A2-JonBarron

From CS 294-10 Visualization Sp10

Jump to: navigation, search

Image:JonBarron-A2.png


We used a design inspired by the "multi-window plot" of age and sex referenced by Tufte ("the Visual Display of Quantitative Information", page 113). Total donations are plotted on the left, while relative donations (donations normalized by population size) are shown on the right. We only list nations, as the other organizations have no population size, so we did not know how to plot with this data without unfairly biasing the perception of the viewer. We separate each donation into total pledges, and pledges that have already been committed (which is some fraction of the total pledge). We do this using a technique similar to the framed-rectangle chart of Cleveland and McGill, except that, since we have more space to work with and since length is the most effective mechanism for displaying quantity, our rectangles are very long and thin. We feel this strikes some balance between simply using the length of lines to indicate amount (primarily advocated for by Tufte) and the framed-rectangle charts (which allow us to show part/whole relationships). Additionally, we plot the average donation as a red line (mean donation across all countries on the left, mean donation-per-person on the right), such that donations can be easily compared to some baseline, and so that the nations who contributed more or less than average can easily be spotted.

We encode each nation's population in two ways: nations are sorted vertically by population, and groups of nations are color-coded by their size (specifically, by log10(population)). We initially explored using the vertical location of each bar to indicate log(population), but because many nations have fairly similar populations, this lead to distracting grouping. Though our method obscures the exact population of each nation, that data is not nearly as important as being able to quickly compare each nation to other nations of similar size, or determine which nation is bigger, which are what our method allows. Population is encoded in discrete colors, where saturation increases with population. Color seemed like a natural dimension for grouping these lines, and as mentioned in the readings, varying saturation is more effective than varying hue (and varying brightness would create readibility issues with the black text), and discrete colors are easier to distinguish than continuously varying colors. The horizontal lines the viewer sees when the discrete colors vary also serve to partition the data into powers-of-10 boxes, which are labeled in the lower left of each box.

Care was taken to reduce chartjunk. We eschewed borders around the bar charts and ticks on the x axis, and minimized the background grid to just a sparse set of vertical marks to make it easier to determine the precise dollar amount. We regret the obnoxious legend, and the numerous labels at the top, but they seem necessary.

The two contrasting bar plots draws your eye to the difference between absolute donation and relative donation. The US looks very generous in absolute terms, but not at all in relative, per-person terms, while countries with between 1 million and 10 million citizens contribute heavily per-person, even though they contribute very little in absolute terms. Additionally, having the two plots side-by-side and aligned by country makes it easy to quickly switch absolute/relative perspectives for any given country. For example, China looks roughly average in absolute donation, but a quick glance to relative donation shows that they are pitifully below average. In contrast, Luxembourg seems exceptionally generous in relative terms (well above average donation-per-person, with all of it fully committed), but in absolute terms on the left, we see that the donation is almost entirely insignificant. The framed-rectangle charts allow us to quickly see that though some countries (Canada, Spain) appear to have donated very heavily, that the majority of their donations are not yet committed.

Jon Barron - Feb 10, 2010 02:25:47 pm

Visualization made (data imported, sorted, and plotted) in Matlab, revised (colorized) in Adobe Illustrator.



[add comment]
Personal tools