Chapter 5. Conceptual Framework and Hypothesis Development
5.1 Theory versus conceptual framework
A theory offers an explanatory account of relationships or mechanisms. A conceptual framework is the study-specific model that identifies the constructs, proposed links, boundary conditions and covariates relevant to the research questions. A diagram alone is not a framework: each arrow needs a theoretical and empirical rationale.
5.2 Building the framework
- Define every construct precisely and distinguish it from related concepts.
- Identify a theory or coherent set of theories that explains why a relationship should exist.
- Review evidence for each proposed link, including inconsistent findings.
- Specify the outcome, predictor(s), mediators, moderators and covariates only where conceptually justified.
- Show expected direction when theory supports it; do not add arrows merely to make the diagram look sophisticated.
- Write a short narrative explaining each relationship and its assumptions.
- Translate each relationship into a research question or hypothesis and an analysis plan.
5.3 Example framework
Illustrative model: perceived supervisor support (predictor) → work engagement (outcome). Organisational tenure and job type might be considered covariates if there is a theoretical reason and sufficient data. A mediator such as psychological safety or a moderator such as workload should be included only if the study has a defensible theory, valid measures, adequate sample size and a suitable analysis plan.
| Element | Meaning | Key caution |
|---|---|---|
| Predictor | Variable used to explain or predict variation in an outcome | A predictor is not automatically a cause |
| Outcome | Primary variable the study seeks to explain or estimate | Define scale and scoring before analysis |
| Mediator | Proposed pathway through which a relationship may operate | Cross-sectional mediation has serious temporal limitations |
| Moderator | Variable that changes the strength or direction of a relationship | Test an interaction; explain conditional effects |
| Covariate | Variable included for a defined analytical reason | Do not adjust indiscriminately; adjustment can introduce bias |
5.4 Hypothesis writing
A directional hypothesis predicts the direction of a relationship (for example, higher perceived support is associated with higher engagement). A non-directional hypothesis predicts a relationship without specifying direction. The choice should follow theory and prior evidence, not convenience. A null hypothesis often states no association or no difference in the population, but reporting should focus on estimates and uncertainty rather than treating a p-value as proof that a null is true.
5.5 Causal diagrams and confounding
A causal diagram can help clarify assumptions about predictors, outcomes and confounders. A confounder is a common cause of the exposure/predictor and outcome. Including every available variable is not a solution: controlling for a mediator may remove part of the relationship of interest, while conditioning on a collider can create a spurious association. Covariate selection should be grounded in the causal question and domain knowledge.