The New York Times analyzed the first and second night Democratic debates by asking columnists and contributors across the political spectrum to rate each debater on a 1-10 scale (10 being highest) and then averaging each candidate's values. I analyzed all of scores using Microsoft Excel to find the standard deviations for each candidate. Since standard deviation provides the average distance of all the values from the mean, a candidate with a high standard deviation represents a high variance in opinion among the commentators. In other words, the higher the standard deviation, the more the pundits disagree with each other. What? How do you all relax?
My wife challenged me to create "a pretty graph that visualizes both mean rating and standard deviation, with a minimum of three colors." She was trolling me about my PhD thesis advisor who always admonished us to make our graphs and charts more visually appealing. This was not always easy to do when you were discussing plant Golgi body morphology, but, I digress.
I took on her challenge. Not as easy as it sounds. Converting hard data to a visually compelling story-telling format is more art than science. My first attempt was shot down both by my wife (not enough colors) and my son (improper use of a line graph to illustrate discrete variables). They were both right, dammit.
So, here is my second (and hopefully final) version.
Just to stuff a little more data into the graph, I color-coded the bars “most progressive,” “mid progressive,” and “least progressive,” based on a recent Business Insider reader poll. Bill de Blasio and Marianne Williamson were not ranked in the poll, so I used my best guess. De Blasio is coded blue since he at least tries to portray himself as the most progressive candidate. Marianne Williamson is rainbow because she is Marianne Williamson. In plotting the standard deviation range instead of simply the mean, my goal was to illustrate the variety of opinion for each candidate. This is the advantage of averaging multiple experts instead of relying on single opinions. I’m sure Nate Silver would agree.
Also note that Williamson had the highest standard deviation, followed by Delaney, which means that the pundits argued the most on how to rate them; they either loved them or hated them. Bill de Blasio, on the other hand, had the lowest standard deviation, which means that the pundits were united in their opinion that he did horribly.
Finally, I fully realize that comparing all the scores together is a clever fiction. Since the candidates did not debate on a single stage, comparing Biden’s score to Warren’s score (for example) is not empirically valid. Nonetheless, the average score is a useful way of illustrating how well each candidate on “the debate stage” in general.
No comments:
Post a Comment