Statistics Communication & Stakeholder Relations 3 — Questions and Answers
Question 1: A stakeholder wants you to remove 'inconvenient' outliers to make the trend line look cleaner. What is the ethical response?
- Remove outliers if they fall beyond 2 standard deviations
- Investigate outliers for data quality issues but report results both with and without them transparently (Correct answer)
- Delete them since they distort the mean
- Replace them with the median value silently
Correct answer: Investigate outliers for data quality issues but report results both with and without them transparently
Ethical practice requires investigating outliers, documenting decisions, and reporting sensitivity analyses rather than silently manipulating data.
Question 2: Which format is most effective for communicating uncertainty to a general audience in a written report?
- Standard error tables
- Phrases like 'we estimate X, with a plausible range of Y to Z' (Correct answer)
- Variance-covariance matrices
- Z-score distributions
Correct answer: Phrases like 'we estimate X, with a plausible range of Y to Z'
Pairing a point estimate with a plain-language range communicates uncertainty accessibly without requiring statistical literacy.
Question 3: A policy analyst asks which statistical test was used and why. What should the statistician's answer address?
- Only the software package used
- The data type, distributional assumptions, and research question that guided test selection (Correct answer)
- The p-value threshold chosen
- The number of iterations in the simulation
Correct answer: The data type, distributional assumptions, and research question that guided test selection
Test selection rationale should cover data characteristics, assumptions, and how the test addresses the specific inferential question.
Question 4: A stakeholder is confused why two analysts got different results from the same dataset. Which explanation is most likely correct?
- One analyst made arithmetic errors
- They used different analytical methods, assumptions, or preprocessing steps (Correct answer)
- The dataset changed between analyses
- Statistical software always produces random results
Correct answer: They used different analytical methods, assumptions, or preprocessing steps
Different modeling choices, variable coding, or handling of missing data can legitimately produce different results from the same raw data.
Question 5: What is the primary risk of presenting only statistically significant findings to stakeholders?
- It reduces the length of the report
- It creates publication bias and a misleading picture by omitting null results (Correct answer)
- It makes the analysis easier to replicate
- It violates ANOVA assumptions
Correct answer: It creates publication bias and a misleading picture by omitting null results
Selective reporting of significant results inflates apparent effect sizes and misrepresents the overall evidence, a form of publication bias.
Question 6: A marketing team asks if an A/B test result is 'good enough to act on.' What statistical concept should guide your answer?
- The Bonferroni correction value
- The combination of statistical significance, effect size, and practical business impact (Correct answer)
- Whether the sample was randomly selected
- The variance inflation factor
Correct answer: The combination of statistical significance, effect size, and practical business impact
Actionability depends on significance (low false-positive risk), effect size (meaningful magnitude), and whether the impact justifies implementation cost.
Question 7: When a stakeholder says 'the data speaks for itself,' a statistician should clarify that:
- All datasets are self-explanatory
- Analytical choices, assumptions, and context always shape how data is interpreted (Correct answer)
- Only biased data requires interpretation
- Visualizations eliminate subjectivity
Correct answer: Analytical choices, assumptions, and context always shape how data is interpreted
Data interpretation always involves choices about methods, assumptions, and framing, making critical evaluation of those choices essential.
Statistics Communication & Stakeholder Relations 3
A stakeholder wants you to remove 'inconvenient' outliers to make the trend line look cleaner.
What is the ethical response?