A1-ChristopherVolz
From CS294-10 Visualization Fa07
Contents |
[edit] Bad (but not terrible) Visualization
Source: The Atlantic 300 no. 2 (September 2007)
[edit] Explanation and Critique
The above image (taken from the theatlantic.com -- attempts to scan the image proved problematic due to the graphic spanning two pages) has a number of good visualization qualities but its overall design serves to confuse the reader more than make the data clear. The greatest problem is the use of two different forms of visualization for the data the chart is attempting to compare. Arguably, since the primary goal of this chart is to show the relative numbers of government documents classified and declassified in relation to each other then having those data points plotted along different axes results in a near total failure of the chart, and there is no real reason to do so since both data sets are in the same scale (millions of documents). The fact that the bar graph obfuscates the line graph in areas further reduces the utility. One can only assume it was designed to be stylistically engaging and not meant for close scrutiny. Another complaint is the use of shading between the x-axis (time) and the line-graph. It doesn't appear to correspond to administration and tends to distract the reader from the information rather than make it more clear. If the bar graph were replaced with a separate colored line graph and the shading removed, however, this would probably be a very good chart as you could directly compare the numbers of documents classified with those that were declassified along a time axis and presidential administration. The use of labels to highlight significant data points is also useful.
[edit] Deconstruction
This chart is working with a large number of related data sets. It includes the number of government documents classified, the number of pages declassified, the passage of time, the administration in office at the time, and a number of explanatory labels either providing additional information or revealing specific data points in the chart. The values are arranged in an interesting format with the number of documents classified plotted against time using a line chart and the number of pages classified plotted against time using a stylized bar graph along the z-axis. The use of shaded blocks is used to encode the administration in office against time and acts as a backdrop for the line denoting the number of documents classified. An inherent flaw in this chart, however, is that to a certain degree, it is comparing different measures of the same thing. That is, one data set is the number of documents classified over time. Another data set is the number of pages declassified over time. But there is no correlation between pages and documents (though I would hazard that a representative sample could be taken from the number of pages declassified and then the data could be normalized against that). In order to mask this contradiction the document utilized three axes for measurement which not only made it difficult to compare the values directly, it also hid the fact that the values shouldn't be compared directly.
[edit] Redesign
This redesign kept a number of the elements of the original but instead plotted the data sets (pages and documents) along the same axis to show their values relative to each other over time. Again, since this is a measurement of two related but non-identical data sets the comparison is not as honest as it could be if the data were better. A partial grid demarcates each year, but the horizontal rule-lines were intentionally left out, largely because the values from the original chart weren't explicitly provided, I had to eyeball the values for my redesign. Arguably, the vertical rule lines could be eliminated in the upper portion of the graph, though when I experimented with that it looked funny. The time and quantity scales were both kept to consistent increments, the only deviation being the upper limit on quantity so the viewer could better estimate the value of the upper portion of the graph. The administrations in office were explicitly labeled and changes in administration were further encoded by the use of a different shade of background color.
[edit] Good Visualization
Source: The Atlantic 297 no. 2 (March 2006)
[edit] Explanation and Critique
This chart is relatively simple and thus probably difficult to screw up. It tracks changing trends over time of where people with bachelor degrees choose to live in three broad locations: downtown, suburbs, or city. On the positive side the graph is very clear to interpret and clearly reveals the broad trends associated with study; namely that people with bachelor degrees are increasingly moving downtown. The real negative of this graph, however, is that while it reveals broad trends it does little to provide insight into the data itself.
[edit] Deconstruction
The data set for this visualization is the percentage of residents (y-axis) in three urban areas (City, Downtown, Suburb) who have received a bachelor's degree plotted against time. This data is then plotted as three separate line graphs using color as a key differentiator between the three plots. The passage of time is encoded along the x-axis through the use and positioning of labels and, similarly, the percentage of residents with a bachelor's degree is encoded along the y-axis. Each plot is separately labeled. What is left unclear in the graphic, however, is if there is any overlap in the data sets between the three areas of residence (that is, is "downtown" also part of "city"?)
