A1-AndrewMcDiarmid
From CS294-10 Visualization Fa07
Contents |
[edit] Bad Visualization
Source: ESPN Magazine 10, no. 17 (August 27, 2007)
This table and pie (donut?) chart show Louisiana State University's 2007 football schedule (LSU is named out of frame) and LSU's record against the highlighted opponent (Virginia Tech), respectively. The chart reveals that VT has won one game, and LSU zero. This lack of data make the visualization unnecessary, though in others accompanying this one in the magazine, the teams have played more often, thus making the chart marginally more useful. Specific problems include the illegibility of "VT" in the chart, the lack of labeling, the weak visual connection between the chart and the highlighted table, and the unexplained box to the left of "Series Record Wins."
[edit] Deconstruction
The data model for this visualization includes only the number of wins (0) and losses (1) LSU has against Virginia Tech. The image model is a pie chart. I had assumed that the color(s) used in the pie chart would correlate with each school's colors, but that seems not to be the case. VT has won 100% of the games, but the corresponding area of the pie chart is not the team's signature maroon. Thus the image model includes only shape, with the ratio of wins to total games correlated to each teams area of the circle (here a full 100% for VT), and the areas are defined by the labels 'VT' and 'LSU.' As stated above, the square to the left of the 'Series Record Wins' label is entirely unnecessary.
[edit] Redesign
Though not in this case because of the lack of data, I think it would be useful to include time data for series history, to give a sense of recent and historical trends. For this reason, and in light of this visualizations accompanying a table of text (LSU's schedule), I have used Tufte's sparklines approach.
In an in-line time series, ticks above the line are LSU wins, and ticks below the line are LSU losses. LSU's loss to VT came in 2002.
With such an in-line presentation, data for the series between LSU and each opponent (or at least more than one) could be highlighted in this way. I have included LSU's series record against Ole Miss, a team LSU has played most years since 1900 to demonstrate an idea of the data density possible with this visualization. Tick marks that span the x-axis indicate ties. Gaps indicate years in which the teams did not play.
The data used to populate this redesign come from Stassen.com College Football Information.
[edit] Good Visualization
Source: The New York Times" (August 12, 2007)
[edit] Description
This map from The New York Times presents the aggregated number of terrorist attacks in Afghanistan since January 2002. While not ideal in every dimension, it illustrates two techniques well. First, the graphic presents information organized geographically (by province), and placing the information directly on a map echoes that organization. Second, the circle sizes vary appropriately. That is, for a five-fold increase in data there is a corresponding five-fold increase in a circle's area, not radius, thus avoiding one of the pitfalls described by Tufte. Annotations are used to show clustering among provinces and to describe the percentage increase in attacks between 2002 and 2006. This is somewhat effective, but could be better, since it is this increase that seems to be focus of the article, given the graphic's title.
[edit] Deconstruction
The data model includes the number of terrorist attacks (ratio), time (interval), and the names and geographic outlines of provinces (nominal). The image model consists of circles whose areas grow in relation to the aggregate number of attacks in a given province. Color and transparency are used to show overlapping of data-circles, and do not correspond to variables in the data model. Time (for calculating attacks per year) is not mapped onto the image model, and is only addressed in the various captions. Perhaps dotted lines inside each data-circle--representing at least 2002 and 2006--could have been added to graphically show the increase in terrorist activity discussed in the captions.
