Methods

Method directory

• Case study

  A case study investigates one person, a small set of people or another clearly bounded case in depth. Multiple kinds of evidence can establish a detailed account rather than reducing the case to one score. Rare or unusual cases can challenge a general theory, but a striking case does not tell us how common a pattern is.

  What can detailed evidence from one person with an unusual memory impairment reveal about memory?

  Research designSee an example

• Confidence interval

  A frequentist confidence interval is produced by a procedure designed to cover the fixed population parameter at a stated long-run rate under its assumptions. For a 95% procedure, about 95% of intervals from repeated applications would contain that parameter. The interpretation concerns the procedure, not a probability assigned to the parameter after observing this interval.

  What does the interval show about estimation precision under the chosen procedure and model?

  AnalysisSee an example

• Construct validity

  Construct validity concerns whether evidence supports interpreting a score as a measure of the proposed psychological construct. It involves a reasoned network of expectations about content, structure, relationships to other measures, and relevant group differences. An attractive label or a reliable score is insufficient.

  Do observations support the interpretation we want to make from this score?

  MeasurementSee an example

• Correlational study

  Researchers measure variables as they occur and examine how they vary together. The defining feature is measurement without assigning the exposure of interest. The data may come from questionnaires, records or behavioral observations. Correlational research can describe associations, test predictions and contribute to larger causal arguments, but association alone leaves competing explanations unresolved.

  Do students who report more regular sleep also report better concentration?

  Research designSee an example

• Cross-sectional study

  A cross-sectional study measures participants during one assessment period. In developmental psychology it commonly compares people of different ages at that time. It can also describe a population or examine concurrent associations. The design provides a snapshot: the researcher does not directly observe the same individual moving from one age or state to another.

  Do adults in different age groups report different preferences for learning resources?

  Research designSee an example

• Deception and debriefing

  Deception gives participants misleading information or conceals a feature that matters to their understanding. Debriefing explains the research and addresses misconceptions; it cannot retrospectively make any deceptive design acceptable.

  Is withholding or misrepresenting information necessary and defensible, and how will participants’ understanding and welfare be restored?

  Research ethicsSee an example

• Effect size

  An effect size quantifies the magnitude of a difference or relationship. It may use original units, such as seconds, or a standardized scale. Choosing the metric and reporting uncertainty makes results more useful for interpretation, comparison, and future study planning than a significance label alone.

  How large is the estimated difference or relationship on a scale that answers the research question?

  AnalysisSee an example

• Hypothesis testing

  Hypothesis testing evaluates data relative to a specified statistical model and decision rule. A p-value describes how unusual the observed test statistic, or a more extreme one, would be under the null model and its assumptions. It is not the probability that the null hypothesis is true.

  How incompatible are these data with the specified null model, given the assumptions and analysis plan?

  AnalysisSee an example

• Informed consent

  Informed consent supports a voluntary, understandable decision about research participation. It is a communication process that continues during a study, rather than merely a signature or a checkbox.

  What would a prospective participant need to understand to make a meaningful choice about this particular study?

  Research ethicsSee an example

• Longitudinal study

  A longitudinal study measures the same participants across time. This makes it possible to investigate individual change and stability rather than reconstructing change from different people measured once. Longitudinal describes a time structure, not a guarantee of causal identification. Repeated questionnaires, interviews, observations or experimental follow-up can all be organized longitudinally.

  How does an individual’s sense of belonging change during the first year of university?

  Research designSee an example

• Meta-analysis

  Meta-analysis statistically combines estimates from separate studies when their questions, designs, and outcomes can support a meaningful synthesis. Common methods use weighted averages. The pooled number is an answer to a specified model-based question, not an automatic verdict that overrides bias or differences between studies.

  Can compatible study estimates be combined, and what does the combined estimate mean given their differences?

  AnalysisSee an example

• Naturalistic observation

  Researchers systematically record behavior in the setting where it normally occurs, without introducing the focal manipulation. The method can reveal patterns that a laboratory task misses. Natural surroundings do not automatically make a sample representative or a measurement unbiased. Observation still requires explicit choices about whom to watch, when to watch and what counts as an event.

  How do students share speaking time in an ordinary study group?

  Research designSee an example

• Open data and materials

  Open research resources make the evidence and procedures behind a claim inspectable and reusable when sharing is appropriate. Useful transparency includes documentation, analysis code, and materials, with justified restrictions where participant protection or rights require them.

  Could another researcher understand what was done, rerun the analysis, or plan a new test from the resources provided?

  Open scienceSee an example

• Operationalization

  Operationalization translates a conceptual variable into specified observations, tasks, questions, or manipulations. The resulting score is an indicator of a construct. Naming a questionnaire does not replace specifying what is recorded, when it is recorded, and how responses become an analyzable variable.

  What exactly will count as evidence of the construct in this study?

  MeasurementSee an example

• Preregistration

  Preregistration records a research plan before the relevant outcomes are known. It helps readers distinguish decisions made in advance from decisions prompted by observing results, while keeping exploration a legitimate part of research.

  Which hypotheses, measurements, exclusions, and analysis decisions can be specified before learning the outcomes that will test them?

  Open scienceSee an example

• Qualitative interview

  Qualitative interviewing explores how participants describe experiences and make sense of them. A semistructured interview uses guiding questions while allowing relevant follow-up. An interview is a data-collection method rather than a complete research design: the study also needs a sampling rationale, an analytic approach and a clear account of how interpretations were developed.

  How do first-generation students describe their experience of asking for academic help?

  Research designSee an example

• Quasi-experiment

  A quasi-experiment evaluates an intervention or change without random assignment to the relevant conditions. Researchers may compare existing groups, examine change before and after an intervention, or use a time series. These designs can inform causal questions when the comparison and assumptions are defensible. Simply collecting a before-and-after score does not identify the intervention’s effect.

  Did introducing a new tutoring program improve performance beyond changes that would otherwise have occurred?

  Research designSee an example

• Randomized experiment

  A researcher deliberately changes a condition and uses a chance process to allocate participants to conditions. The comparison asks whether the intervention changes an outcome while reducing systematic baseline differences between groups. Randomization strengthens causal inference; the quality of measurement, implementation, follow-up and the comparison condition still matters.

  Does a brief retrieval exercise improve later recall compared with rereading the same material?

  Research designSee an example

• Registered Reports

  Registered Reports move substantive peer review to an earlier point, before the relevant results are known. A qualifying format offers in-principle acceptance based on the question and proposed methods, subject to stated conditions.

  Would the planned study produce informative evidence regardless of whether its results support the researchers’ preferred hypothesis?

  Open scienceSee an example

• Reliability

  Reliability concerns consistency under specified measurement conditions. Relevant conditions include repeated occasions, multiple items, and different observers. Reliability evidence belongs to scores obtained in a particular setting and population; a familiar instrument name is not a permanent guarantee of dependable measurement.

  How much would scores vary because of the measurement process rather than the differences of interest?

  MeasurementSee an example

• Replication

  Replication examines a prior claim using new data. This guide reserves computational reproducibility for obtaining consistent results from the original data and computational workflow, while recognizing that disciplines sometimes use the terms differently.

  What new evidence would increase or decrease confidence in the prior claim, and which study conditions matter to that interpretation?

  Open scienceSee an example

• Research data privacy

  Privacy planning asks what information a study needs, who can access it, and what could happen if it is disclosed or linked with other information. Removing names alone does not settle those questions.

  What could reveal a participant’s identity or sensitive circumstances, and which protections match the study’s actual risks?

  Research ethicsSee an example

• Research ethics review

  Research ethics review examines a proposed study’s treatment of participants before the relevant activities begin. In covered US research, an institutional review board applies regulatory criteria; other jurisdictions use their own systems.

  Are the proposed risks, recruitment practices, consent arrangements, and protections acceptable for this study and participant population?

  Research ethicsSee an example

• Sampling

  Sampling is the process of selecting cases from a defined population or access route. A good plan connects recruitment to the inference sought. Probability sampling supports design-based population estimation; nonprobability approaches require a defensible inferential basis and explicit assumptions rather than a blanket claim of representativeness.

  Who can enter the study, who actually enters, and to whom can its findings reasonably apply?

  MeasurementSee an example

• Statistical power

  Statistical power is the probability that a specified testing procedure rejects its null hypothesis when a particular alternative and other assumptions hold. It is a property of a planned design under specified conditions. It is not a guarantee that a study will discover the true effect.

  How often would this design detect an effect of the size we have a reason to care about?

  AnalysisSee an example

• Survey research

  Survey research gathers participants’ reports about experiences, beliefs, feelings or behavior, with explicit attention to sampling. Surveys can describe a population or examine associations, and a survey can contain an experiment. A questionnaire is a measurement tool, so calling a study a survey does not fully describe its design or justify a causal conclusion.

  How common is a particular study habit among students at a specified university?

  Research designSee an example

• Systematic review

  A systematic review answers a defined question through planned, transparent methods for finding, selecting, appraising, and synthesizing relevant studies. Its credibility depends on those procedures and the evidence located. A review can be systematic without combining effects statistically, and a statistical synthesis is not a substitute for a sound review.

  What does the relevant body of evidence show when studies are sought and assessed using explicit methods?

  AnalysisSee an example

• Within-subjects design

  The same participants complete each condition of a comparison. This allows the researcher to compare a person with that person under another condition, reducing some variation arising from stable individual differences. In an experiment, condition order needs deliberate control. Repeated measurement alone does not make a study experimental, and not every manipulation can be reversed or repeated meaningfully.

  Does the same participant perform differently with two versions of a task?

  Research designSee an example