Math – Applications for Living XVI: The Evils of Line Charts

In our Math – Applications for Living class, we just finished our first chapter on statistics, which included some of the standard graphical displays (frequency charts, bar graphs,  and line charts) as well as scatter diagrams for bivariate data.  Based on observing students working with line charts, I commented to the class that line charts have a risk — human perception may suggest a much stronger pattern than is really present in the data.

Here is an arbitrary example of a line chart (pseudo-random data):

 Join Dev Math Revival on Facebook:

The same data presented in a bar graph looks like this:

A quick search on research related to this yielded a study done by Jeff Zacks and Barbara Taversky at Stanford (1999); see      http://dcl.wustl.edu/pubs/ZacksMemCog99.pdf.  Briefly, their study found that bars tended to processed discretely while lines tended to be processed as a pattern.  My examples above are not ideal examples … the correlation is fairly high, perhaps higher than the bar graphs suggest. 

It seems that scatter diagrams have some of the same risk as line charts, in the sense that my students want to visually connect the dots as in a line chart.  We are working on identifying patterns correctly in bivariate data, though my expectations are not high.  Human perception might have a strong desire to establish a pattern, even when some data needs to be excluded.

No Comments

No comments yet.

RSS feed for comments on this post. TrackBack URI

Leave a comment

You must be logged in to post a comment.