A Response to Flowing Data

This morning, after being out of the office for a long weekend because of NATO in Chicago, I checked traffic on this blog. It was far higher than usual, most of it coming from this post by Kim Rees at Flowing Data, who was very condescending about my post on visualizing ACT Scores.

I like good debate about data visualization. About almost anything, really. But this seemed a bit off, for several reasons:

It appears Ms. Rees didn’t read the blog post that accompanied the chart. Had she understood the intent of the graphic–to tell a story rather than to provide precise comparisons among groups–she might have not had anything to write about. Fortunately, most of the comments on the site pointed this out, even though not everyone thought the graphic was perfect. I actually join them in acknowledging this.
She didn’t give me a chance to respond prior to “picking on” me, something she says she does not like to do right before doing it.
Her suggestions for alternatives (a violin chart and a bean chart) would not even come close to doing what this graphic does. In fact, I’ve been interested in visualizing data for several years, and I can’t even begin to tell you how to interpret these charts. In addition, they’re ugly, as are most of the charts on the site she points to.

But in the interest of doing better, I took another stab at it. I used Tableau, of course. Here’s one take on it:

I’ve broken the data into four large groups by ethnicity and then into smaller groups by self-reported income. The length of the bar shows the number of records; the color of the bar shows the average ACT of the students in that group (dark gold is low, dark purple is high), as does the label. Better?

This, of course, fails to show the distribution; you get a sense of it, but it does give you some greater detail. What about distributions? Try this, and see if it makes the point, even though it’s impossible to compare the number of records in each group, even if you were to label them. These are grouped into a smaller number of income bands for clarity, and show stacked, 100% bars colored by groups of ACT Scores. Labels show individual scores, which are grouped by color for visual impression.