The Truthful Art — Methods of Encoding & Exploring Data (Ch. 5&6)
When reading these chapters, I could not help but wonder — if choosing the right method of encoding the data is crucial to effectively interpreting a visualization, then why do so many people use pie charts? It felt like the most commonly referenced chart type in grade school, and it still seems to be a go-to in annual reports and professional presentations. Though perhaps I should not be so quick to judge or dismiss the pizza-shaped art form. One of the major takeaways from this chapter is that there are appropriate uses for all methods of encoding data, depending on the goal of the visualization. And a pie chart, comparing two amounts, can be reasonably effective at demonstrating a comparison.
I was somewhat surprised to read the criticism of radar charts. Particularly because they are increasingly trendy as a way for “stat nerd” sports fans or writers to present information about players. It’s most often see when comparing players before the NFL draft. Top players are invited to a scouting combine, performing various feats primarily used to measure different aspects of athleticism (40 yard sprint, standing jump, bench press, etc.). These players are then captured by a radar chart comparing percentiles within each position. Each year I might look through some of these, particularly for Miami players looking to make it to the league. With the initial example used in the book, I completely agree that it was not a particularly useful method of encoding for making comparisons. I feel like it does improve in a more interactive setting, like in the Mockdraftable site.

As someone who has (mostly) successfully navigated a graduate program in statistical modeling, I’m more than familiar with the premises of exploratory data analysis. It is common nature to run descriptive statistics on a dataset as the first order of business. I actually tend to do it twice: once to check for quirks or obvious outliers (a distorted mean or range, for example), and a second time after data cleaning to get a sense of the data with which I am working. And I occasionally will plot out a histogram or scatter plot before doing anything else. But this chapter has me reconsidering why I don’t do it every time. For one, I think it makes it easier to catch outliers that don’t show up in a normal list of descriptive statistics (though R does have some good functionality for determining outliers, you just have to be step outside the boundaries of base R packages). Using some of the tools we are learning in class (e.g, iNZight) is something I will likely start incorporating into my routine when approaching new datasets.
