Statistics Communication & Stakeholder Relations 4 — Questions and Answers
Question 1: A statistician needs to explain Type I and Type II errors to a clinical trial sponsor. Which real-world framing is most effective?
- Define alpha and beta mathematically
- Type I: approving an ineffective drug; Type II: rejecting an effective drug (Correct answer)
- Explain the F-distribution
- Describe power curves in detail
Correct answer: Type I: approving an ineffective drug; Type II: rejecting an effective drug
Grounding abstract error types in domain-specific consequences (approving/rejecting drugs) makes them concrete and relevant for stakeholders.
Question 2: A finance team requests weekly statistical reports but rarely reads them. What is the best communication strategy?
- Continue sending full technical reports
- Switch to a one-page executive dashboard highlighting key metrics and anomalies (Correct answer)
- Stop producing reports
- Add more statistical detail to increase perceived value
Correct answer: Switch to a one-page executive dashboard highlighting key metrics and anomalies
Condensed dashboards that surface key insights and exceptions are more likely to be consumed and acted upon by busy stakeholders.
Question 3: What does 'statistical significance' fail to communicate that stakeholders often most need to know?
- The sample size used
- The practical or economic magnitude of the effect (Correct answer)
- Whether normal distribution was assumed
- The confidence level chosen
Correct answer: The practical or economic magnitude of the effect
Statistical significance only addresses whether an effect is distinguishable from zero, not whether it is large enough to matter practically.
Question 4: A city planner asks whether a survey of 500 residents can represent a city of 2 million. What is the key concept to explain?
- The law of large numbers requires millions of respondents
- With probability sampling, a well-designed sample of 500 can produce reliable estimates regardless of population size (Correct answer)
- Only a census can represent a large city
- 500 is too small for any inference
Correct answer: With probability sampling, a well-designed sample of 500 can produce reliable estimates regardless of population size
Statistical precision depends more on sample size than population size when probability sampling is used correctly.
Question 5: A stakeholder is upset that the model's predictions were wrong in one notable case. How should you frame model uncertainty?
- Admit the model is flawed and should be discarded
- Explain that probabilistic models express uncertainty across many cases, not certainty for individual ones (Correct answer)
- Promise to retrain the model on that single case
- Add more predictor variables immediately
Correct answer: Explain that probabilistic models express uncertainty across many cases, not certainty for individual ones
Statistical models provide probabilistic predictions over populations; individual outcomes naturally vary and a single miss does not invalidate a well-calibrated model.
Question 6: Which principle should guide how much statistical detail to include in a stakeholder report?
- Always include all technical appendices in the main body
- Match detail level to what the audience needs to make the specific decision at hand (Correct answer)
- Include as much as possible to demonstrate rigor
- Omit all methodology to keep it brief
Correct answer: Match detail level to what the audience needs to make the specific decision at hand
Effective statistical communication is audience-centered: include enough detail to support the decision, with full methodology in supplementary materials.
Question 7: A nonprofit director asks why your survey results have a 'margin of error of ±4%.' How do you explain this?
- It means the survey was 4% inaccurate
- It means the true population value likely falls within 4 percentage points of the sample estimate, based on sampling variability (Correct answer)
- It measures response bias
- It reflects the error rate of the survey software
Correct answer: It means the true population value likely falls within 4 percentage points of the sample estimate, based on sampling variability
Margin of error quantifies sampling variability: the range within which the true population value is expected to fall with the specified confidence level.
Statistics Communication & Stakeholder Relations 4
A statistician needs to explain Type I and Type II errors to a clinical trial sponsor.
Which real-world framing is most effective?