A1-DavidSun
From CS294-10 Visualization Fa07
Contents |
[edit] Good Visualization
[edit] Explanation
This visualization shows the best investment option out of a choice of 32 investment classes. The options are organized in using a binary tree with the best option at the root of the tree. All the options appear as the leaves. The diagram compares the options on a pair-wise basis and selects the better opinion for the next round of comparison. This organization of information allows the user to quickly see the outcome of a complex decision process without the need to stepping through it. Although the comparisons are only between pairs of options, the reader can easily use the result of the next round of comparisons to compare arbitrary pairs of options (I take this back: it's clear that you can rank any internal node with its subtrees but nothing can be derived between elements on the same level but belong to different subtrees). For example, Japanese Equities, being the winner between its match up with Latin American Equities must also be the winner in comparison to Asian Equities and Preferred Stocks.
[edit] Deconstruction
The investment options are nominal variables. The relative strengths between these investment options is encoded within the structure of the binary tree: the root of the tree is the best option; an internal node is better than the options residing on both the left and the right subtrees. The relative ordering exhibited by the binary tree is a strict partial order. We cannot totally order the investment options based on the information given. Hence this ordering may or may not qualify as an ordinal quantity depending how strict you want to be.
[edit] Bad Visualization
[edit] Explanation
This particular visualization is attempting to communicate to the reader the subjective ratings of a wine critic on Cabernet Sauvignon originated from the Napa Valley over the past decade. The critic also gave commentaries on the particular taste of the wine in question. The goal of the original article was to give an expert's opinion on a popular yet expensive product. This visualization appeared in Business Week (June 18, 2007).
[edit] Deconstruction
The visualization includes numerical ratings (quantitative variable) of Napa Cabernet for each of the past 10 years, i.e. from 1997 - 2007. The ratings are made on a scale from 1 to 100. To simplify the interpretation of these numbers, they are mapped on to a set of ordinal rankings: Extraordinary (96-100), Outstanding (90-95), Above average (80-89), Average(70-79), Below average(60-69), Unacceptable (<59). To assist in visually grasping the meaning of the ratings, the ordinal rankings are encoded using different colors. The rankings are augmented with information about the taste of the wine. These commentaries are organized into 5 groups whose descriptions are encoded by the first letter of the word in the sentence (nominal). For each year, the wine, its ranking, and flavor information are clustered together and depicted in the shape of a grape.
[edit] Critique
There are number of problems with this visualization. It attempts to organize the data points as a batch of grapes in an attempt to draw a clever connection with the subject of the article. However, this metaphor was not obvious (at least upon the first couple of viewings) which made the organization of the data very unnatural to look at. Additionally, the numerical ratings and colors were confusing. Again, it wasn't obvious how to interpret relationship between the coloring and rating numbers until very inspection of the legend. The additional encoding of flavor information as characters beside the numbers didn't help either. Additionally the flavor encodings are almost arbitrary as it was based on the first character of the phrase.
[edit] Redesign
The redesign is aimed to achieve the following goals:
- Find more consistent encoding for the original data.
- Reduce the cluttering around the legends.
- Hopefully reveal any trends not readily perceivable from the original.
The redesign consists of plotting the original set of ratings against the years in which the corresponding wines were produced: the years are mapped on to the x-axis and the ratings are mapped on to the y-axis. A scatter plot was chosen over a line plot due to the discrete nature of the year variable on the x-axis and the small number of data points. The flavor of the wine (nominal) is re-encoded by employing different colors, which is reflected on the data markers.
This new design is better than the original in several ways. Firstly, we can actually see an interesting trend in the form of a dip in the ratings from 1998 to 2000. While there could be many different interpretations of this pattern (especially given the ratings are very subjective), this trend was not obvious from reading the original visualization. A second improvement is in the form of using easily discernible color encoding for the different tastes versus the previously arbitrarily choose letters of the alphabet. The new design gets rid of the qualitative ranking (Extraordinary, Outstanding, Above average, etc.) found in the original since it was somewhat redundant given the numerical rankings and helps to reduce the cluttering of legends.
