CSE512 (Week1)

I am taking the cse512 from Washington University. The instructor is Jeffrey Heer, a regular contributor in the DataViz field. I decided to post here my progress, with the hope to get feedback and who knows, maybe motivate someone else to join me.

In the first week, we had a practical and reading assignments. See the class site for the book's reference. I should warn you, it is not an easy one to find. Yes, you can get it from Amazon and pay 100$ but I find that a bit excessive. You'll have more luck from your local University Library (UT in my case). I couldn't take the book home, but I was able to make digital copies of the first chapter. So much good stuff. If anyone out there has any interest, I can write up a blog post summarizing my notes.

It was a dense but interesting reading. I learned new terms that will I'll use the book as a reference in the future for sure.

The practical assignment involves creating a visualization for a dataset. A very interesting dataset actually. Read the assignment details here.

Design's description

The first thing you notice as you start playing with the data is the wide range of values you can have for the antibiotic's concentration to be effective. The table does a decent job communicating the data, but I think my viz allows users to consume the data with less cognitive effort.

We want to be able to pick one of the antibiotics, ideally the one that requires less concentration so we can save money without compromising effectiveness. We may not have the most efficient antibiotic and we may have to use the second best one.

I wanted to use distances for my visual encodings since those are processed faster by the human brain. But mapping the data directly will not work here because of the wide range of values in the variables. To solve that, instead of using linear mappings for my scales, I use logarithmic scales so we the separation of the values are more meaningful and in turn, help to communicate the data more effectively.


I am happy with the final result but there is room for improvement. I would separate the horizontal location of the three measurements so we make more clear collisions (two antibiotics that require the same concentration). If we do that, we could move the text's labels to the graph so we eliminate ink. I would definitely implement this if this was a real project.

I could add the units I am using on the axis.

I could add some color to the visual element that encodes the gram staining. I experimented with that but I found it hijacked some of the attention from the meat of the dataset, the MCIs.

Finally, there is plenty of room for improvement in the aesthetics department, although I don't think I'd work on that since it would not help make the visualization more effective, just more pretty.

The result