Eyes on the Prize, Part Two
April 18th, 2008(This week is the second of a two-part post on the data-sharing site “Many Eyes”. If you haven’t, you should definitely read part one (about the site’s visualizations) first. This week, by contrast, is going to be about the specifics of those visualizations and the datasets I uploaded.
This two-part post is also the third in a series on data-sharing websites. Here are parts one and two of that series.)
Now that you understand a bit more about the (many) types of data visualizations Many Eyes has to offer, we’ll get into them more in-depth and discuss the process of uploading data as well. Like the first post, I’ll go through the different visualizations in turn.
We’ll start with the world and country maps. Though I didn’t try to make one specifically, they seem like they’re done well. Here’s a good example of one for the U.S. Very intuitive graphs, great use of color, not a lot of visual clutter, etc. Hopefully at some point they will let you break down country maps by county/township/whatever so that you can get even more granularity. That way the country charts can be just as data-dense as the world charts. It’s a minor criticism, though. (And it might be hard to standardize that kind of data to overlay on a map anyway.)
The line charts look good too, this being a good example. The lines are a bit crammed together, but the point of the visualization is clear: people in the U.S. are spitting out kids like Pez dispensers compared to the rest of the developed world. Anyway, the line charts have many good features, such as light gridlines and labels, muted colors, intelligent use of intervals for both axes, and so on. Considering how easy many of these principles are to mess up (and how often it happens), I commend Many Eyes for getting line charts right. Since line charts are such a basic and important tool in the statistical graphics toolbox, this is quite important.
As I said in part one, I’m less keen on the cousin of line charts, the stacked line chart. That said, like with line charts, I guess they’re implemented about as well as they could be. When used with categories you can create some interesting stacked line charts, but ultimately they seem like stretched-out pie charts to me. As such, they suffer from the same problem - you just can’t gauge the parts to the whole very well. On top of that, filling in the areas underneath lines creates a lot of color-based visual clutter while drastically reducing the “data-ink” ratio that I speak about here so often. Remember, kids, just say no to stacked line charts.
Bar charts, like line charts, are done well at Many Eyes. And they’re good for mostly the same reasons the line charts are good. One gripe, though: you can see from the example image I linked to that they rotate the horizontal axis labels 45 degrees. (Many Eyes does this for lots of different charts.) That’s bad for readability. Ultimately, it’d be better if Many Eyes just showed less categories when you run out of room like that, and have the ones they do show be oriented normally. Still, I guess if you have to have all the categories show up at once, that’s the best you can do. Like line charts, these are basic and important charts, so it’s good that they’re done well.
For my part, I uploaded a bar chart that was based on the data I used in my ethnic cuisines post. Like Swivel, uploading data was a bit of a pain. I was honestly a bit surprised that you could only copy and paste data into a box. Swivel at least had options for uploading Excel spreadsheets and comma-separated-values (CSV) files.
Plus, Swivel had an Excel toolbar to help you get the data in the right format. The best you get at Many Eyes is a tutorial, which I appreciate, but which Swivel had as well. Annoyingly, you can’t even right-click in the copy-and-paste box. You have to use the keyboard shortcut to paste your data in.
Even worse, I found it very hard to format some of the tables and graphs how I wanted. For example, in a bar chart, you can’t flip the rows and columns in the way you would expect to make a horizontal bar chart. That gets a big thumbs-down from me. (Could I be the Roger Ebert of data? Well, at 142 pounds I might have to bulk up a bit to fit the bill.) Again, in comparison Swivel handles that flipping effortlessly. As for tables, why are the headers so hard to format? I don’t want those columns to be so wide, and I definitely don’t want centered column headers.
Oh well. Usability was an issue with Swivel, and it’s an even bigger issue on Many Eyes. As usual, it’s easier to re-upload a dataset than to try and fix it. For example, when I tried to format one of my datasets, Many Eyes insisted on creating another version of my data, which I could not figure out how to delete for the life of me. Plus, I still didn’t get to format the data the way I wanted. These data-sharing sites definitely seem to be designed more around viewing than uploading.
Anyway, all whining aside, let’s get back to the visualizations. The block histograms, bubble charts, and matrix charts mostly seem to be on the level. There are good examples here, here, and here. The histograms have vertical gridlines, but I suppose those are an OK addition given how discrete and “blocky” histograms are. Bubble charts and matrix charts suffer from the usual problem of having circles to judge size (as I mentioned in part one), but those are problems that are probably inherent in using those types of graphs. Also, sometimes it’s hard to read the labels on the bubble charts. Again, I don’t know what you’d do about that, exactly. (Making a legend is not a particularly helpful suggestion, since you have to cross-reference it so often.)
Scatter plots and network diagrams look good, as you’d expect. You can find good examples here and here. Yet again, there’s not much to screw up here, but Many Eyes does a good job of keeping things simple and uncluttered. Scatter plots form the last part of the “holy triumvirate” of charting, along with line and bar charts, and they’re done well. That means the most important and basic visualizations work like they should and give Many Eyes a solid basis to create charts. (So you see, I’m not all negative here, lest some accuse me of being a “hater”.)
I’m still not sure what makes a good pie chart, though. It’s almost like an oxymoron to me, as you probably already know if you’ve read my other posts. (Feel free to label me a “pie chart hater” if you like, I don’t mind.) Still, Many Eyes does choose muted colors and good, clear legends, so I guess they’re implemented as well as they can be. I am willing to admit that, on occasion, there are good pie charts, but they are few and far between, and they could probably be done just as well (if not better) using a different kind of visualization. And whatever you do, please don’t mix pie charts with other chart types. My eyes are still watering from that visualization I linked.
In contrast to pie charts, the treemaps seem great. Here’s an example. It’s best if you keep them monochromatic, like the one I linked. Multi-colored treemaps can be much more visually distracting and hard to read.
If you tried out Sequoia View, the hard disk visualization tool I mentioned in part one, you’ll note their treemaps have one color as the default. (An astute reader pointed out that “Sequoia” is a not-so-subtle pun on treemaps, as opposed to the incorrect explanation I offered. Thanks Frank!) You can use multiple colors on treemaps to great effect, but it seems like the dark, highly-contrasting colors Many Eyes uses are not the way to do so. Otherwise, I like their treemaps. As I said in part one, consider using these in place of pie charts when you need to show the relationship of parts to a whole.
Last up are tag clouds and word trees. All you need to display here are words with simple formatting, so it’s no surprise that they’re done well in Many Eyes. They get the sizes and network links mostly right too, so whatever you need to know is pretty clear by looking at them. Here’s a humorous tag cloud example, and also a word tree of Martin Luther King Jr.’s iconic “I Have a Dream” speech.
Even better, as I also mentioned in part one, Many Eyes recently rolled out a tag cloud comparison visualization. Many Eyes basically takes two separate word clouds and, by using sizes to indicate frequency and color to indicate which text, compares them. Continuing with the theme of those other text visualization links (i.e. famous black people), I made my own comparison tag cloud. Being both a huge rap fan and a junkie for political news, I decided to compare the poetry of 2pac with Barack Obama’s recent high-profile speech on race relations.
To be honest, I’m pretty proud of the visualization. You can see at a glance that 2pac is more emotional, positive, idealistic, and less focused on race than Obama. (Is this why Obama is catching so much heat for his infamous “bitter” comment?) Yet again, though, the viewer side is better than the user side. I think the visualization is great, but uploading the data was kind of a pain.
To make a comparison tag cloud, you have to label two texts and put them back-to-back in the same file, and then upload it. That constitutes the dataset. What you should be able to do, however, is just take any two texts and compare them on the fly. Since you can’t, though, it’s a lot harder to do a tag cloud comparison than it should be. I hope that, since they just rolled out this particular visualization, they’ll implement my suggestion later, especially since this visualization is such a great idea.
And that’s it for my write-up on Many Eyes. I hope I don’t sound too critical of Many Eyes here. All hater-esque tendencies aside, it really is a great and extremely ambitious data-sharing site. There’s just so much to talk about that I had to gloss over a lot of my positive comments in order to focus on the fair number of criticisms I had. The usability, customizability, and dataset editing and uploading could all use a lot of work, but most everything else is great. Plus, like Swivel, Many Eyes is still in beta, so I’m sure I can cut them some slack and stop all the hatin’.
If you want to keep up with the site, check out the main page (where they feature visualizations) and the Many Eyes blog. I certainly plan on doing so, and I look forward to more great things from Many Eyes in the future!
| | | del.icio.us |
April 22nd, 2008 at 10:25 am
I’m pretty sure SequoiaView was named after the ‘big tree’, since it was meant to explore large hierarchy trees. I should know, I implemented it..
Thanks for letting me know, Frank. I’m just wrong there. I corrected it in my article. Don’t know what I was thinking there. I guess I came up with some explanation years ago when I first used it (ignoring the obvious explanation) and I never questioned it.
That’s a crazy anecdote you dropped there at the end of your comment. I wonder if you meant you implemented the treemap in SequoiaView, or on Many Eyes?
- Dave