Lecture 6: Deceptive visualizations

Lecture 6: Deceptive visualizations#

STATS 60 / STATS 160 / PSYCH 10

Recap#

Wednesday#

  • Data visualization (histograms, scatter plots, dot maps and choropleths).

  • Summaries of center (mean, median, mode).

Today#

  • Examples of bad and misleading data visualizations.

  • Quiz 2.

Line charts#

Household income and rent#

  • The following graph when viral on Twitter/TikTok, but it is misleading.

  • Let’s do a Fermi calculation based on the graph:

    • Suppose a household spent 20% of their income on rent in 1985.

    • Based on the graph, what percentage of their income would they be spending on housing in 2023?

Household income and rent#

  • In the previous graph, only income is adjusted for inflation.

  • Here is what the trend looks like when both are adjusted for inflation.

  • In general, alarming data visualizations are some of the ones to look out for.

  • Fortunately, the issue with the previous graph was noticed in the community notes.

Native and foreign born workers#

  • What is misleading about this line chart, original posted in the Financial Times?

  • Be careful of double axes!

Global CO2 emissions#

  • Which of these two graphics more accurately represents the data?

Gun deaths in Florida#

  • The line chart on the right shows an increase in gun deaths after Florida enacted its ‘Stand your ground’ law.

  • But the inverted y-axis makes it look like a decrease!

Bar charts#

The most read books#

  • It violates the principle of proportional ink!

  • Ironically, Business Insider has an article on bad charts!

Proportional ink#

  • There are many ways the principle of proportional ink can be violated:

    • Using 3D perspective.

    • Changing the x or y axes.

    • Using a ‘stylish’ shape.

Examples#

Proportional ink#

  • The principle of proportional ink also applies to other types of visualizations.

Scaling height and width#

  • Scaling both height and width also violates the principle of proportional ink as shown in this example from Wikipedia.

Scaling 3D objects#

  • The problem gets even worse with 3D objects!

  • What would happen to an object if you scaled its height, width and depth each by a factor of 3?

  • The object will increase in size by a factor of \(3 \times 3 \times 3 = 27\)!

Scatter plots#

Genetic scores and education attainment#

  • One great thing about scatter plots is that they show both trend and variation.

Genetic scores and education attainment#

  • Sometimes it is tempting to hide the variation.

Variation in scatter plots#

  • How does the difference between these two visualizations relate to our class on Wednesday about summaries of center?

  • The visualization on the right hides the variation. The center does not tell the whole story.

Ducks#

The big duck#

  • A data visualization that puts style over substance is called a “duck.”

The Big Duck in Southampton, Long Island

Ducks from USA today#

A duck in a scientific paper#

Good visualizations#

Data visualizations#

  • Despite all these examples, data visualizations remain the one of the best ways to explore, understand and communicate data.

  • There are many websites dedicated to producing and sharing high quality data visualizations:

People living in an extreme poverty#

The following graphic is from the article on progress towards ending extreme poverty from Our World in Data.

Interactive data visualizations#