There’s More Than One Way to Skin a Dataset
Last week I asked if you could improve a mediocre bar chart showing party majorities by county. There was a resounding yes as many of you deconstructed and then reconstructed your own graphs. For reference, here's the original chart:
Here are the key flaws to the original that you all caught:
- The x-axis tick marks were in really weird places;
- The y-axis label was misleading because the data were number of counties;
- Red and blue would make more sense for Democrats and Republicans;
- Counts for counties don't match the years, because they are reversed;
- We see a different story when we bring in data for undecided "other" and "declined to declare."
What was the graph trying to show? It was trying to show party registration in California over the past five presidential elections. Did it succeed? No. It failed miserably; however, you did much better. Here are all the reworks.
Brijesh made a stacked chart for Democrats and Republicans:
Tyler made a horizontal stacked bar chart with a useful majority line down the middle:
Blair provided some R code:
David used a tornado chart, which turned out well:
Amos went with a stacked line chart:
Kevin sent this one in:
John put together a few versions - this being one of about five:
Jorge went with simplicity:
Stack created a time series for the Dems and Reps:
Jake put up a fan favorite:
Nate, the graphic designer, embedded a stacked line chart inside the California boundaries:
This is the one I made at the workshop:
Personally, I like Jake and David's the best, but who gets the golden star for best graph? I'll let you be the judge.