This is Jessica. Upon learning this morning that Lee Wilkinson passed away I also felt compelled to write something on the extent to which his work has influenced interactive visualization research.
The Grammar of Graphics was an incredibly ambitious undertaking – Wilkinson set out to create a system that could produce any statistical graphic he’d ever seen, and that could deepen understanding of the meaning of graphics. The GoG demonstrates the minimum set of components necessary to generate a statistical graphic, under an understanding that a graph is a function: data, algebra, scales, statistics, geometry, coordinates, and aesthetics. I often tell students I teach that in visualization research we hate chart taxonomies, and GoG is perhaps the best demonstration of how much more deeply we can think about visualization. Wilkinson pointed out the “deep structure” in visualization, observing, for instance, that a pie chart is just the result of passing rectangular marks through a polar transformation. He was inspired by Bertin’s work on graphical symbolism, and GoG systematizes thinking about the design space of visualizations in a way that is ultimately generative as well. You might be able to use an interactive implementation of it to make some crazy graphs but nothing that isn’t meaningful.
From what Wilkinson describes (e.g., in this recent podcast) some of the hard work behind the Grammar of Graphics was in the editing: making sure the system was complete and correct while also minimally complex. There are only three operators in the algebra – cross, nest, and blend – but they suffice. Tableau’s underlying table algebra and ggplot2 are examples of major components of the grammar used in today’s most popular visualization tools. At the same time, the book covers uncertainty, time, graph drawing, interactive control, and just about every other major branch of visualization research in some way, synthesizing important distinctions that would otherwise take someone a while to glean from the literature.
My own admiration for Grammar of Graphics is partly why I chose to get into visualization back as a grad student. I remember thinking his concept of a frame was really important but underappreciated in any discussions I’d heard about visualization. I read it for the first time as a Ph.D. student and have been calling it my favorite book for years. Whenever I go back to reread chapters I always come away with some new appreciation. I even bring in a copy to pass around in my interactive visualization course, trying to get students to sense its influence and hopefully read it. Just looking at the examples is like an education in visualization.
I didn’t know Lee well at all, but recall meeting him for the first time at the IEEE VIS doctoral colloquium, back in 2012. I remember he came in very late, but just in time for my presentation on uncertainty visualization, and his enthusiasm for the ideas (basically hypothetical outcome plots) was the highlight of my week. A few years later I remember talking to him at Tableau about the same ideas, and that time he was more critical, arguing that they would never catch on, but I appreciated that too. Lee has been critical of the visualization research community at times over the years, and while it was sometimes it was tough to hear, it was always clear he cared deeply and was optimistic about the field’s progress and open-minded intellectual attitude. His perspective will be missed.
Am i the odd one in liking native R graphics better than the ggplot metaphors?
I mean I like the ggplot looks just not the way you specify the visualisation. The syntax seems unnatural and just doesn’t seem to grow on me.
Somehow native R seems familiar having come from a gnuplot era. Just never really warmed up to the ggplot dplyr grammar.
I felt that way at first but have come to appreciate ggplot. Even after six years or so I still sometimes have to look up things that I feel should be easier, especially involving legends (the labels especially, but sometimes I want to change the placement or whatever). Actually I could name several things. But (1) for at least 95% of the graphics I want to make, ggplot will produce a better plot in less time than I would make using basic R, and (2) now that I’m used to the grammar/syntax almost all of it makes sense to me.
In short, I know what you mean and I would have agreed with you at one point, but no longer.
Thanks! That’s actually helpful to know.
So maybe I should just persist at it and will get over the learning curve.
+1
Personally, I’ve always found ggplot and tidyverse more intuitive than base R, though I nonetheless object to tidyverse’s frequent use of non-standard evaluation. But I see the point that to an end user, it’s really just “memorize syntax and commands of A vs syntax and commands of B”, and a preference for one or the other is pretty arbitrary. The real advantage comes when you build drastically different or all new visualizations; then, scales -> statistics -> geometry -> coordinate transformation provides a structure to your library that makes developing things enormously less complex. Otherwise, you’re just groping around in the darkness
Great post.
There is a typo or some words are missing in this sentence.
Tableau’s underlying table algebra and ggplot2 are examples of major components the grammar in today’s most popular visualization tools.