A1-Jimmy

From CS294-10 Visualization Fa07

Jump to: navigation, search

[edit] Good Visualization

Image:chen_good.jpg

Explanation
This visualization for the US presidential polls is taken from Newsweek. It visualizes the poll of the potential republican candidates for the 2008 presidential election. The data dots are the poll values on daily basis, which are given different colors for different candidates. The lines, which are average of the data, show the poll trends of each candidate. The graphs in the right visualize the polls divided by different states.

Deconstruction
The data model can be seen as a relational model with a set of records. Each record, which is visualized as data dots in the graph, has 3 attributes: poll value, time, and candidate. The image model visualizes the trivariate data mentioned above. The time is encoded in x-axis (quantitative), the poll values encoded in y-axis (quantitative), and the candidates are encoded in different colors (nominal). The poll trend is encoded in lines with different colors that are the average values of the data dots. In addition, the smaller graphics in the right are divided into different locations. The time and location are encoded in text labels on top of each graphics. The size of data set is the number of data entries (tuples) in the relational model. Assume the polls are done daily from December to August, and there are five candidates. The size of data set would be approximately 30 X 9 X 5 = 1350. The uninformative element in this graphic is the grid. However, it could help clarify the data by segmenting the graphic by month and values. And the color of the grid is light gray that wouldn't obscure the data in the foreground.

Critique
The graphics are clear and informative. The lines and dots are marked with different colors, which makes it clear to compare the supports of different candidates. Another good thing is the use of lines rather than data dots. The lines make it easy and clean to recognize each candidate's support in a single graph, without mixing up with the dots. And the lines could also visualize the poll trend. In addition, it is good to make the color of grids as light gray. This clearly segments the graph with time and values, while not obscuring the actually data in the graph.

It's also good to have additional graphs for the polls in individual states, as shown in the smaller graphs in the right. The minor problem for the small graphs is that the date information is not clear enough. We don't know whether the grids are by month, day, or anything else. It'll be better to put the date in the bottom of the small graphs.


[edit] Bad Visualization

Image:chen_bad.jpg

Explanation
This visualization for the stock data of the five companies is taken from San Jose Mercury News. It visualized the fluctuation of stock values of the five companies in the past three months. The labels in the right of each stock chart show the closing value of the day.

Deconstruction
The data model for each stock chart is a set of tuples with two attributes (time, value). The image model visualizes the bivariate data mentioned above. The time is encoded in x-axis (quantitative), and the stock values (quantitative) encoded in y-axis. The company name (nominal) is encoded on top of each stock chart. In addition, the latest stock value and the fluctuation rate is encoded in the text label on the right of each graph. The data set is the stock values of five companies from 5/29 to 8/29. So the size of data set is approximately 90 X 5 = 450. (assume the values are sampled on a daily basis) The uninformative element is the horizontal lines that segment the chart by stock values. However, these lines are useful as they make it easier for viewers to know the stock values more precisely at each point of time.

Critique
Although there's no problem if we look at each stock chart separately, it's misleading when these charts are put together. As these charts are aligned next to each other, the viewers would tend to compare the stock trend of these companies. However, the problem comes from values of y-axis: the scale and range of the stock values for each chart are not consistent. This makes it misleading when we compare the stock values based on these curves. For example, the range of stock values in the LSI chart is from $4 to $10, but that of PDL Biopharma chart is from $10 to $40. If we compare the two companies based on the curve itself, we may have the impression that the two companies are similar in stock values, while they are actually very different. Such design variation is not a good style of visualization. As readers rely heavily on the visualization to compare the stock values, the scale and range of the charts should be consistent so as to show what the data really is.

Redesign Image:chen_redesign.jpg The visualization is redesigned in two parts: the scale and range of the values in y-axis, and the positions of the graphs. Since the original graphs have different scale and range in the y-axis, it is difficult to compare the stock values of the companies. Moreover, the vertical alignment of the graphics also makes it difficult to compare the stock values. So the new graphs are redesigned to unify the values in y-axis, from $0 to $35. And the positions of the graphs are rearranged so that they are aligned horizontally. This redesign would make it easier for the readers to compare the stock fluctuations of the companies because the horizontal alignment let the viewers easily compare the stock values in y.


[add comment]
Retrieved from "http://vis.berkeley.edu/courses/cs294-10-fa07/wiki/index.php/A1-Jimmy"