Follow

5/12/17: Data is good, when we know how to use it.

But we savvy folks can be duped, too, when we’re not on the lookout for “visualization lies” that are presented without us viewing them critically.

Nathan Yau, a statistician with a PhD from UCLA, provides a beautiful overview of how we can view visual data critically, at FlowingData.com. In his post, “How to Spot Visualization Lies,” Yau provides multiple examples to improve our graph-savvy approach, so that we are not potentially misled by the data we view.

In the graph below, Yau illustrates the “Truncated Axis” violation. The graphs below depict the same data, but the one on the left visually shows a greater disparity between the items on the y-axis. This could be misleading, and should be viewed critically. As Yau efficiently explains, “Bar charts use length as their visual cue, so when someone makes the length shorter…the chart dramatizes differences.”

The adage “correlation is not causation” could never be more accurate today. Many intentionally use charts to confuse we average folks to infer that correlated items link causation. A great example of this fallacy is provided by the ice cream sales and crime rates. When ice cream sales rise, so do crime rates (Peters, 2013), but are we to assume ice cream sales cause crime rates to surge? Not likely. Rather, as Peters details, these 2 events are correlated by a third, unseen factor: warmer temperatures.

The Dual Axes chart uses 2 different scales to depict information, and could be interpreted to imply causation if we are not vigilant.

Pictures like the one shown below create a view of “absolutes.” We must consider relative values when viewing this data. Everything is relative: “you can’t say a town is more dangerous than another because the first one had two robberies and the other only had one. What is the first town has 1,000 times the population than the first?”

This is echoed fiercely in Steven Pinker’s tome, The Better Angels of our Nature. Pinker explains the commonly held belief of the high occurrence of homicide in the last 2 centuries is an absolute fallacy. Because of how violent crime is reported, and how often we read about it, many have the view that our situation is chronic. Homicide rates in 5 Western European regions in the 1200’s were 100 to of 100,000 people per year. At the start of 1900, they were less than 1 per 100,000.

Graphs depicting limited scope may unfairly illustrate anomalies, but we need to see this data with history. Is this a pattern?

Yau provides several other examples in his well-written overview.

Obviously, data visualization is a critical way we make sense of the world. We need to learn both how to visualize our data, and how to consume it in a way that makes us smarter, and informs our decisions in positive ways.

Yau concludes, “A chart doesn’t make something true. Data doesn’t make something true. It bends. It shows many things. So keep your eyes open.”

-Tonya Riney, PhD

VP, Client Services, IntelliBoard

 

 

 

Have more questions? Submit a request
Powered by Zendesk