引導式測驗 · 中等 等級
辛普森悖論測驗
測試聚合與子組逆轉、分母權重、伯克利招生示例以及因果解釋的限制。
問題 1 的 8
什麼模式定義了辛普森悖論?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- 什麼模式定義了辛普森悖論?
- 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
- 不同的子群組規模如何導致反轉?
- 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
- 柏克萊大學的錄取分析說明了什麼?
- 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
- 當治療效果在子群組與整體中出現反轉時,分析師應該檢查什麼?
- 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
- 分析師應如何解釋辛普森類型的反轉?
- 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