A1-KevinLim
From CS294-10 Visualization Fa07
Contents |
[edit] Bad Visualization: Honda Advertisement (Economist, 01 Sep 2007)
[edit] Original
http://flickr.com/photos/k7lim/1313233947
Explanation and Critique: This first visualization is an advertisement for a car company that wants to show that it is more efficient than its competitors. It uses a photograph as an ornate way to show a comparison of two numbers, but by doing soe, introduces a significant ambiguity: where are the actual points/baseline located? The points are literally nebulous... clouds. And it is unclear whether the baseline is at the foot of the mountains, or at the top of them, or somewhere else. Regardless of where I measured the baseline, there is a lie factor. If measuring from the top of the mountain as the baseline (to give them the benefit of the doubt in terms of lie factor), the lower number is 10cm up, and the higher number is 14 cm up. The actual data says that the higher number should be 20% higher (25 vs. 30).
[edit] Deconstruction
The data set for the visualization is exactly two numbers: the average fuel efficiency for Honda and for the entire auto industry as expressed in miles per gallon. The image model here is to use photographic images as stand-ins for a standard bar graph for the comparison of one-dimensional data. Here, the clouds stand in for the data points that they want the viewer to compare. Their distances away from the "ground", or the sky/mountains between the clouds and the bottom of the page, encode the value of the fuel efficiency. The values are shown over a nominal variable of basically "Honda" and "Industry Average." The mountains, if they are to be considered an "element" are uninformative. They create this second quasi-axis, which is confusing and provides no data.
[edit] Redesign
I feel a little silly how straightforward my redesign is. Simply taking Tufte's notions of Lie Factor and Maximizing Data-Ink, I've gone ahead and made a very simple visualization that shows the data from the ad.
Lie Factor: I drew this simple bar graph and aimed for a Lie Factor of one. I therefore measured out a direct data-to-drawing mapping- the mapping is 1 cm = 1 mpg. I decided to make the bars two different colors, just because it was more visually pleasing, without adding excess ink or distraction.
Maximizing Data-ink: I chose not to make fat bar graphs, because I'm simply showing a comparison of two quantitative values. I was tempted to not even draw the zero-line (leaving the bottom of the board as the implicit zero) but decided to make it explicit.
Making the nebulous markers into absolute ones: Part of the deceptiveness of this persuasive visualization is the fact that it's hard to pin down where the data-points or the axis actually IS. I made it very clear, marking out the point and the axes explicitly.
Labels: I have torn out the labels I found to be appropriate directly from the magazine, left as is. The one that's hard to read says "10-year Corporate Average Fuel Economy." What I've done is emphasize the source of the data, where previously it was left rather hidden. I feel this is more compelling, and that if you're going to have a data set as simple as two numbers, you might as well make it clear what those numbers actually mean.
[edit] Good Visualization: Capital Punishment Map of the US (Economist, 01 Sep 2007)
http://flickr.com/photos/k7lim/1313143503
Explanation and Critique: This image is a visualization of capital punishment/death row data from the individual states in the US. Tufte praises visualizations that are about the data, and allow for exploration, and this map had me "zooming" in and out, looking for trends, exploring the regions I'm familiar with, comparing my stereotyped view of regions with the data.
[edit] Deconstruction
The data set includes the number of inmates on death row and the number of executions since 1976 for each state in the US (and the District of Columbia). Also the data set includes whether or not a state imposes the death penalty at all. The data set also includes states that technically have the death penalty but don't practice it due to the declaration of unconstitutionality. The data set also includes the names, the latitude, longitude and geographic shape of each state.
The map uses the X and Y axes to encode latitude and longitude.
The map uses hue to show a state's policy and practice towards capital punishment: Green means a state has no capital punishment (therefore not executing anyone). Brown means a state has capital punishment laws, but has not executed anyone within the timeline (since 1976). Red means a state has capital punishment and has used it since 1976. The brown hue may seem redundant, but I think it's a fair point to show that the theory and practice of capital punishment diverge. It's kind of clever in my opinion to squeeze these two variables (official policy and actual practice of capital punishment) into one visual variable (there are only three values since no state is against capital punishment on the books, but actually practicing it).
The visualization shows the exact number of current inmates on death row, and the exact number of inmates executed since 1976 via explicit number labels. Each capital punishement state bears both numbers, and the two are differentiated using position and a negative color scheme (red on white for executions show below, white on red for executions shown above). The use of position and the use of the color reversal may seem redundant, but I found it useful in that I didn't have to look back at the legend.
Finally, in the case of New York, the visualization explicitly labels with an asterisk in the legend that that state has the death penalty but that the policy has been struck down in the courts.
I found the dotted lines to be unnecessary, vibrating and distracting.
[edit] Brian Gawalt - Sep 05, 2007 12:40:29 pm
It's interesting for me to notice that in your "bad" image, one of the two clouds has been recreated from the other via Photoshop. It means the distortion, the imbalance between the two columns, wasn't an artifact of a stock photograph of two clouds who happened to be at conveniently different altitudes. The visualizer took pains to make sure "30" looked vastly higher than "25".
