, what materials are required, and what outcome would count as informative.
Mixed Methods Joint Display: Create a small synthetic dataset and a short set of interview excerpts on the same topic, analyse each strand, and build a joint display that integrates convergent and divergent findings.
Field Research Portfolio: Visit a relevant research setting such as a laboratory, archive, library, community organization, museum, field site, or research institute and produce a portfolio containing observations, an interview with permission, methodological notes, and a reflection on how setting shapes evidence.

Learning Assessment

Design Justification: Given a research question, compare at least two plausible designs and justify which one produces the most credible inference under realistic ethical and resource constraints.
Bias Diagnosis: Analyse a hypothetical study with nonresponse, attrition, and imperfect measurement, explain how each problem could affect the findings, and propose remedies that do not overstate what can be fixed.
Causal Reasoning: Draw a simple causal diagram for an exposure, outcome, and at least two plausible confounders, then explain what evidence would be needed to support a causal claim.
Qualitative Audit Trail: Analyse a short transcript, develop a coding scheme, revise it after comparison with a second case, and document how and why your interpretation changed.
Mixed Evidence Synthesis: Interpret a scenario in which survey results and interviews disagree, generate at least two explanations for the divergence, and propose a follow-up analysis or study.
Open Research Plan: Decide which materials, code, data, metadata, and analysis decisions from a hypothetical project could be shared, which should remain restricted, and justify the balance between transparency and participant protection.

Evidence of Learning

Your evidence of learning should show more than vocabulary recall. A strong portfolio demonstrates knowledge of research design, sampling, measurement, ethics, analysis, and inference; skills in formulating questions, evaluating evidence, designing instruments, interpreting data, documenting decisions, and communicating uncertainty; products such as a protocol, sampling plan, interview guide, questionnaire, coding framework, visualization, joint display, or reproducible analysis; and transfer through the ability to choose and defend appropriate methods for a new problem in your own discipline.

A high-quality portfolio also makes the limits of each product visible. You should be able to explain what the design can establish, what assumptions it relies on, what populations or contexts it addresses, what ethical constraints apply, and what further evidence would increase confidence.

OERs on the Topic

The English Wikipedia article on Research provides a broad starting point and links to related concepts such as scientific method, research design, qualitative research, quantitative research, statistics, and research ethics.

You can deepen your study through open materials on Scientific method, Research design, Statistics, Qualitative research, Survey methodology, Open science, Reproducibility, and Research ethics. When using any open resource, evaluate its authorship, revision history, references, licensing, and disciplinary fit.

Linked Learning Areas

Research Methods

Research methods connect the logic of inquiry with practical decisions about evidence. You begin by defining a researchable question and reviewing existing knowledge. You then choose a methodological approach and design, identify a population or field of cases, select a sampling strategy, operationalize concepts or define qualitative prompts, plan ethical data collection, and analyse evidence using methods appropriate to the data and question. Finally, you evaluate uncertainty, validity, trustworthiness, generalizability or transferability, document limitations, and communicate conclusions without claiming more than the evidence can support.