Visualizing Data

Welcome to my data visualization portfolio.  As an engineering professor and academic for 35 years, I made 100’s of graphics to help visualize data for presentation in archival journal articles, conference presentation, grant proposals, and classroom demonstrations.  None of these were particularly interesting because, in that setting, it is best to stick to formats familiar to the audience.  You don’t want to confuse a paper or proposal review or make them do extra work, and you don’t want to confuse the audience.

After leaving Penn State, I decided to spend some of my free time refining my skills at visualizing data sets with the freedom of being creative and using new methods and formats to present data.  While I have no formal training in data analysis, data science, or data visualization, I have spent decades thinking about how to best convey information to an audience.  I have also spent decades using Wolfram Research’s Mathematica software for my teaching and research.  I have discovered that it is also flexible enough to allow me to be creative and easily implement new ideas.

I encourage your feedback, but please offer suggestions for ways to improve something instead of just saying, “This sucks,” or “This is terrible.”  That kind of criticism is not at all constructive.  Also, be away that when you’re designing things, albeit an automobile, a Mars rover, or a data visualization, there are always going to be trade-offs.  I will often have an exchange like the following with someone:

Putting the bars for each state next to each other so you can read them both at the same time without scrolling to the other side. You could even superimpose them. I have to zoom in so far to read the numbers, I can’t see the the state they correspond to.

  • Putting the the bars for each state next to each other makes it harder to compare states, which is what interested me. Superimposing them doesn’t work for a variety of reasons.
  • I probably should put the numbers at the base of each bar. The precise numbers really aren’t that important in my opinion.

Using a scale that actuality fits all the data.

  • If I scale to AK and HI then everything else would get squashed and be harder to compare. I probably should have just left them off but then people would complain about that.

Spacing them so it’s easier to read.

  • I have no idea how I’d do that.  Got a suggestion?

What the hell do the big numbers at the top even mean? Hundreds of miles? Some labels (in a reasonable place that obviously corespond to what they should) would be nice.

  • The numbers at the top mean the same as the numbers at the bottom. Labeling them twice is redundant.  I included the number to help the viewer.  Maybe I should just have left them off.
  • There are trade offs to all those suggestions. I had considered most of them when I made this and picked the option I thought was best. Had I done it exactly as you suggest someone else (and probably you too) would have complained.

Then people get upset that I’m dismissing their feedback.  That’s not it.  I’m simply explaining why I made the trade-offs I did.  When people have never been through the design process of a certain visualization and been forced to make trade-offs, they don’t really understand.  I tried almost every one of the suggestions offered above and chose not to go that way for one reason or another.  I am constantly reminded of the sayings:

Those that can’t play coach.  Those that can’t coach sit in the bleachers yelling obscenities at the players and coaches.

Too much high-gain feedback leads to an unstable system.