引導式測驗 · 中等 等級
數據科學測驗的統計面試問題
測試 p 值解釋、顯著性和實際重要性、多重比較以及數據科學實驗背後的假設。
問題 1 的 8
Under the NIST definition, what does a p-value describe?
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。
- Under the NIST definition, what does a p-value describe?
- The size of the business impact caused by the treatment
- The probability of a test statistic at least as extreme as observed, assuming the null hypothesis is true
- The probability that the null hypothesis is true
- The probability the alternative hypothesis is true
- A candidate sees a small p-value. Which statement should they avoid?
- “The result is evaluated under a specified null hypothesis.”
- “The observed statistic would be unusual under the null model.”
- “I should also assess effect size and study assumptions.”
- “This is the probability that the null hypothesis is true.”
- Why should a significance threshold be chosen before examining results?
- To avoid selecting a decision rule after seeing which result looks favorable
- To ensure the measured effect is large enough to matter
- To remove the need for a sampling plan
- To guarantee the null hypothesis is false
- A team compares several outcomes and many pairs. What statistical issue should it consider?
- The smallest p-value is always the true effect
- Adding outcomes removes the need for a hypothesis
- Multiple comparisons can make repeated unadjusted testing misstate the overall error rate
- Each comparison automatically becomes independent evidence
- Which method can control an intended overall error level for planned comparisons?
- Replace the outcome with a post hoc metric
- Use an appropriate multiple-comparison procedure, such as Bonferroni when suitable
- Choose whichever comparison has the lowest p-value after analysis
- Ignore all comparisons except the one with the largest effect
- A result is statistically detectable but has a negligible effect on the product outcome. What should the analyst say?
- Statistical evidence and practical importance are different questions
- A small p-value proves the change is worth shipping
- Statistical significance measures revenue impact directly
- Effect size is irrelevant once a threshold is crossed
- Before interpreting an A/B test result, which design question matters?
- Which graph color makes the result look clearest
- Whether the reported p-value has enough decimal places
- Which team prefers the treatment
- How assignment occurred and whether groups or observations satisfy the analysis assumptions
- What can a non-significant result establish by itself?
- The treatment has no practical or statistical effect
- The experiment had enough power for every relevant effect
- It does not prove there is no effect; the result may be inconclusive
- The null hypothesis is certainly true