Earlier this month, we held the webinar, Learn the essentials of data visualization, with Stephanie Evergreen, a US-based expert on using research to present data effectively, and Andy Kirk, a UK-based apostle for better designed data visualization. In this one-hour webinar, Evergreen and Kirk addressed the basics of data design and demonstrated various ways for showing different types of data. The archive of their webinar appears above.
The webinar included a lively question and answer session with the audience, but there was not enough time to get to all of the questions so the speakers kindly provided follow-up answers here. A Storify of the Twitter activity that occurred during the webinar appears at the bottom of this post.
What is your favorite chart to represent survey results and why?
Stephanie: It totally depends on the survey question you asked and the story you are trying to tell within the results. That said, my current favorite way to show survey results is the diverging stacked bar. I love this one so much because it really emphasizes the split between desirable and less-than-desirable responses. My book has an entire chapter on how to visualize survey results with many more options for you.
There are different types of bar graphs. Which work best under what conditions?
Andy: Let’s look at the two major bar chart variations. The clustered bar chart has pairs of bars that help to present a convenient comparison between two different value series about a major category, let’s say median male vs. female for a range of different job types.
The stacked bar chart is about a part to whole breakdown showing the composition of several categories within a whole, e.g. the breakdown responses to a survey questions across a likert agree > disagree scale.
I find network graphs extremely difficult to understand, but they seem very popular, especially in academic circles to analyze relationships between people.
Andy: They are difficult to understand and often used without much sense of explanation to accompany them to help surface the key insights. However, often network based displays are actually trying to convey complexity, which is the intent. So the sense of being able to read their details is never really viable (unless the graph is interactive). Rather it’s about getting a gist of the major patterns (cluster, influences, gaps, outliers, etc.)
Are some of these presentation formats accepted by APA?
Stephanie: I studied the APA style guide (and others) for my first book (the MOST FUN part of writing – haha, no). Surprisingly, it doesn’t say much about chart types. While it does specify some formatting issues (like no more than 4 lines per line graph), there’s nothing in there preventing you from trying something new. The biggest obstacle is probably your department chair or the journal editor, who are used to seeing things in a certain way, regardless of how helpful those visuals may be.
Since journals generally publish in black-and-white and on paper, I would think symbols and numericals would be really critical to demonstrate. Is there any specific way to include symbols in data visualization?
Stephanie: I think by symbols you mean things like the markers on a line in a line graph. I’ve seen these used in the past, where each line has a different shape, in an attempt to distinguish the lines from one another. Please don’t do this. Such distinction is only necessary to try to help the reader connect the line to its related legend entry. It’s much easier for the reader to produce a cleaner looking graph that embeds the legend. Here’s a blog post on how to do so and I offer other strategies in my book.
How is Excel keeping up with your suggested graph types?
Andy: Excel still isn’t reflecting the best visualization practices. Each new version that seems to come out maintains the bad stuff (3D cone charts, awful defaults) and fails to catch up with the techniques in the contemporary field. That said, one can still overcome these settings and manually control a lot of the features to enhance Excel’s capability. For example, we can hack the charts and create new ways of showing data by thinking outside the box (such as using conditional formatting in worksheets to create heat maps/matrix diagrams). I would urge you to check out Jorge Camoes’ book ‘Data at Work’ as well as Stephanie’s to really learn how to make Excel sing a better tune. 🙂
You cited William S. Cleveland and Robert McGill’s 1984 article in the Journal of the American Statistical Association, “Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods.” Are there other academic readings you recommend?
Stephanie: Cleveland and McGill’s article is pretty legendary but yes, there’s so much more. Both of my books are pretty heavily referenced to point you to the research that backs up these choices.
Andy: With the release of my book there will be a dedicated sub-site on my site (visualisingdata.com) that will house lots of further reading/references including other key papers and research articles.
Do your books explain how to create the charts in particular software programs?
Stephanie: Andy’s book is a beautiful guide that provides a workflow on how to go about the visualizing process. My book hones in on how to choose the right chart type, given the considerations we discussed in the webinar, and then how to make it step-by-step inside Excel.
Beyond more traditional charts and graphs, any tips for using images, icons, other non-typical tools?
Andy: Direct tips are a little hard to express without direct examples to refer to but let me surmise that alternative media types and visual marks are all part of the palette for representation data, for representing information and for supplementing our traditional charts with other channels for conveying story and facilitating understanding.
Do you have any specific suggestions for better maps for analysis of geographic distribution?
Stephanie: I also said I’d send along some resources for qualitative visualization. If you recall, my biggest suggestion was to diagram it! Here’s a blog post that shows how this worked with one of my clients. You might also want to check out my qualitative viz Pinterest board. While much of this seems like stuff that takes a graphic designer, what ideas can you apply to your own work, maybe on a smaller scale?