Manuel B. Garcia is a professor of information technology and the founding director of the Educational Innovation and Technology Hub (EdITH) at FEU Institute of Technology, Manila, Philippines. Read More

Contact Info

1607, FEU Tech Building,
P. Paredes St, Sampaloc,
Manila, Philippines
mbgarcia@feutech.edu.ph

Follow Me

Variable Role Selector

Describe how a variable functions in your research question or model to identify whether it is an outcome, predictor, grouping variable, covariate, confounder, mediator, moderator, control, exposure, or latent construct.

Identify the role of a variable

Recommendations update as you describe its function.

No data stored
1What is the main function of the variable?
2How is the variable obtained?
3When does it occur relative to the focal outcome?
4How does it relate to the focal predictor and outcome?
5How will it be used in the analysis or design?

Describe the variable

Its likely role and interpretation will appear here.

Role depends on the modelThe same measured variable can serve different roles in different research questions, analyses, or causal models.

Variable roles come from the research question

A variable is not permanently independent, dependent, or controlled.

Its role depends on how it functions in the study. The same construct may be an outcome in one model, a predictor in another, a mediator in a mechanism model, or a moderator in an interaction analysis.

The selector distinguishes statistical roles from stronger causal interpretations, which require temporal ordering, design justification, and assumptions beyond the variable label itself.

Outcome

The response or result being described, compared, predicted, or evaluated.

Predictor

A measured or assigned variable used to explain variation in an outcome.

Mechanism

A mediator transmits a relationship; a moderator changes its strength or direction.

Adjustment

Covariates, controls, and confounders are included for different reasons.

Common variable roles at a glance

Use these labels only after defining the research question and model.

RoleWhat it doesExample
Dependent variableOutcomeRepresents the response being explained or comparedPosttest score, recovery status, engagement level
Independent variableTreatmentDefines an assigned experimental conditionTeaching method, dosage, interface condition
PredictorExposureExplains or predicts an outcome without necessarily being manipulatedStudy time, prior achievement, environmental exposure
Grouping variableDivides observations into categories for comparisonProgram group, school type, grade level
CovariateProvides adjustment or improves precisionBaseline score, age, pretest value
ConfounderCan distort the predictor–outcome relationshipSocioeconomic status in a study of school type and achievement
MediatorRepresents a mechanism through which an effect operatesMotivation mediating the effect of feedback on performance
ModeratorChanges the strength or direction of a relationshipPrior knowledge moderating the effect of an intervention
Control variableIs held constant or statistically controlledTask duration, device type, class level
Latent variableIs estimated from multiple observed indicatorsAnxiety, trust, self-efficacy, cognitive engagement

Before assigning a variable role

  • Write the research question. Identify what is being explained and what is doing the explaining.
  • Establish temporal order. Predictors, exposures, and mediators should occur before the outcomes they are proposed to influence.
  • Separate design from analysis. Manipulated treatments differ from merely observed predictors.
  • Justify adjustment. Do not add variables automatically without considering their causal role.
  • Define measurement. Specify whether the role refers to an observed score, category, composite, or latent construct.

Avoid treating every adjustment variable as a confounder

A covariate is a broad statistical term. A confounder has a more specific causal meaning: it is related to both the exposure and outcome and is not simply a consequence of the exposure.

Adjusting for mediators, colliders, or post-treatment variables can change the estimand or introduce bias. Variable selection should therefore be based on the research question and a defensible causal model, not only automated statistical criteria.

Good practice: create a conceptual or causal diagram before deciding which variables to adjust, test as interactions, or model as mediators.

Frequently asked questions

Identify what the variable represents in the research question, when it occurs, whether it is manipulated or observed, whether it predicts or explains another variable, and whether it changes, transmits, or distorts a relationship.

An independent variable is a variable used to define groups, conditions, exposures, or predictors. In experiments it may be manipulated, while in observational studies it is usually measured rather than assigned.

A dependent variable is the measured outcome or response that the study seeks to explain, compare, predict, or evaluate.

The terms overlap, but predictor is often preferred in regression and observational research because it does not imply experimental manipulation or causal independence.

A grouping variable divides observations into categories or conditions for comparison, such as treatment group, grade level, sex, institution type, or study condition.

A covariate is an additional measured variable included in an analysis to improve precision, account for baseline differences, or adjust an estimated relationship.

A confounder is a variable associated with both the exposure or predictor and the outcome that can distort their estimated relationship if not appropriately handled.

A mediator is an intervening variable through which a predictor or exposure may influence an outcome. Mediation requires a defensible causal and temporal sequence.

A moderator changes the strength or direction of the relationship between a predictor and an outcome. Moderation is commonly tested using an interaction term.

A control variable is held constant by design or statistically adjusted so the focal relationship can be examined with less influence from alternative explanations.

A latent variable is a construct that is not observed directly and is estimated from multiple indicators, such as motivation, anxiety, engagement, or socioeconomic status.

Yes. The role depends on the research question and model. Age may be an outcome in one study, a predictor in another, a moderator in another, or a confounder requiring adjustment.

Disclaimer: This selector provides general educational guidance. It does not establish causal relationships or replace methodological supervision, subject-matter expertise, statistical consultation, or a complete causal analysis. The final variable roles should reflect the exact research question, design, timing, measurement, and assumptions.