Training Data Visualization Misleading Graphs & Design Principles
4 / 10

Misleading Graphs & Design Principles

23 min Data Visualization

Misleading Graphs & Design Principles

Common Misleading Techniques
  • Truncated y-axis: starting the axis above zero exaggerates differences
  • Distorted scales: unequal intervals on an axis
  • Cherry-picked range: showing only part of the data to support a narrative
  • 3-D effects: perspective distortion makes comparisons unreliable
  • Dual axes: two different y-scales can imply false correlations
  • Area/volume encoding: humans misjudge areas (doubling radius quadruples area)
Tufte's Principles
  • Maximize data-ink ratio
  • Avoid chart junk and non-data ink
  • Show data variation, not design variation
  • Use labels directly on the graphic when possible
  • Small multiples over one complex chart
Lie Factor

$$\text{Lie Factor} = \frac{\text{Size of effect shown in graphic}}{\text{Size of effect in data}}$$

Ideally = 1. Values far from 1 indicate distortion.

Example 1

A bar chart shows profits of $\$100M$ vs. $\$105M$ but the y-axis starts at $\$98M$. Why is this misleading?

The 5% increase looks like a 70% increase visually. Starting at 0 would show the true proportion.

Example 2

Sales doubled, but a graphic uses icons where the taller icon is 2× the height and 2× the width. Lie factor?

Area quadrupled (2×2). Lie factor $= 4/2 = 2$. The graphic exaggerates the change by a factor of 2.

Example 3

A graph shows temperature rising from 14.0°C to 14.8°C over 100 years with a steep line. Is it misleading?

Depends on context. A truncated y-axis exaggerates the visual trend. Show the full scale or clearly label the narrow range.

Practice Problems

1. What is chart junk?
2. Should a y-axis always start at zero?
3. Why are 3-D pie charts problematic?
4. A coin icon doubles in diameter to show 2× value. What's the lie factor?
5. What is a dual-axis chart, and when is it dangerous?
6. What are small multiples?
7. How can color mislead in a map?
8. Why label data directly instead of using a legend?
9. What does a log scale help visualize?
10. Is it ethical to truncate an axis?
11. Name three elements of good graph design.
12. How would you fix a misleading bar chart with a truncated axis?
Show Answer Key

1. Decorative, non-data visual elements that distract

2. For bar charts yes; for line charts showing small changes, a truncated axis with clear labeling can be acceptable

3. Perspective makes front slices look larger than back slices

4. Area $\propto r^2$; visual effect $= 4\times$, data effect $= 2\times$. Lie factor $= 2$.

5. Two different y-scales; can make unrelated series appear correlated, scales can be manipulated

6. Repeating the same chart structure with different subsets of data for comparison

7. Poor color scales (rainbow) can create false boundaries; sequential scales are better for continuous data

8. Reduces eye movement and cognitive load

9. Growth rates, data spanning several orders of magnitude

10. Only if clearly indicated and if zero isn't meaningful for the context

11. Clear title, labeled axes, source attribution (also: consistent scale, minimal chart junk)

12. Start the y-axis at 0, or use an axis break with clear notation

Scatter Plots, Correlation & Trend Lines MATLAB — Numerical Computing & Visualization