Discussion on Graph Design I.Q. Test
Perceptual Edge
We are overwhelmed by information, not because there is too much, but because we don’t know how to tame it. Information lies stagnant in rapidly expanding pools as our ability to collect and warehouse it increases, but our ability to make sense of and communicate it remains inert, largely without notice.
Computers speed the process of information handling, but they don’t tell us what the information means or how to communicate its meaning to decision makers. These skills are not intuitive; they rely largely on analysis and presentation skills that must be learned.
Perceptual Edge focuses on the tools and techniques of visual business intelligence to help you make better use of your valuable information assets.
Here’s the beginning story why I roamed into the graph design test. In last fall, I took the class called “Data Visualization” by Professor Nazzaro who has inspired me a great deal to pursue data science career. We critically discussed the types of bad and good graphs based on Tufte design principles. Tufts believes that the visualizations should tell the story in their simplest manner, without the need to be highly decorative and unnecessarily colorful. In fact, using the color in the graphs needs to be carefully done because while the graphs look visually appealing, the story you intend to convey may be incomplete. Henry David Thoreau seems to validate Tufte’s idea of keeping it simple as well.
What went wrong with pie chart? Studies have found that it’s more difficult for people to judge differences in area, such as slices of pie, than it is for them to judge differences in length, such as the lengths of bars.
Compared to the simple bar graph, Pie chart does not distinguish the area of investment portfolio breakdown, for example, the area of international stock is almost the same as large-cap stocks, thereby not noticeably clear. Suppose the value labels were added, pie chart would still not be as consistent as bar chart. No further color is needed for bar graph in breaking down portfolio investment.
2D vs. 3D line graphs. It should make sense to us that 2D line graph is easier to understand as time progresses. 3D graph over-complicates the visualization by taking into account of volume, which does not exist in the dataset. Therefore, we should stick to 2D instead of 3D and always keep in mind that we’re not living in a gaming world!
The grid, fill colors, unnecessary precision, and redundant use of the dollar signs in the top table all distract from the data and make it unnecessarily difficult to read and compare values.
When you want to see the change/behavior over time, line graph is preferred to bar graph. This allows you to identify the peaks, though, and rise and fall as time continues.
Graph A lies. Because the bars start at 2,000 instead of 0, their respective height differences have been greatly exaggerated. We need to make sure that y-axis is correctly and incrementally adjusted and starting from 0.
Map A makes use of color gradation, while Map B randomly picks out colors that confuse the story.
3-D effects make graphs harder to read and can hide some of the values altogether, through occlusion. The “small multiples” display allows for easy comparison of all of the bars. Ask yourself when you look at 3-D bar graph, whether or not you can tell the expense in travel of R&D sector.
The distinction here is how the xy-axis are presented. As a reader, I believe you do not want to turn your head 90 degree to learn what the graph is telling you. To keep it simple, one should have xy-axis be horizontal.
The above graph excessively uses colors in describing the median employee salary by Department and state. The background color does not provide any additional information, so Tufte’s principle suggests using white background color instead of colorful one that does not add anything.
Issue of wrong choice of coloring appears again because the below graph does not tell anything about the green color. What does yellow and red represent? Only if you want to emphasize or highlight a particular value should you color it.
Key Take-Away
1.) Perpetual edge complied some common examples of mistakes in data visualization, and it reflects that misleading data visualizations are quite common and prevalent in the media. We need to be critical and inquisitive to criticize the graphs, find ways to better convey the ideas, and encourage a feedback loop as well as peer review.
2.) Complex graphs at times mislead the story; in fact, Tufte principles suggest that visualizations should be simple and straightforward to the idea rather than visual appreciation.
3.) It is important to use your judgment in deciding which graph is appropriate in certain context. Knowing types of variables — quantitative or categorical- help us choose the appropriate ones, for instance, the relationship between two quantitative variables (e.g. SAT score and GPA) is best shown by scatterplot. Another example deals with a univariate categorical variable (e.g. a breakdown of hair colors). Particularly, there are many visualizations that account for a single categorical variable like pie chart and bar chart. Hence, other factors will come into play such as xy-axis, color, 2D vs. 3D, grid, and precision etc. Please refer to http://support.minitab.com/en-us/minitab/17/topic-library/basic-statistics-and-graphs/introductory-con cepts/data-concepts/cat-quan-variable/for further explanation on the difference between categorical and quantitative variables.
