A1-BrianGawalt
From CS294-10 Visualization Fa07
[edit] Good Visual
From Newsweek's July 30, 2007, cover story on Islam in America:
The graphic above does a great job of conveying quantitative, ordinal, and categorical data. Vertically stacking the three bars captures the progression of the age brackets, while the distance measure along the horizontal allows for easy comparison of the quantitative percentage the different responses pulled in. The thin bars also avoid some of the pitfalls and illusions of comparing by area (where big differences wind up discounted). Most of all, the use of hue I think is excellent here. Neutral gray is employed for the "don't know" abstainers, while there remains a progression from "Too much" to "Not enough" as far as the color brightness is concerned.
Deconstruction
A survey of Americans was taken soliciting opinions on the level of Muslim immigration into the U.S., and the visualization displays the resulting data broken down by age bracket of the respondents. The data model has three dimensions. The first, age bracket of respondent, is ordinal. The second, response type, is nominal. The third, response rate, is quantitative (ratio). The image model handles each differently. Age bracket of respondents is encoded by vertical position of the bars. Response type is encoded two ways: horizontal and the hue and value of each bar segment. Response rate is encoded by the length of each bar segment.
[edit] Bad Visual
From the text Introduction to Wireless and Mobile Systems by D. P. Agrawal and Qing-An Zeng:
This graphic is busy with overlapping, thin, black ellipses. Making matters worse, the thin, black, elliptical at the center is in fact the surface of the Earth, a fact that does immediately register with the viewer. You can see in the original a slight attempt to make it a bolder line and distinguish it from the other lines (which instead represent satellite loci), but this comes up short. The textual labels of the orbital altitudes are needlessly wedged in the center. The labels for the different orbits are spread haphazardly about the ellipses, but a simple vertical realignment could provide the reader with quick insight into which orbits are farther or nearer Earth.
Deconstruction
The visualization encodes (simplified) sizes and shapes of the planet and several satellite orbits. This is based on a data model with two quantitative (ratio) dimensions ("longitude" and "latitude"), a nominal dimension (planet or orbit), and an ordinal dimension (LEO < HEO < MEO < GEO). Also included is a nominal dimension labeling special facts about each object (e.g., "Iridium satellites float here" or "This is the planet Earth"). The image encodes the quantitative data using elliptical shapes and areas for each object and the nominal and ordinal data with text.
Redesign
By stacking the labels, and varying the thicknesses of the arrows, I've introduced a stronger encoding of the ordinal data to the image model. The "~" characters now convey which radii labels are approximate (the HEO, LEO, and MEO orbits) and which are exact (GEO, the geostationary orbit, which can be achieved only at the given altitude). By changing the hue of the interior of Earth, I've added a superior encoding of the nominal dimension. Tilting the orbits maintains the original quantitative encoding while reducing clutter among the lines. The orbits' lines stay the same width in order to keep comparisons between their sizes and shapes fair. As an aside, I felt I should stay true to the constraints of the original image author and work only in the gray scale.
