Causal Inference Questions

Establishing cause-and-effect from observational and experimental data. Covers correlation versus causation, confounding, treatment-effect estimation, and quasi-experimental methods such as difference-in-differences, matching, and instrumental variables. Includes incrementality reasoning when true randomization is not possible.

HardTechnical
83 practiced

Design a sensitivity analysis to quantify how strong an unobserved confounder would have to be to change your estimated treatment effect to zero. Explain Rosenbaum bounds and the E-value, show how you would compute an E-value for an estimated risk ratio, and give a plain-language interpretation a non-technical stakeholder could act on.

MediumTechnical
57 practiced

Explain the difference-in-differences (DiD) identification strategy: what it estimates, and state the parallel trends assumption precisely. Describe at least two empirical checks or falsification tests you would run to assess whether parallel trends is plausible (for example, pre-period trend comparison or a placebo outcome). Give a concrete business scenario, such as a price change applied in one country but not a comparable one, where DiD is the right tool.

HardTechnical
67 practiced

An online experiment recruited users through voluntary sign-up, and the treatment group skews younger than control. Explain how selection into the sample (not confounding) biases a naive treatment-effect estimate, and describe how inverse probability weighting, post-stratification, or transportability could be used to recover an estimate for the target population.

EasyTechnical
119 practiced

What is Simpson's paradox? Construct a concise numeric example with at least two groups where the aggregated association points in the opposite direction from the association within each group, and explain why stratifying by the confounding variable reveals the true relationship.

MediumTechnical
82 practiced

You are analyzing a recruiting funnel: applicants -> interviewed -> hire outcome. Restricting your analysis to only the applicants who were interviewed, you observe a negative relationship between years of experience and interview performance. Explain why this could be collider bias rather than a real effect, and describe how you would detect and avoid introducing this kind of bias.

Unlock Full Question Bank

Get access to all 32 Causal Inference interview questions and detailed answers.

Sign in to Continue

Join thousands of developers preparing for their dream job.