MS-DS Master of Data science Data Visualization and Communication 2 — Questions and Answers
Question 1: Which color encoding principle states that hue should be used for nominal data while luminance/saturation should encode ordered data?
- Gestalt principle
- Bertin's retinal variables principle (Correct answer)
- Tufte's data-ink ratio
- Cleveland's graphical perception hierarchy
Correct answer: Bertin's retinal variables principle
Jacques Bertin established that hue differentiates categories (nominal), while value/lightness encodes order or quantity (ordinal/quantitative).
Question 2: A dashboard shows a KPI card with a large number and no trend context. What is the primary communication failure?
- Overuse of color
- Absence of a reference baseline or comparison (Correct answer)
- Too much whitespace
- Using a sans-serif font
Correct answer: Absence of a reference baseline or comparison
A single number without a baseline, target, or historical trend gives the viewer no way to judge whether the value is good or bad.
Question 3: When visualizing a distribution with heavy outliers, which chart type preserves the most information without distortion?
- Histogram with equal-width bins
- Box-and-whisker plot (Correct answer)
- Bar chart of mean ± SD
- Pie chart of quartile segments
Correct answer: Box-and-whisker plot
A box-and-whisker plot explicitly shows median, IQR, whiskers, and individual outlier points without collapsing the distribution into a single mean.
Question 4: The 'lie factor' metric introduced by Edward Tufte measures what?
- The ratio of false-color pixels to total pixels in a chart
- The ratio of the effect size shown in the graphic to the effect size in the data (Correct answer)
- The number of chartjunk elements per square inch
- The percentage of audience members who misread the chart
Correct answer: The ratio of the effect size shown in the graphic to the effect size in the data
Tufte defined Lie Factor = (size of effect shown in graphic) / (size of effect in data); a value far from 1.0 indicates visual distortion.
Question 5: In a connected scatterplot (path chart), what does the path between points encode that a standard scatterplot omits?
- Correlation strength
- Temporal sequence or ordering (Correct answer)
- Cluster membership
- Statistical significance
Correct answer: Temporal sequence or ordering
The path connects observations in order (usually time), revealing how two variables co-evolved — information a static scatterplot cannot convey.
Question 6: Which technique is most appropriate for visualizing the joint distribution of two continuous variables when the dataset contains 500,000 points?
- Standard scatterplot with transparency
- 2D kernel density estimation (contour plot) or hexbin plot (Correct answer)
- Grouped bar chart
- Parallel coordinates
Correct answer: 2D kernel density estimation (contour plot) or hexbin plot
At large scale, overplotting makes scatterplots unreadable; hexbin or KDE contour plots aggregate point density into readable cells or iso-lines.
Question 7: A presenter wants to compare part-to-whole relationships across five product categories simultaneously. Which chart type is most appropriate?
- Stacked area chart
- 100% stacked bar chart (Correct answer)
- Bubble chart
- Scatter matrix
Correct answer: 100% stacked bar chart
A 100% stacked bar chart normalizes each bar to 100%, making it easy to compare proportional compositions across categories.
MS-DS Master of Data science Data Visualization and Communication 2
Which color encoding principle states that hue should be used for nominal data while luminance/saturation should encode ordered data?