← Study Design & Causal Reasoning: Applied Clinic

Practice & understand

Formative practice with explanations. No registration or certification.

1. A protocol targets all district children but samples only clinic users. What is the most appropriate design-review response?08. Design defenseKeep the target population and ignore the frame differenceIncrease the number of clinic users onlyDescribe the sample as random without further explanationReassess the frame or explicitly revise the target, documenting the consequences

2. A variable lies on a pathway from an exposure to the outcome. What should be considered before adjusting for it when estimating a total effect?05. Causal assumptionsAdjustment is always required if the variable predicts the outcomeIt may be a mediator, so adjustment can block part of the effect of interestA large sample makes its causal role irrelevantIts role can be determined solely from its p-value

3. A household has selection probability 0.10. One of its two eligible children is selected at random. What is that child’s overall probability and base inverse-probability weight?07. Selection weightsProbability 0.05; weight 20Probability 0.10; weight 2Probability 0.50; weight 10Probability 0.20; weight 5

4. One analyst reports the percentage of appointments attended; another reports the percentage of families attending at least once. Why can the results differ even with correct arithmetic?01. Objective and estimandOne percentage must always be a recording errorThe analysis units and denominators answer different questionsBoth necessarily estimate the same population riskThe difference proves confounding

5. A preliminary search finds two articles from the same trial and one separate interview study. What is the best next step?02. Evidence-gap briefCount all three as independent trialsPool their reported percentages immediatelyLink reports from the same study and map what each design contributesConclude that no further evidence exists

6. In a case–control study, investigators deliberately select controls who are unexposed. Why is this problematic?04. Comparison groupsIt guarantees equal follow-upIt necessarily improves precisionIt removes the need to define the source populationControl selection depends on the exposure being compared

7. A team expects non-response related to access barriers. Which plan is most defensible?06. Bias auditIncrease the sample and assume selection bias disappearsReport only results agreeing with the main hypothesisPlan repeat contacts, document participation patterns and specify assumption-linked sensitivity checksReplace absent households with any nearby volunteers

8. A cohort labels people exposed at enrolment because they join a programme several months later. Which design issue needs attention first?03. Design timelinesEligibility, exposure classification and the start of follow-up may be misalignedAll observational studies must use case–control samplingUsing more decimal places in the result will remove the issueA 99% confidence interval automatically corrects classification