Research Design and Planning
From question to design
The candidate must be able to move from a research question to a defensible design. Know the difference between a hypothesis-generating (exploratory) and hypothesis-testing (confirmatory) study, and why a confirmatory study needs a pre-specified primary outcome, a sample-size justification, and an analysis plan fixed before data collection.
Core concepts
Understand the main design families: experimental (randomized controlled trials), quasi-experimental, observational (cohort, case-control, cross-sectional), and diagnostic accuracy studies. For each, know the unit of analysis, the timing of exposure and outcome measurement, and the principal threat to validity. Learn the vocabulary of bias by design: selection bias, confounding, information bias, and loss to follow-up. Know the purpose of randomization (balance on known and unknown confounders), allocation concealment (protecting the sequence), and blinding (protecting assessment). Distinguish these three; questions frequently conflate them.
Traps
- Assuming randomization removes confounding after the fact; it does so only at baseline, and differential attrition can reintroduce it.
- Confusing a pilot or feasibility study with a definitive trial; pilot studies estimate feasibility parameters, not treatment effects.
- Treating a cross-sectional design as able to establish temporal order.
- Forgetting that a study can be internally valid but not generalizable.
Worked rule of thumb
When a question describes a study, first ask: was the exposure assigned or observed? If assigned, was allocation concealed and were assessors blinded? If observed, what was the comparison group and how were confounders addressed? That sequence resolves most design questions.
Planning documents
Know what belongs in a protocol: objectives, design, eligibility, sample size, outcomes, statistical methods, and governance. The sample-size calculation must match the primary outcome and the planned test.