A1-KenghaoChang
From CS294-10 Visualization Fa07
Contents |
[edit] Kenghao Chang - Sep 04, 2007 04:32:36 pm
[edit] Good visualization
[edit] First example
I'm going to talk about the visualization i found especially useful in maps.
The first image is the map in downtown Las Vegas, in which there are drawings of the hotel buildings on the main street, Las Vegas Boulevard. When i travel to somewhere new, i always like to take this kind of map as a guide. It can help me get into the context easily: not only knowing what the famous spots are, but also knowing where they are at the same time, through a single map. In addition, when i'm trying to locate myself in the map, i don't have to walk to intersections and read the street name printed in the small signs. Instead, I can use the buildings around me to locate myself (by mapping the look of the building into the map drawing), as what i do when i try to locate myself in my hometown.
[edit] Deconstruction
In the deconstruction of the map, the data set includes data structures to record information about streets, buildings, and spots (where colored stars are drawn in the map).
A street can either be recorded as a line with only two end-points if the street is straight, or as a sequence of segments (or even a set of curves) if the street is not straight. Each street is labeled with a name, and it can be addionally marked as a highway, strip trolley route, or downtown trolley route. For each building drawn, the data set includes 2D coordinates of its location, the name, and the image link to its drawing. As for the spots, the data set includes 2D coordinates, its name, and the category it belongs to, for example, "attractions", "points of interest", "hotels/casinos", and "shopping centers".
For the image model, a 2D coordinate in the data set is associated with a location point (x and y) in the map. Each street segment connects two location point with a line (or a curve). A street would be colored as blue if it's a highway, colored as green if it's marked as a strip trolley route, or colored as red if it's marked as a downtown trolley route. Street names would be carefully labeled so that no two names of the interected streets would stay too close to each other. For buildings, a drawing image would be placed in the coordinates of each building. For spots, names (text) and colored star (color + shape) would be placed on the coordinates. The star is colored as green if the spot is for "attractions", as purple if it's for "points of interests", etc.
In general, the colors represent nominal values. Locations (coordinates) represent quantitative values.
[edit] Second example
I also found the map below useful when i tried to bike to Tilden park at the first time. This map shows the street grades: the street has 0-5 degrees if it's colored as white, there is 5-10 degrees in a street if it's colored as pine, and so on. In other words, the darker the street color, the steeper the street is.
When i tried to make a route planning, i knew it's my first time doing it and i didn't want to push myself too hard. Then, i used this visualization to quickly trace and find out a route from campus to tilden park, which has the lightest color.
[edit] Deconstruction
The deconstuction here essentially follows the previous example. However, there is an additional attribute assoicated with each street segment: the street degree. As for image model, street segment would be plotted as different color: the higher the street degree, the darker the street's red color is.
Here, quantitative street degree is segmented into three levels (the steepest, steep, and flat). Levels are ordinal data. In the image model, ordinal data is represented by color.
[edit] Bad visaulization
From Business Week, September 25, 2006, p. 36-37.
The image tries to convey the variance of oil price overtime, which is a price-by-time 2D plot. To convey the image more efficiently (or more novel), the author uses the gas pipe and gas outlet to tell the audience that this plot is showing something about oil. However, this innovative graphic design made negative effect, at least for me.
If we read the labels on the oil pipe (or the line of trend), it says the oil price started from a high point of $79.32 on July 14, then went down to $75.35 on July 21, and kept dropping to $63.97 on September 13. This is basically a trend going down over-time. However, at the first moment i saw this piece of visualization, I interpreted it as a going-up trend instead.
When i first saw this image, i saw a pipe, a outlet, and some bold, red numbers with $ signs. So, I knew it's something about the trend of oil price. Then, by the intuition that oil should flow all the way to the outlet, which is from right to left in the image, i got the first impression that i should read the trend from right to left, which makes me interpet that there is a up-going trend (from 63.97 to 79.32). However, after i read those smaller letters about the dates (which was ignored because of small size), i know i was wrong: the trend should be read from left to right and should goes down.
Even i was fully aware of the convention that trends should be read from left to right, i was mislead by the other convention: fluids flows from right to left in the image, i didn't interpret the idea of the image correctly.
[edit] Deconstruction
The data set in the visualization is simply a 2D data points: (date, oil price). In the image model, data points are plotted in a 2D plane, turing date as x-axis and oild price as y-axis. In other words, 2D quantitative data is represented by locations. The data is printed on the side of each location point. In addition, the adjacent data points (in the sense of time order) are connected by a line segment.
[edit] Redesign
I re-designed the bad one by removing the ambiguity caused by the flow convention.
In the original design, people would be mislead by the right-to-left flow direction and therefore mis-interpreted that the oil price was going up, rather than going down as if they read the trend from left to right. In addition, the font size of the date is too small to help people identifying the trend. As a result, i changed the icon and made it looks like the oil is flowing from left to right. And, i increased the font size for dates.
However, after i read the text by Tufte, i found this new design falls into the "chartjunk" category to some extent. The redundant icon such as oil pumps and oil pipes might help people uderstand the picture is showing something about oil. However, at the same time it might mislead people. Therefore, i'm wondering whether the best redesign is simply put a 2D chart with time as the x-axis and oil price as the y-axis.
