Envisioning a syllabus on the Display of Quantitative Information
February 14, 2012 12:36 PM Subscribe
I am planning on teaching a course on data visualization to some Ph.D. students in the social sciences, and could use some ideas about designing the syllabus. If you have taken or taught such a course, I would especially value your input, but I would also love to hear from any among you who have experience doing data visualization.
To clarify, I expect to emphasize statistical graphics and (to some extent) multidimensional, large-n data display, while entirely avoiding "infographics." Also, Tufte -- at least The Visual Display of Quantitative Information -- goes without saying.
I have some ideas of my own, but it I think it would be useful to gather from a wider range of experience before getting started. I've come up with a list of specific questions, but I would be happy to read any relevant suggestions.
= What is the best way to structure such a class? What topics need to be covered, and in what order?
= What book chapters and articles should students read? What are the canonical papers on statistical graphics?
= Are there any discipline-specific visualization papers from your area of interest that could serve as good (or bad) examples of designing graphics with a specific purpose in mind?
= I am expecting to have them work and develop projects in R, but what types of weekly assignments and semester-long projects would be most useful?
= Are there good examples of such syllabi available online for reference? If any of you have taken such a course, what was useful, and what was not?
Thanks for your time and ideas!
To clarify, I expect to emphasize statistical graphics and (to some extent) multidimensional, large-n data display, while entirely avoiding "infographics." Also, Tufte -- at least The Visual Display of Quantitative Information -- goes without saying.
I have some ideas of my own, but it I think it would be useful to gather from a wider range of experience before getting started. I've come up with a list of specific questions, but I would be happy to read any relevant suggestions.
= What is the best way to structure such a class? What topics need to be covered, and in what order?
= What book chapters and articles should students read? What are the canonical papers on statistical graphics?
= Are there any discipline-specific visualization papers from your area of interest that could serve as good (or bad) examples of designing graphics with a specific purpose in mind?
= I am expecting to have them work and develop projects in R, but what types of weekly assignments and semester-long projects would be most useful?
= Are there good examples of such syllabi available online for reference? If any of you have taken such a course, what was useful, and what was not?
Thanks for your time and ideas!
Best answer: What is the best way to structure such a class? What topics need to be covered, and in what order?
Maybe you should take Tufte's one-day class. It might help plant seeds about what ideas you want to present, and how you want to present them. He really hates PowerPoint, for example, to the extent that he goes out of his way to make answering why a part of the class. If you use slides, that might challenge you to try alternative presentation methods, which may have the side effect of making your class that much more interesting to students.
What are the canonical papers on statistical graphics?
This review paper of Tukey's work in informatics has a ton of references to get you started.
Are there any discipline-specific visualization papers from your area of interest that could serve as good (or bad) examples of designing graphics with a specific purpose in mind?
Miriah Meyer and Martin Krzywinski are two informaticists who are doing important and creative work in visualizing biological data. I did a post about Krzywinski's hive plots not too long ago. Both sites have plenty of PDFs and citations to review.
I am expecting to have them work and develop projects in R, but what types of weekly assignments and semester-long projects would be most useful?
For a semester-long project, perhaps think about the Google Data Visualization Challenge as a jumping point. For weekly assignments, start with an introduction to R, and then do mini-challenges every week after, as well as reading assignments. Pull reading from the classics, but also add in scientific papers from various fields that make use of specific techniques.
You might also have students critique papers — how does this figure fail to communicate the paper's idea clearly and unambiguously? That sort of thing.
Are there good examples of such syllabi available online for reference? If any of you have taken such a course, what was useful, and what was not?
I have never taken an informatics course. I relied on reading Tufte and Tukey to get a foundation. I read a lot of scientific papers and keep an eye out when interesting visualizations show up. I ask myself: Why did the researcher(s) choose this visualization approach over another? How does this visualization help push the paper's narrative? How does it hurt? I quietly incorporate new and revised knowledge into my own work.
posted by Blazecock Pileon at 1:04 PM on February 14, 2012 [2 favorites]
Maybe you should take Tufte's one-day class. It might help plant seeds about what ideas you want to present, and how you want to present them. He really hates PowerPoint, for example, to the extent that he goes out of his way to make answering why a part of the class. If you use slides, that might challenge you to try alternative presentation methods, which may have the side effect of making your class that much more interesting to students.
What are the canonical papers on statistical graphics?
This review paper of Tukey's work in informatics has a ton of references to get you started.
Are there any discipline-specific visualization papers from your area of interest that could serve as good (or bad) examples of designing graphics with a specific purpose in mind?
Miriah Meyer and Martin Krzywinski are two informaticists who are doing important and creative work in visualizing biological data. I did a post about Krzywinski's hive plots not too long ago. Both sites have plenty of PDFs and citations to review.
I am expecting to have them work and develop projects in R, but what types of weekly assignments and semester-long projects would be most useful?
For a semester-long project, perhaps think about the Google Data Visualization Challenge as a jumping point. For weekly assignments, start with an introduction to R, and then do mini-challenges every week after, as well as reading assignments. Pull reading from the classics, but also add in scientific papers from various fields that make use of specific techniques.
You might also have students critique papers — how does this figure fail to communicate the paper's idea clearly and unambiguously? That sort of thing.
Are there good examples of such syllabi available online for reference? If any of you have taken such a course, what was useful, and what was not?
I have never taken an informatics course. I relied on reading Tufte and Tukey to get a foundation. I read a lot of scientific papers and keep an eye out when interesting visualizations show up. I ask myself: Why did the researcher(s) choose this visualization approach over another? How does this visualization help push the paper's narrative? How does it hurt? I quietly incorporate new and revised knowledge into my own work.
posted by Blazecock Pileon at 1:04 PM on February 14, 2012 [2 favorites]
I have never taken an informatics course
Well, I took Tufte's class, but that was a one-off and not what I would call a course, so much.
posted by Blazecock Pileon at 1:06 PM on February 14, 2012
Well, I took Tufte's class, but that was a one-off and not what I would call a course, so much.
posted by Blazecock Pileon at 1:06 PM on February 14, 2012
The O'reily book Visualizing Data is pretty interesting. The book stemmed from Ben Fry's MIT dissertation and describes a methodical approach to visualizing data.
posted by jchaw at 2:09 PM on February 14, 2012
posted by jchaw at 2:09 PM on February 14, 2012
Best answer: Marti Hearst and Jeff Heer (two prominent researchers in information visualization) co-taught a course at Berkeley a while back. Lots of slides that I'm sure they'd be happy to let you use if you just credit them.
http://courses.ischool.berkeley.edu/i247/f05/
Jeff Heer and his colleagues also have a great paper which gives an overview of the different kinds of visualizations around today. I use this as an overview in my courses.
http://queue.acm.org/detail.cfm?id=1805128
I know Tufte is really popular, but a big drawback is that he doesn't ground his work in any kind of empiricism. His philosophy is more of proof by authority. At the CHI conference one or two years ago, during the Madness session where everyone gives a 30 second summary of their work, people applauded when some researchers showed that chart junk can actually improve memorability (as opposed to Tufte's philosophy that all chart junk is wrong, wrong, wrong!). Tufte is definitely something to include, though I'd also throw in some caveats and discuss limitations.
I don't know a lot about using R for visualizations, though there are several blogs out there where people discuss their work. Searching for "r blog visualizations" yields a lot of results.
In the social web class I'm teaching now, our current plan is to have the students use either many-eyes or NodeXL to create some kind of interesting visualization. Our goal is to give a flavor of what is possible rather than doing rigorous research, but they could also be helpful in your class for broadening the horizons of what is possible while minimizing the pain involved in using R.
posted by jasonhong at 2:52 PM on February 14, 2012
http://courses.ischool.berkeley.edu/i247/f05/
Jeff Heer and his colleagues also have a great paper which gives an overview of the different kinds of visualizations around today. I use this as an overview in my courses.
http://queue.acm.org/detail.cfm?id=1805128
I know Tufte is really popular, but a big drawback is that he doesn't ground his work in any kind of empiricism. His philosophy is more of proof by authority. At the CHI conference one or two years ago, during the Madness session where everyone gives a 30 second summary of their work, people applauded when some researchers showed that chart junk can actually improve memorability (as opposed to Tufte's philosophy that all chart junk is wrong, wrong, wrong!). Tufte is definitely something to include, though I'd also throw in some caveats and discuss limitations.
I don't know a lot about using R for visualizations, though there are several blogs out there where people discuss their work. Searching for "r blog visualizations" yields a lot of results.
In the social web class I'm teaching now, our current plan is to have the students use either many-eyes or NodeXL to create some kind of interesting visualization. Our goal is to give a flavor of what is possible rather than doing rigorous research, but they could also be helpful in your class for broadening the horizons of what is possible while minimizing the pain involved in using R.
posted by jasonhong at 2:52 PM on February 14, 2012
I'm a graduate student currently trying to learn more about data visualization, and Nathan Yau's Visualize This has been very useful.
posted by tinymegalo at 4:54 PM on February 14, 2012 [1 favorite]
posted by tinymegalo at 4:54 PM on February 14, 2012 [1 favorite]
The topic of visualization with R came up in one of my classes the other day, and based on some cursory searching of the info I wrote down, I would recommend checking out Hadley Wickham, who wrote ggplot2 as part of his dissertation, I believe. There's a syllabus on his site, and a bunch of other good-looking resources.
Other people/stuff to search for, in no particular order: Heike Hofmann, Di Cook, Andreas Buja, rggobi package to interface with ggobi, software called Mondrian, and "Exploratory Data Analysis" by Tukey.
posted by mean square error at 6:48 PM on February 14, 2012
Other people/stuff to search for, in no particular order: Heike Hofmann, Di Cook, Andreas Buja, rggobi package to interface with ggobi, software called Mondrian, and "Exploratory Data Analysis" by Tukey.
posted by mean square error at 6:48 PM on February 14, 2012
David Comberg at Upenn is teaching a class on Information Design and I follow the blog. It's a Fine Arts course so it's design oriented but you might be able to find some things in the syllabus or the resources link helpful.
posted by tangaroo at 7:54 PM on February 14, 2012
posted by tangaroo at 7:54 PM on February 14, 2012
This thread is closed to new comments.
I'd have a class on how difficult it is to get visualizations published.
Google visualization + .edu + syllabus to try to find syllabi.
posted by k8t at 12:59 PM on February 14, 2012