From CS294-10 Visualization Fa07
== Bad Visualization ==
"Access Denied: Broadband service is available worldwide, but it's beyond most people's budgets." Wired 9/07
Assignment 1b answers:
This visualization attempts to show the cost of broadband around the world, in US dollars. The data is fairly simple: (country (nominal), cost of broadband in US$ (quantitative)) is mapped to onto a world map, with the area of the country on the map being colored a particular color, depending the range of its broadband cost. Countries with no broadband are depicted as black.
This visualization fails in a number of ways. First, the pink oceans are overwhelming, and add nothing to the visualization -- in fact, they make it hard to read. Second, by using color to represent an quantitative dimension, i found myself constantly referring to the legend, rather than looking the actual visualization. There was no implicit understanding that blue meant ok cost, other than by referring to the legend. Finally, the wrong data is depicted. In all the textual bubbles, the author refers to the cost in country X being Z times the monthly wage. This is interesting. This is what should've been depicted, as what readers are interested in is knowing "can someone in country X afford broadband?" not whether they as the American reader might be able to afford it. The current layout requires that the viewer make some inferences about what is affordable and what is not.
Unfortunately, I couldn't find data converting this into affordability of broadband for the average consumer in each country. As stated above, that would've been ideal. Instead, I looked to see what was likely to be affordable in most places: that was the $0.00 to $1.00 range. How things distributed in this range are marked in different intensities of green. Similarly, broadband that's certainly expensive ($1.00 and up) is marked in shades of red. Finally, the no broadband category, which is an interesting category on its own since it means that the country does not offer access to broadband, is still marked in black. I also included country borders in this graphic, which were oddly removed in the original.
below i've depicted africa and europe, as they had the most interesting variations. note that you can quickly tell within europe what's cheap versus expensive (as compared to the original) and comparing europe to africa is dramatic. Then, within africa, you get a great sense of
My photoshop skills are a bit lacking: i wanted to break the legend up into three sections, with the words "Affordable" around the green hued areas, and "Expensive" around the $1 to $500 area, and "No Broadband" at black, but gave up. Also, the $1 to 9.99 looked more red on my LCD but looking at it now on another monitor it's almost white. I realized it is quite hard to cover a range of 5 elements with different saturations of one hue!
A problem that still remains is answering questions that might be implied by the visualization. Since the country names are not on the graphic, it's hard to go from the graphic to getting more info about any particular country. Interaction would support this. Two letter abbreviations for countries might work as well, though I found that unintuitive and distracting.
== Good Visualization ==
New York Times weather info, 9/2/07.
Assignment 1b answers
As shown in the graphics, the temperature trends for various major US cities are depicted. These graphics depict an enormous amount of data compactly. The days of the week (ordinal) are depicted on the x-axis. The temperature (quantitative) are depicted on the y-axis. there are four types of temperatures (nominal): record high/low, normal high/low, actual high/low, forecast range high/low. the data is printed in greyscale, so the these four nominal categories of temperature measurements are distinguished by color value (for instance, the records are lightest while the current temperatures for this year are black). To further distinguish forecasts from actual temperatures, the thickness of the line depicting the range between high/low is varied. So forecast high/low has a thin black bar connected highs and lows, and a slightly thicker bar to indicate the high and low ranges, whereas measured temperatures have a thick bar.
The graphic is extremely useful. Looking at the san francisco graph one immediately notices that (1) it's hotter than usual and (2) it's expected to not be so hot in the upcoming week (as compared to usual) (3) and we're generally no where near record highs or lows. in addition (though it's not as useful) side-by-side comparisons of cities are possible because they use the same scale between graphs. so we can see that san diego temperature is pretty tight (not much difference between high and low compared to san francisco) and it's generally slightly warmer and not as cool. unfortunately, the side-by-sides are just alphabetical, and it gets harder to compare, say, detroit, with san francisco (the variability of temperature is perhaps the only visually easy thing to compare in non-side-by-side plots).