Research glossary

A test of whether three or more group means differ, by comparing variation between groups to variation within them.

A test of association between categorical variables, comparing observed counts in a contingency table with the counts expected under independence.

A range of plausible values for a population parameter, constructed so that the procedure captures the true value a stated share of the time (usually 95%).

A third variable that influences both the presumed cause and the outcome, creating a spurious or distorted association.

The degree to which a measure actually captures the concept it claims to measure.

A standardized measure (−1 to +1) of how strongly two variables move together. Association, not causation.

A permanent, unique identifier for a scholarly object (10.xxxx/…) that resolves to its current location — the stable way to cite and link research.

A standardized measure of how large a difference or relationship is — Cohen’s d, Pearson’s r, odds ratios — independent of sample size.

A qualitative methodology for building theory from data through iterative coding, constant comparison, and theoretical sampling.

An author metric: h papers with at least h citations each. Balances productivity and impact, with well-known biases.

A specific, testable prediction about the relationship between variables — falsifiable in principle, stated before the data are seen.

A journal metric: the mean citations in a year to items the journal published in the previous two years. A journal average — not a paper’s or author’s quality score.

The independent variable is what you manipulate or compare; the dependent variable is the outcome you measure to see the effect.

Internal validity is whether the study really shows what it claims within its setting; external validity is whether the result generalizes beyond it.

The survey format asking respondents to rate agreement on a symmetric scale (e.g. strongly disagree → strongly agree), typically with 5 or 7 points.

A synthesis of existing research on a topic that maps what is known, where the field disagrees, and what gap the new work fills.

The statistical pooling of effect sizes from multiple studies into one overall estimate, with heterogeneity quantified.

Research that combines quantitative and qualitative strands in one study, with an explicit design for how they connect.

A test that makes minimal assumptions about the data’s distribution, typically by working on ranks — Mann-Whitney, Wilcoxon, Kruskal–Wallis, Spearman.

The symmetric, bell-shaped distribution that arises whenever many small independent influences add up — the default assumption behind parametric statistics.

The default assumption a statistical test tries to discredit — usually that there is no effect, no difference, or no relationship.

Publishing that makes research freely readable — gold (journal-side, often via APCs), green (self-archived copies), diamond (free for both sides).

Translating an abstract concept into something concrete you can measure — defining exactly what counts as "stress", "engagement", or "success" in your study.

The probability of observing data at least as extreme as yours if the null hypothesis were true. It is not the probability that the hypothesis is true.

The evaluation of scholarly work by independent experts before publication — the quality filter, imperfect but load-bearing, of academic publishing.

A manuscript shared publicly (arXiv, bioRxiv, SSRN, OSF) before — or while — undergoing journal peer review.

Publicly time-stamping your hypotheses, design, and analysis plan before collecting data, so confirmatory and exploratory results stay distinguishable.

The reporting standard for systematic reviews — a 27-item checklist and the flow diagram tracking records from identification to inclusion.

The tendency for significant, positive results to get published while null results stay in file drawers — skewing what the literature shows.

An experiment in which participants are randomly assigned to intervention or control, making the groups comparable and causal inference possible.

A family of models that predict an outcome from one or more predictors, quantifying each predictor’s contribution while holding the others constant.

The consistency of a measure — across items (internal consistency), time (test–retest), and raters (inter-rater).

The finding, across several fields, that many published results fail to reproduce when independently repeated — and the reform movement it triggered.

The number of participants or units in a study — determined by power analysis (quantitative) or saturation (qualitative), not convention.

Systematic error introduced when the people (or units) in your sample differ from the population you want to describe.

The point in qualitative data collection where additional interviews or observations stop yielding new codes, themes, or insights.

A measure of how spread out values are around their mean, in the same units as the data.

The standard deviation of an estimate across hypothetical repeated samples — how precisely you know a mean, proportion, or coefficient.

The probability that a study will detect an effect of a given size if it really exists — conventionally targeted at 80%.

The convention of treating a result as unlikely to be pure chance when its p-value falls below a preset threshold, usually 0.05.

A literature review conducted as a documented method: registered question, comprehensive search, explicit criteria, and reproducible screening.

A test of whether two means differ: one sample against a known value, two independent groups, or paired measurements.

A widely used qualitative method for identifying and interpreting patterns (themes) across a dataset, canonically via Braun and Clarke’s six phases.

Strengthening a conclusion by approaching it through multiple methods, data sources, researchers, or theories.

A Type I error is a false positive (rejecting a true null); a Type II error is a false negative (missing a real effect).