Guide-linked quiz · Medium Level
Simpson's Paradox Quiz
Tests aggregate-versus-subgroup reversals, denominator weighting, the Berkeley admissions example, and limits on causal interpretation.
Question 1 of 8
What pattern defines Simpson’s paradox?
Preview every question8 questions
Answers and explanations appear as you take the quiz.
- What pattern defines Simpson’s paradox?
- An association reverses or disappears after data are combined across subgroups
- A trend appears only after random noise is added
- Every subgroup has identical sample size
- A causal effect is proven by an aggregate rate
- How can different subgroup sizes contribute to a reversal?
- They guarantee each within-group rate is unbiased
- They make subgroup denominators irrelevant
- They change the weights used in the aggregate rate
- They change the labels on the outcome variable
- What did the Berkeley admissions analysis illustrate?
- The aggregate rate always proves discrimination
- The department-level table alone fully identifies causal mechanisms
- All departments had equal applicant composition
- Aggregate rates can differ from department-level patterns because application patterns and department admission rates vary
- When a treatment effect reverses by subgroup and overall, what should an analyst inspect?
- The chart color used for the treatment group
- Denominators, group composition, assignment process, and the causal role of subgroup variables
- Only the group with the largest sample
- The overall percentage rounded to fewer decimals
- Why is stratifying on every available variable not a safe automatic fix?
- Stratification always removes every source of confounding
- Subgroup data cannot be displayed as rates
- A mediator is always a measurement error
- Conditioning on a mediator or collider can change the causal question or introduce bias
- What does an aggregate rate represent when subgroups have different sizes?
- A weighted average of subgroup rates using their denominators
- The unweighted mean of subgroup percentages in every case
- The rate for the largest subgroup only
- A causal effect that ignores group composition
- How should analysts interpret a Simpson-type reversal?
- It proves aggregate data are always wrong
- It shows randomization failed in every subgroup
- It is a prompt to investigate composition and causal assumptions, not automatic proof of one mechanism
- It proves that subgroup data are always the causal truth
- What should a useful report include when aggregate and subgroup results differ?
- A causal claim without describing the assignment process
- Both views, their denominators and weights, and the estimand or design assumptions
- Only the aggregate rate because it uses all observations
- Only the subgroup with the strongest effect
Keep testing yourself
More quizzes picked for your level.