Start here · Beginner
Research, statistics and publication glossary
Definitions of research, review, statistics, citation, qualitative and engineering terms with links to tutorials.
16 min read · Nov 14, 2026
Key points
- Learn the meaning of technical terms in context.
- Distinguish related concepts that are often confused.
- Use the linked tutorials to apply each concept.
Contents
Definitions of research, review, statistics, citation, qualitative and engineering terms with links to tutorials.
Use this glossary while reading the tutorials, not as a substitute for studying a method's assumptions. Each entry links to an article that explains its use and provides relevant sources. Definitions describe common research usage; specialist fields may use a term more narrowly. A familiar statistical name does not by itself identify an appropriate analysis.
Questions and study design
| Term | Meaning | Tutorial |
|---|---|---|
| Research question | The specific uncertainty a project intends to address using a defined body of evidence. | Question |
| Objective | A statement of what the study will estimate, describe, compare or interpret. | Question |
| Hypothesis | A testable proposition or expectation; it is distinct from the general objective. | Question |
| Estimand | The precise quantity targeted by an analysis, defined by its population, comparison, outcome and relevant conditions. | Question |
| Population | The collection of people, objects, events or settings about which a question is asked. | Design |
| Sample | The subset actually observed; how it was selected affects what can be inferred. | Design |
| Experimental unit | The smallest unit independently assigned to a treatment in the design. | Replication |
| Observational unit | The entity on which a measurement is recorded; it may differ from the experimental unit. | Replication |
| Technical replicate | A repeated measurement of the same underlying material or unit, useful for assessing measurement variation. | Replication |
| Pseudoreplication | Treating dependent observations as independent replication in a way that overstates the information in the design. | Replication |
| Random allocation | Assignment by a random mechanism, used to support treatment comparisons under the design. | Design |
| Random sampling | Selection from a population by a probability mechanism; different from assigning treatment at random. | Design |
| Blinding | Keeping specified information, such as treatment identity, unavailable to people whose decisions it could influence. | Design |
| Control condition | A comparison condition chosen to help interpret what changes when the treatment or procedure differs. | Design |
| Confounding | Distortion of a causal comparison because exposure groups differ in relevant background causes of the outcome; use a causal model to distinguish confounders from mediators or colliders. | Regression |
| External validity | The extent to which an inference applies beyond the studied people, materials, conditions or settings. | Discussion |
Measurement and variables
| Term | Meaning | Tutorial |
|---|---|---|
| Operational definition | The concrete rule used to identify, manipulate or measure a concept. | Question |
| Construct | A theoretical concept, such as confidence, that must be represented through an appropriate measurement approach. | Methods |
| Validity of a measure | The support for interpreting measurements as representing the intended concept for a particular use. | Methods |
| Reliability | Consistency of measurements under specified repeat conditions; consistency alone does not establish validity. | Methods |
| Outcome | The response measured or interpreted to address the question. | Question |
| Predictor | A variable included to describe or predict variation in an outcome; its inclusion does not establish causation. | Regression |
| Covariate | An additional measured variable used in design or analysis; its role should be justified. | Regression |
| Categorical variable | A variable whose values indicate categories, such as material type or response option. | Describing data |
| Ordinal variable | A categorical variable with an ordering whose category gaps need not be equal. | Describing data |
| Continuous variable | A variable modelled as able to take values across a numerical interval, even when recorded with finite precision. | Describing data |
| Measurement error | Difference between a measured value and the relevant reference or target quantity, considered under a measurement model. | Engineering |
| Denominator | The total or reference quantity used in a fraction, proportion or rate; changing it changes the interpretation. | Tables |
Descriptive statistics and displays
| Term | Meaning | Tutorial |
|---|---|---|
| Mean | The sum of numerical values divided by their number. | Describing data |
| Median | A central ordered value that divides the distribution into lower and upper halves, with conventions for even sample sizes. | Distributions |
| Standard deviation, SD | A measure of spread around the mean, expressed in the same units as the observations. | Uncertainty |
| Variance | A measure of spread based on squared deviations from the mean; its units are squared. | Uncertainty |
| Standard error, SE | The standard deviation of an estimator's sampling distribution, usually estimated from the data and model. | Uncertainty |
| Quantile | A value associated with a specified cumulative proportion of a distribution; sample calculation conventions vary. | Distributions |
| Interquartile range, IQR | The difference between the 75th and 25th percentiles, describing the middle half of a distribution. | Distributions |
| Skewness | Asymmetry of a distribution, often visible as a longer tail in one direction. | Distributions |
| Outlier | An observation unusual relative to a specified distribution or model; unusual does not mean erroneous. | Robustness |
| Histogram | A display that groups numerical observations into bins and shows their frequencies or densities. | Distributions |
| Density | Probability per unit along a continuous scale; area, rather than height alone, corresponds to probability. | Distributions |
| Error bar | A graphical interval whose meaning must be stated, such as SD, SE or a confidence interval. | Figure design |
| Percentage point | The unit of an absolute difference between percentages; 20% to 30% is ten percentage points. | Writing |
| Relative change | Change divided by a stated reference value; 20% to 30% is a 50% relative increase. | Writing |
Statistical inference
| Term | Meaning | Tutorial |
|---|---|---|
| Parameter | A quantity in a model or population, such as a population mean or regression coefficient. | Uncertainty |
| Estimator | A rule that maps data to an estimate of a target quantity. | Uncertainty |
| Estimate | The numerical result obtained by applying an estimator to the observed data. | Uncertainty |
| Effect size | A quantitative description of a contrast or relationship, in original or standardised units. | Effect sizes |
| Confidence interval, CI | An interval from a procedure with stated repeated sampling coverage under its assumptions. | Uncertainty |
| Credible interval | A Bayesian interval containing a stated posterior probability for a quantity under the specified model and prior. | Choosing analysis |
| Prediction interval | An interval for a future observation or quantity, accounting for relevant predictive variation. | Regression |
| Null hypothesis | The model restriction or reference proposition evaluated by a hypothesis test. | P values |
| P value | Under the specified null model, the probability of a test statistic at least as extreme as observed in the defined sense. | P values |
| Significance level, alpha | A prespecified threshold associated with the test's error control; it is not a measure of importance. | P values |
| Type I error | Rejection of a true null hypothesis under the testing framework. | P values |
| Type II error | Failure to reject a false null hypothesis for a specified alternative and testing setup. | Sample size |
| Statistical power | The probability of rejection under a specified alternative, design and analysis. | Sample size |
| Degrees of freedom | A quantity reflecting independent information available for estimating variation or defining a reference distribution. | Choosing analysis |
| Familywise error rate, FWER | The probability of at least one false rejection within a defined comparison family. | Multiple testing |
| False discovery rate, FDR | The expected proportion of false rejections among all rejections, with the proportion defined as zero when there are none. | Multiple testing |
| Adjusted p value | A p value transformed according to a specified multiple testing procedure and family. | Multiple testing |
| Statistical significance | A result meeting a stated statistical decision threshold; it does not establish importance or truth. | P values |
| Equivalence testing | Testing whether an effect lies within prespecified bounds representing differences considered practically negligible. | Choosing analysis |
| Bootstrap | A resampling approach used to approximate properties of an estimator; the resampling scheme must respect the design. | Robustness |
Models and associations
| Term | Meaning | Tutorial |
|---|---|---|
| Paired analysis | Analysis that retains the relationship between observations matched within the same unit or matched set. | Worked paper |
| Welch procedure | A comparison of independent means that estimates uncertainty without requiring equal population variances. | Choosing analysis |
| Nonparametric method | A method that avoids some particular parametric distribution assumptions; it still has assumptions and a target quantity. | Choosing analysis |
| Correlation | A measure of a specified form of association; Pearson correlation measures linear association. | Correlation |
| Regression coefficient | A model parameter describing how the fitted outcome changes with a predictor under the model's specification. | Regression |
| Residual | An observed outcome minus the value fitted by the model. | Regression |
| R² | In ordinary regression with an intercept, the fraction of sample variation around the outcome mean represented by the fitted model. | Regression |
| Interaction | A modelled dependence of one predictor's association on another variable, on a specified scale. | Regression |
| Odds ratio | A ratio of odds between conditions; odds are probability divided by one minus probability. | Regression |
| Risk ratio | The ratio of event probabilities in two specified groups over a defined period. | Effect sizes |
| Multilevel model | A model that represents variation across levels such as measurements within people or students within schools. | Choosing analysis |
| Overfitting | Fitting patterns particular to the development data that do not generalise reliably to new cases. | Computational research |
Missing data and reproducibility
| Term | Meaning | Tutorial |
|---|---|---|
| Missing completely at random, MCAR | A missingness mechanism independent of observed and unobserved data values in the model. | Missing data |
| Missing at random, MAR | A missingness assumption under which dependence on missing values is removed after conditioning on the relevant observed information. | Missing data |
| Missing not at random, MNAR | A missingness mechanism that still depends on missing values after conditioning on the observed information in the model. | Missing data |
| Complete case analysis | Analysis restricted to records with the measurements required by that analysis; its validity depends on the setting and assumptions. | Missing data |
| Multiple imputation | Creation and analysis of several plausible completed datasets, with uncertainty combined under an appropriate imputation model. | Missing data |
| Sensitivity analysis | Examination of how conclusions change under alternative reasonable assumptions, specifications or data decisions. | Robustness |
| Preregistration | A timestamped record of planned questions and methods made before the relevant outcomes are examined. | Protocol |
| Exploratory analysis | Analysis used to investigate patterns or develop ideas, distinguished transparently from prespecified confirmatory work. | Analysis plan |
| Data dictionary | Documentation of variable names, meanings, units, coding, allowed values and missing value conventions. | Data |
| Provenance | The documented origin of data or material and the transformations applied to it. | Data |
| Computational reproducibility | Obtaining consistent computational results from the same inputs and documented methods; usage varies across fields. | Reproducibility |
| Replication | A new investigation of a prior finding using new observations; specify what aspects of the design are repeated or changed. | Replication |
Reviewing and synthesising literature
| Term | Meaning | Tutorial |
|---|---|---|
| Narrative review | A literature based account organised around a question, interpretation or framework, with a transparent description of its scope and selection. | Narrative reviews |
| Systematic review | A review using explicit, planned methods to identify, select, appraise and synthesise evidence addressing a defined question. | Systematic reviews |
| Scoping review | A structured review used to map the extent, types, concepts or gaps of evidence within a defined scope. | Scoping reviews |
| Meta analysis | Statistical synthesis of sufficiently comparable estimates from multiple studies; it is not required in every systematic review. | Synthesis |
| Search strategy | The documented combination of concepts, terms, fields, limits and database syntax used to retrieve records. | Searching |
| Boolean operator | A search connector such as AND, OR or NOT; its effect depends on the database syntax and grouping. | Searching |
| Eligibility criteria | Prespecified rules for deciding which studies or reports address the review question. | Systematic reviews |
| Screening | Assessment of retrieved records and reports against eligibility criteria. | Systematic reviews |
| Record, report and study | A record is a database entry, a report is a document, and a study is an investigation that may produce several reports. | Systematic reviews |
| Evidence matrix | A structured table of study questions, designs, samples, findings and limitations used to support comparison and synthesis. | Narrative reviews |
| Risk of bias | The potential for systematic distortion of a result due to features of study design, conduct, analysis or reporting. | Systematic reviews |
| Heterogeneity | Differences among studies in clinical, methodological or statistical features; the intended meaning should be specified. | Synthesis |
| I² | An estimated proportion of variation in study estimates attributable to heterogeneity rather than sampling error; its value depends on study precision and is uncertain with few studies. | Synthesis |
| Publication bias | Selective availability of studies or reports related to their results or other characteristics. | Synthesis |
| PRISMA | A reporting guideline family for systematic reviews and related review types; reporting compliance does not itself establish methodological quality. | Review reporting |
| Synthesis | An explanation of relationships across evidence, including agreement, differences, uncertainty and implications. | Worked review |
Qualitative and engineering terms
| Term | Meaning | Tutorial |
|---|---|---|
| Reflexivity | Examination of how the researcher's position, assumptions and interactions shape qualitative research decisions and interpretations. | Qualitative research |
| Positionality | The researcher's social, professional and theoretical standpoint in relation to the inquiry. | Qualitative research |
| Code | An analytic label applied to a relevant feature of qualitative material; its role depends on the analysis approach. | Qualitative analysis |
| Theme | A developed pattern of meaning or organising concept, defined according to the chosen qualitative approach. | Qualitative analysis |
| Purposive sampling | Selection of cases because they can illuminate the research question, rather than by a probability sampling rule. | Qualitative methods |
| Saturation | A family of concepts concerning adequacy or diminishing new information in some qualitative approaches; not a universal sample size rule. | Qualitative methods |
| Audit trail | A record showing how research decisions and interpretations developed; its form depends on the method. | Qualitative reporting |
| Verification | Checking whether an implementation or solution correctly follows its specified equations, rules or requirements. | Engineering |
| Validation | Assessing whether a model or system represents the intended phenomenon or meets its intended use sufficiently well. | Engineering |
| Calibration | Establishing a relationship between instrument indications and reference quantities under specified conditions. | Engineering |
| Baseline | A reference method or system used to make a performance comparison interpretable. | Engineering |
| Benchmark | A defined task, dataset or evaluation procedure used for comparison; performance depends on its relevance and fairness. | Engineering |
| Data leakage | Information entering development or evaluation in a way that makes reported performance unrepresentative of genuine new cases. | Computational research |
| Cross validation | Repeated fitting and assessment across data partitions, with preprocessing and grouping handled within the appropriate partitions. | Computational research |
| Principal component analysis, PCA | A transformation to orthogonal directions of variation in numerical data; preprocessing affects the resulting components. | Advanced plots |
| Censoring | Partial observation of an event time, such as knowing that an event had not occurred by the final observation. | Survival plots |
Citations and publication
| Term | Meaning | Tutorial |
|---|---|---|
| Citation | An in text attribution connecting a claim, idea, method or quotation to its source. | References |
| Reference entry | Bibliographic information identifying a cited source so readers can locate and evaluate it. | References |
| Primary source | A source presenting the original work or evidence relevant to a claim; what is primary depends on the question. | Reading |
| Secondary source | A source that describes, interprets or synthesises other work; a review is often secondary for empirical findings. | Reading |
| Paraphrase | An accurate restatement of an idea in a new formulation that still requires attribution. | Originality |
| Direct quotation | The source's exact wording, clearly marked and attributed with an appropriate location when available. | Originality |
| Plagiarism | Presenting another person's words, ideas or other work without appropriate acknowledgement. | Originality |
| DOI | A digital object identifier linking to a registered scholarly object; having one does not guarantee quality or peer review. | References |
| ORCID | A persistent researcher identifier used to help distinguish people and connect their contributions. | Authorship |
| Et al. | An abbreviation meaning “and others”, used according to the citation style when shortening an author list. | References |
| Preprint | A publicly shared manuscript version typically made available before journal peer review; check the specific version and status. | Publication |
| Peer review | Evaluation by relevant reviewers within a publication process; it does not guarantee that all errors are detected. | Publication |
| Reporting guideline | A structured recommendation for what information a report should contain for its study type. | Methods |
| CRediT | A taxonomy of contributor roles; assigning roles does not by itself determine eligibility for authorship. | Authorship |
| Competing interest | A relationship or interest that could influence, or be perceived to influence, research or reporting judgment. | Declarations |
| Correction | A published change addressing an error while identifying the affected content and preserving the record appropriately. | Publication |
| Retraction | A formal notice withdrawing reliance on a published work for specified reasons; read the notice rather than inferring its cause. | Publication |
| Supplementary material | Additional methods, data or other content accompanying the main article; it should remain accessible and connected to the argument. | Structure |
| Data availability statement | A description of where data can be accessed or why access is limited, including relevant conditions. | Reproducibility |
| Revision response | A document explaining how each editorial or reviewer comment was addressed, with reasons for any disagreement. | Peer review |