# Analytic Visualization and the Pictures in Our Head

In 1922, American journalist and political philosopher Walter Lippmann wrote about the “pictures in our head,” arguing that we conceptualize distant lands and experiences beyond our own through a mental image we create. He coined the word “stereotypes” to describe these mental pictures, launching a field of political science focused on how people form, maintain, and change judgements.

While visual analytics is far from the study of stereotypes, in some ways it relies on the same phenomenon. As described in Illuminating the Path, edited by James J. Thomas and Kristin A. Cook, there is an “innate connection among vision, visualization, and our reasoning processes.” Therefore, they argue, the full exercise of reason requires “visual metaphors” which “create visual representations that instantly convey the important content of information.”

F. J. Anscombe’s 1973 article Graphs in Statistical Analysis makes a similar argument. While we are often taught that “performing intricate calculations is virtuous, whereas actually looking at the data is cheating,” Anscombe elegantly illustrates the importance of visual representation through his now-famous Anscombe’s Quartet. These four data sets all have the same statistical measures when considered as a linear regression, but the visual plots quickly illustrate their differences. In some ways, Anscombe’s argument perfectly reinforces Lippmann’s argument from five decades before: it’s not precisely problematic to  have a mental image of something; but problems arise when the “picture in your head” does not match the picture in reality.

As Anscombe argues, “in practice, we do not know that the theoretical description [linear regression] is correct, we should generally suspect that it is not, and we cannot therefore heave a sigh of relief when the regression calculation has been made, knowing that statistical justice has been done.”

Running a linear regression is not enough. The results of a linear regression are only meaningful if the data actually fit a linear model. The best and fastest way to check this is to actually observe the data; to visualize it to see if it fits the “picture in your head” of linear regression.

While Anscombe had to argue for the value of visualizing data in 1973, the practice has now become a robust and growing field. With the rise of data journalism, numerous academic conferences, and a growing focus on visualization as storytelling, even a quiet year for visualization – such as 2014 – was not a “bad year for information visualization” according to Robert Kosara, Senior Research Scientist at Tableau Software.

And Kosara finds even more hope for the future. With emerging technologies and a renewed academic focus on developing theory, Kosara writes, “I think 2015 and beyond will be even better.”

This site uses Akismet to reduce spam. Learn how your comment data is processed.