Instructions and covered skills
Read each scenario carefully before selecting a response. Focus on the causal relationship being claimed, the evidence provided, and variables that may affect interpretation. Do not assume that a correlation establishes cause and effect. Consider timing, comparison groups, assignment methods, and possible alternative explanations. Work in a quiet setting and turn off notifications before you begin. Stay focused on the information in each question rather than adding assumptions that are not supported by the scenario.
Key Areas
This assessment evaluates the ability to reason from observed evidence to defensible causal conclusions. It covers the distinction between correlation and causation, including recognition that an association can arise from shared causes, selection processes, changing conditions, or outcome-to-exposure relationships. Candidates interpret causal questions in terms of an intervention, an outcome, a target population, and a meaningful comparison condition.
Key capabilities include identifying confounders: variables that influence both the proposed cause and the outcome. The assessment also examines the ability to recognize mediators, colliders, and variables that should not automatically be controlled in an analysis. Questions use practical settings such as marketing campaigns, workplace programs, health services, product changes, and public policy.
Study-design reasoning is central. Candidates consider random assignment, matched comparisons, before-and-after analyses, difference-in-differences, regression adjustment, and natural experiments. They assess whether groups were comparable before an intervention, whether exposure timing is credible, and whether a comparison group represents what would likely have happened without the intervention.
Recommended Preparation
Prepare by practicing how to state a causal claim precisely: identify the action or exposure, the outcome, the affected population, and the counterfactual comparison. Review common sources of bias, including confounding, selection bias, attrition, measurement changes, spillovers, seasonality, and regression to the mean. For each scenario, ask what else could explain the observed difference and what information would separate those explanations.
It is also useful to sketch simple causal diagrams showing how variables may relate to one another. Review when adjustment can help, when it can block part of an effect of interest, and when conditioning on a common effect can introduce a false association. Finally, practice matching the strength of a conclusion to the design: randomized evidence can support stronger causal claims than an uncontrolled observational comparison, while both may still require careful attention to implementation and measurement.