Statistics Communication & Stakeholder Relations 2 — Questions and Answers
Question 1: A statistician is presenting regression results to a non-technical board. Which approach best communicates a significant predictor?
- Report the p-value and t-statistic only
- State 'for every 1-unit increase in X, Y increases by B units on average' (Correct answer)
- Show the full covariance matrix
- Explain degrees of freedom in detail
Correct answer: State 'for every 1-unit increase in X, Y increases by B units on average'
Plain-language interpretation of the slope coefficient makes regression results actionable for non-technical audiences.
Question 2: A client insists that a correlation of 0.3 between advertising spend and sales 'proves' advertising causes sales. What is the statistician's best response?
- Agree, since p < 0.05
- Explain that correlation does not imply causation and suggest an experimental design (Correct answer)
- Increase the sample size to confirm causation
- Report Spearman's rho instead
Correct answer: Explain that correlation does not imply causation and suggest an experimental design
Correlation only measures linear association; establishing causation requires controlled experimentation or causal inference methods.
Question 3: Which visualization is most appropriate when presenting the distribution of a continuous variable to stakeholders unfamiliar with statistics?
- A QQ-plot
- A histogram or density plot (Correct answer)
- A correlation matrix heatmap
- A residual plot
Correct answer: A histogram or density plot
Histograms and density plots intuitively show shape, center, and spread without requiring statistical background knowledge.
Question 4: A project manager asks for a '95% confidence interval' explanation. Which analogy is most accurate?
- There is a 95% probability the true parameter is in this specific interval
- If we repeated the study many times, about 95% of such intervals would contain the true parameter (Correct answer)
- The sample statistic has a 95% chance of being correct
- 95% of the data fall within this range
Correct answer: If we repeated the study many times, about 95% of such intervals would contain the true parameter
A confidence interval is a procedure property: 95% of intervals constructed this way capture the true parameter across repeated samples.
Question 5: A stakeholder asks why you used a log transformation on the outcome variable. What is the clearest justification?
- It makes the dataset larger
- It stabilizes variance and addresses right skew, improving model assumptions (Correct answer)
- It removes all outliers automatically
- It converts the variable to a percentage
Correct answer: It stabilizes variance and addresses right skew, improving model assumptions
Log transformations are applied to reduce heteroscedasticity and normalize right-skewed distributions to satisfy regression assumptions.
Question 6: When reporting statistical results in an executive summary, which element should always be included alongside a p-value?
- The full R syntax used
- An effect size measure (Correct answer)
- Raw data tables
- The eigenvalues of the model
Correct answer: An effect size measure
Effect sizes (e.g., Cohen's d, R²) convey practical significance, since p-values alone do not indicate magnitude of an effect.
Question 7: A department head questions why sample results differ from a known population value. Which concept best explains this to them?
- Measurement bias
- Sampling variability (sampling error) (Correct answer)
- Nonresponse bias
- Type II error
Correct answer: Sampling variability (sampling error)
Sampling variability is the natural fluctuation in statistics across different samples drawn from the same population.
Statistics Communication & Stakeholder Relations 2
A statistician is presenting regression results to a non-technical board.
Which approach best communicates a significant predictor?