From CS294-10 Visualization Fa07
 "Good" Visualization
- This visualization comes from an article in the most recent National Geographic which discusses the variation in cave species of a number of kinds of organisms. These images, specifically, are intended to illustrate the gradation of ocular sensors found across species at increasing distances from the cave entrance. As such, this increasing distance is implicitly plotted along the X axis. The images themselves are photographed from a similar angle and arrayed so that the exemplars of each of the three species can be directly compared to one another. Captions on each image identify each species and provide information about their locations within the cave and ocular ability. This juxtaposition of the images and their ordering immediately provides context and allows a viewer to observe features of each of the species (e.g. the appearance of eyes and vibration-sensing hairs and variations in translucency) and associate them with increasing depth. This allows the visualization to convey fairly powerfully the effect of cave depth on speciation without introducing any sort of visual axis or other extraneous visual marks.
 "Bad Visualization"
- This visualization shows the number of days between theatrical release, the delivery of Oscar screener copies, and the appearance of an illicit DVD-quality copy online, for the past five winners of the Academy Award for Best Picture. The visualization plots time in days along the X axis. A pink bar indicates the number of days until the appearance of an illicit copy and a purple bar indicates the number of days until the release of Oscar screeners. As a caveat - this graphic comes from the "Infoporn" section of Wired Magazine, which generally features visualizations that are somewhat exaggerated. By design, these charts often make use of extended visual metaphors and feature elaborate visual flourishes. This chart, however, features a number of design decisions which simply make it difficult to parse. For example:
- The chart's X axis plots time in days. However, this is only indicated once, inside of the top bar in small type. Dark and light shading in the bars themselves show weeks, but this is never indicated. - The bar in the fourth example bends and wraps making visual comparison of the bars difficult.(The pink bars, however, each remain on one line and it appears that the chart was scaled to their length.) - The overlap between the pink bars indicating time to the appearance of an illicit copy and the purple bars indicating time to the release of Oscar screeners is ambiguous. This makes it difficult to distinguish between cases when Oscar screeners were never delivered (e.g. 2004) from ones where they were delivered alongside the theatrical release (e.g. 2005). - The inclusion of data for only 5 films - the Best Picture winners for each of the five years - does not seem to be enough to support its case - that Oscar screeners are generally not responsible for illicit DVD releases.
The data contains four five components. The name of the film (nominal) the year it was awarded Best Picture (ordinal) and the dates (quantitative) on which 1) the film was theatrically released, 2) Oscar screeners for the film were release, and 3) a DVD-quality version of the film appeared on the internet. More important than the specific dates are the intervals between them, specifically the intervals between the theatrical release and the release of the Oscar screeners and appearance of an illicit copy. All of this data is depicted as univariate with each variable having only one value for each film. The dataset is very small, containing only 5 tuples.
The data is laid out as a stacked bar chart in which X position encodes time (in days) after the release of the film. Bars extend from the left baseline and terminate at the date when an Oscar screener is released. Color differentiates the portion of the bar representing time before the appearance of an illicit DVD-quality file from the portion representing time after its release. Labels at the ends of each bar and at the point where the illicit copy was released show the specific dates as well as the interval (in days) between that date and the film's opening. Films are listed in a descending Y order according to the year in which they were awarded the Oscar for Best Picture.
A second small graph in the upper right-hand corner presents the average time from theatrical release to online debut for all Oscar-nominated films. This small chart contains only two values per tuple, a year (ordinal) and the average time from theatrical release to online debut for all Oscar-nominated films in days (quantitative). This data is also univariate and is displayed as a column chart with years arranged left to right in ascending order along the X dimension and time in days in the Y dimension.
A few strange chart decisions, discussed previously in Assignment 1A, render the chart somewhat unwieldy. The winding bar on the fourth example is the most obvious offender, since it renders visual comparison between the bars more difficult. The decision to color code in such a way that bars where screeners were delivered immediately appear identical to bars where they were never delivered also seems dubious. More importantly, however, the chart doesn't provide sufficient evidence to support the accompanying blurb's claim that Oscar screeners are not typically the source of illicit DVD-quality copies.
 Revised Version
My revised version of the chart introduces a formal x-axis with gridlines and eliminates the curvaceous bar for 2006. This allows more accurate visual comparison across years. The stacked bars are replaced by grouped bars to help differentiate time to illicit appearance from time to screener release. The actual dates, while somewhat interesting, aren't relevant to the central point the chart is designed to convey and have been removed. Additionally a fading bar and text notes have been added to help differentiate the case in which screeners were never released from the case in which they were released alongside the theatrical premiere (although the decision to include a bar in the absence of data is certainly debatable).
The resulting graph is still fairly problematic. It sacrifices some of the information delivered by the original, but still monopolizes the same amount of space - probably far more than it needs. Additionally, the graph still shows far to little data to clearly make its case. Whether or not a bar chart is actually the best way to represent this is also questionable. A timeline with interval marks might be more appropriate.