From CS294-10 Visualization Fa08

Jump to: navigation, search

Good Visualization

Nature, 430, 277-384, 15 July 2004. no 6997. Page 314.


This compares several related proportional measures. This type of information might typically be presented in a series of bar or pie charts. But this unified representation allows easy comparison between countries (in a single discipline) and between disciplines. It also makes use of the arrangement of the disciplines around the origin (grouping life and physical sciences).

One noteworthy design decision was the inclusion of the overall measure, rather than just having the breakdowns. One potential misinterpretation of this visualization is to compare enclosed areas, rather than linear distances along the spokes. Including the overall measure, and placing it at the prominant 12 o'clock position (a natural starting point for the viewer) helps avoid this.


The original dataset was probably a list of academic publications by authors in G8 countries. The size of this sample is not specified. This dataset was reduced by computing derived values. Each paper was categorized as belonging to a country (based on authors, presumably) and discipline. The resulting dataset has one value (share of publications) for each country and discipline (plus an overall value for all disciplines). 7 countries and 7+1 disciplines equals 56 datapoints. Additionally, 6 of the categories are classified as life or physical sciences, for a total of 62 datapoints.

  • Country is a nominal variable. It is represented by both color and the shape of the marker.
  • Discipline is a nominal variable. Is is represented by angular position.
  • Discipline category (life science vs physical science) is a nominal variable. It is represented by position of the spokes relative to relative to the vertical mid-line.
  • Share of citations is an interval qualitative variable. It is represented by length along the spokes.

The colored lines connecting the spokes are not informative.

Bad Visualization

Bicycling, October 2003. Page 42.


This visualizes the route of a bike race.

The primary flaw in this visualization is that it represents such a small amount of the relevant information. The numbered text does not always describe the corresponding section of the text (e.g. item 1 describes the course in general). The text includes information about speeds, change in elevation, road type - but none of these are represented visually. The visual information actually conflicts with the verbal description, segment 7 is shown as a perfectly straight line, but the racer explains how significant the curves are. And there is no scale on either map!


This visualization is representing the course of a bike race.

  • Direction is a nominal value. It is represented by arrow icons.
  • Position (latitude/longitude) of the route is a pair or interval quantitative values. It is represented by the position of the black line.
  • Position (latitude/longitude) of other roads is a pair or interval quantitative values. It is represented by the position of white lines.

The size of the dataset is not clear since individual points are not shown. The route can be approximated with about 20 straight line segments or about the same number of street-level turn-by-turn directions. The color of the background (yellow and brown) seems uninformative, it may correspond to some political boundary irrelevant to the race.

Reconstructed Visualization

Click image to zoom


The primary change made is to include a visualization of elevation. This is one of the most important features of a bike route.

  • Elevation is a quantitative variable, represented as length along the y-axis.
  • Distance along the course is an interval quantitative variable, represented as length along the x-axis.

This graph plots the elevation (in feet) against the distance along the route (in miles). The area under the curve is shaded, which is not informative but might make the visualization more intuitive since it maps to the solid ground of the real world. Additionally, this enables the y-scale to be marked by subtracting ink rather than adding gridlines.

The x-axis marks quarter mile increments and is labelled using the same mile markers as the overhead view.

The overhead view has been redone. It includes additional features of the terrain, marked with subtle colors, to provide context. The scale is shown in the lower left, and mile markers are placed along the route. The directional arrows have been removed since they did not seem necessary to follow the route.

  • The mile-markers are difficult to see (especially when zoomed out). They should be done in a color which stands out more.

[add comment]
Personal tools