The words 'business' and 'data visualization' probably put you to sleep before you even reach the end of this sentence. But wake up! There's actually a lot of interesting work to be done in this area, if only we give it a chance.
Computational notebooks offer an alternative to the common GUI-based tools used for data visualization and BI today. In this new paper, I talk about what they are, their pros and cons, and how research could fill in some important gaps.
Encodings play a central role in visualization, but I believe our thinking about them is too simplistic. In a new paper, I argue that we need to distinguish between the encodings that specify how a visualization is drawn and the ones that are readable or actually read by an observer. While they largely or entirely overlap in some charts (like bar charts or scatterplots) they don't in others (pie charts, line charts, etc.). And what exactly do you even specify in more complex visualizations like treemaps?
Who are the people who use data and visualization as part of their work every day? In particular, how do people use data as part of meetings, to present to others, to discuss their findings and recommendations, etc.? My colleague Matt Brehmer and I ran a pair of studies to find out.
In 2016, Steve Haroz, Steven Franconeri, and I published a paper on a technique commonly called the Connected Scatterplot. It turns out that somebody else had research on essentially the same chart 15 years earlier, which we were not aware of. Our work is quite different, but it’s interesting context and it’s also worth reflecting on how we missed this piece of relevant prior work.
How we read pie charts is still an open question: is it angle? Is it area? Is it arc length? In a study I'm presenting as a short paper at the IEEE VIS conference in Vancouver next week, I tried to tease the visual cues apart – using modeling and 3D pie charts.
Why do pie charts look the way they do? What makes this particular way of slicing up a circle the preferred way of showing part-to-whole relationships? In two short papers that I’m presenting this week at EuroVis, I looked at the design space of circular part-to-whole charts, including pie charts.
Visualization doesn't have the replication issues that some other fields are struggling with right now, but is that because our science is so strong or because nobody actually bothers with replications? And what can we do to get ahead of potential problems before we run into a full-on crisis? In a paper to be presented at BELIV, Steve Haroz and I list potential pitfalls and present possible solutions.
There is talk about stories having a beginning, middle, and end, but what does that mean for data stories? How do you create the overall structure for those? In a paper to be presented at EuroVis next week, I discuss a simple but very useful structure that I have found "in the wild," and that I believe to be useful and generalizable.
Embellished charts are common in information graphics. But are embellished bar charts actually harder to read than plain ones? In a short paper to be presented at EuroVis next week, Drew Skau and I present a study that looked into this question.
How should you sequence information in a data story so it makes the most sense? Are some sequences better than others? Does time have to move forward or does it not matter? In a paper to be published at EuroVis next week, with Jessica Hullman at UW and my Tableau Research colleague Heidi Lam, we report on a pair of studies that looked into this.
The two main conferences in visualization are VIS and EuroVis, but recently CHI has also gotten some very interesting submissions (CHI is technically a conference about human-computer interaction, or HCI). This year looked particularly strong,
It's not a secret that I think that we need to ask some harder questions about the foundations that we're building visualization on. In a paper to be presented at the BELIV workshop at VIS next week, I'm making the case for that more extensively than I have so far. The full title of the paper is An Empire Built On Sand: Reexamining What We Think We Know About Visualization.
How do we read pie charts? Do they differ from the even more reviled donut charts? What about common pie chart designs like exploded pies? In two papers to be presented at EuroVis next week, Drew Skau and I show that the common wisdom about how we read these charts (by angle) is almost certainly wrong, and that things are much more complicated than we thought.
The worker IDs Amazon's Mechanical Turk gives you may look pretty random and anonymous, but they can reveal personally-identifiable information. They need to be removed from datasets, especially when they are shared or published.
We've turned the understanding of charts into formulas instead of encouraging people to think and ask questions. That doesn't produce better charts, it just gives people ways of feeling superior by parroting something about chart junk or 3D being bad. There is little to no research to back these things up.
Presentation is often considered a part of visualization, but what does that mean for the kinds of techniques we use? Are they the same as used for analysis? What criteria should we use to pick them? In a new paper, I discuss a class of techniques I call presentation-only.
Stephen Few has written a long and scathing piece tearing down a paper presented at VIS earlier this year. While some of his criticism is justified, it is too focused on one of the authors, and it comes from an idea of visualization research that is far too limited.
I’m very happy to finally be able to announce our paper on the connected scatterplot technique. It describes the technique, provides some historical perspective, and most of all looks into how easy to understand and engaging the technique actually is.
My colleagues at Tableau Research have three papers at InfoVis next week. They cover guided data exploration, color theory, and data partitioning. Here's a little preview.
Information graphics often use variations and embellishments of standard charts that may distort the way people read the data. But how bad are these distortions really? In a paper to be presented at EuroVis this week, Drew Skau, Lane Harrison, and I tested their effects in an experiment.
Unit charts are not common in visualization, and they are often considered a bad idea. The same is true for using shapes other than rectangles. Neither is based on much actual research, however. In a new paper, we look at the specific example of ISOTYPE-style charts – and find them to be quite effective.
A paper on a specific cognitive mechanism may seems like an odd choice as the first paper in this series, but it is the one that sparked the idea for it. It is also the one that has its 30th birthday this year, having been published in August 1985. And it is an important paper, and could play an even bigger role in visualization if properly understood and used.
To explore how we can make it easier to create new visualization designs, we are running a study based on a new approach, called visualization primitives. It lets you map data to the properties of objects like rectangles and ellipses. Build something with data, have fun, and help us figure out if it works!
Visualization is often considered to consist of three phases: exploration, analysis, and presentation. While the former two topics are covered well in the literature, there has been very little work specifically on presentation. In an upcoming paper, Jock Mackinlay and I argue that presentation, and in particular storytelling and communication of data, are the logical next step for the field, and provide some research directions.
After posting my response to Gelman and Unwin's article on visualization and statistical graphics recently, I have now collected links to all the three other responses as well: Stephen Few, Paul Murrell, and Hadley Wickham.
This is how scholarly exchanges used to work: Scientist A publishes a result, Scientist B then writes an angry letter saying that Scientist A is full of it, to which A responds with more insults, etc., and all that published in a fine scholarly journal. I was recently asked to respond to a piece Andrew Gelman and Anthony Unwin had written about visualization for the Journal of Computational and Graphical Statistics, which had some issues.
Visualization is largely defined as the transformation of data into images. Visualization tools don't have a way of assessing their output, though: were there enough pixels to represent all the data? Are there too many overlapping lines? In a paper to be presented at EuroVis next week, Aritra Dasgupta, Min Chen, and I propose a taxonomy of the different sources of uncertainty when working with parallel coordinates.
The point of visualization is usually to reveal as much of the structure of a dataset as possible. But what if the data is sensitive or proprietary, and the person doing the analysis is not supposed to be able to know everything about it? In a paper to be presented next week at InfoVis, my Ph.D. student Aritra Dasgupta and I describe the issues involved in privacy-preserving visualization, and propose a variation of parallel coordinates that controls the amount of information shown to the user.
Can gravity have an influence on how the data in a chart is perceived? How do different kinds of connections between circles change our perception of the distance between them? And what does that mean for how strongly we perceive them to be connected? We conducted some user studies to find out.
Parallel coordinates are a very popular visualization technique for multi-dimensional numerical data. In this paper, we propose a set of metrics to better understand the types of visual structures users commonly look for using this technique. Based on the metrics, we can optimize the display to make it more readable, and allow the user to select dimensions based on their visual structures, rather than their existing ideas about the data.
Visualization needs a new theory. Bertin's ideas about marks and retinal variables have provided a great starting point, but we are now seeing their limitations. We need to turn a new page and move beyond those cosy, familiar ideas, into new territory. A recent paper by Caroline Ziemkiewicz and myself makes an argument why, and provides some possible directions.
Interaction in visualization is incredibly important, but often more tedious than it needs to be. I have developed a new way of brushing in parallel coordinates that uses the multi-touch trackpads on Apple's MacBook and MacBook Pro laptops for faster interaction. The video below demonstrates the technique, and the source code is available.
Design is usually considered a minor point in visualization. Does it make a difference what color scheme you use (as long as it's not an atrocious one), how thick your lines are, whether you put a background behind your chart, etc.? Caroline Ziemkiewicz and I presented a paper at Advanced Visual Interfaces (AVI) where we reported on a study we had performed to find out.
There is almost universal agreement that any extraneous elements in a chart or visualization, elements that do not represent numbers, are detrimental to understanding the data. A paper that was presented at CHI recently described a study to figure out just how bad all this chart junk really was. As it turns out, it's actually rather helpful.
User studies are an important part of visualization, but they also require a considerable amount of effort and time. In a paper presented at the BELIV workshop (part of CHI 2010), we discussed our experiences with running a number of visualization studies using Amazon's Mechanical Turk (MTurk) service. Using MTurk, we are able to run large studies in much less time than usual, and at very low cost. We also show how to avoid gaming the system, which had been reported in earlier work using MTurk.
Venn diagrams are a strange mix of structure and data visualization. In my latest Sightings column (PDF) for American Scientist, I use the example of a visualization challenge from last year to discuss different ways to show the same data about diagnosis techniques for autism in young children. This also sparked the launch of a new site feature: Ask Eagereyes.
In January, my Ph.D. student Caroline Ziemkiewicz told me about an interesting observation she had made: in different papers comparing tree visualizations, treemaps came out as best, worst, or somewhere in the middle. One difference she noticed was how the questions were worded: when a levels metaphor was used, treemaps did badly; a containment metaphor, on the other hand, seemed to favor treemaps. So we decided to investigate – the result will be presented at InfoVis on Monday, October 20.
How do bats fly? What are the aerodynamic conditions around their wings? And how do you visualize all that? I did a short interview with David Laidlaw (PDF), who has collaborated with physicists, biologists, fluid mechanics experts, and others, to create a poster that won last year's NSF Visualization Challenge. The interview was done for American Scientist's Sightings column, which I have been invited to write.
A paper on visualization criticism just appeared in the Visualization Viewpoints section of this month's Computer Graphics and Applications (CG&A). Authors are yours truly, Fritz Drury, Lars Erik Holmquist, and David Laidlaw.
