Displaying ideal points I
I spent almost an entire day working to produce this figure. That’s right, a full day of work to produce this entry in the blog! (higher qualiy pdf version here.)
I got motivated by reading two papers: Scott Desposato’s forthcoming (AJPS) on party switching in Brazil and Andrew Gelman et al’s American Statistician article on turning tables into graphics. Highly recommended reads.
I try to convey a lot of information with the figure. It is the same ideal point estimates I referred to in this other post The lower dots (red) reflect the estimated ideal points of the deputies by the party membership recorded in his/her first roll call vote in the session. The upper dots correspond to the last party membership recorded. The bars reflect where the central 50% of the distributions. Finally, the numbers after the party names (by the way, S.part. means without a party) are the number of deputies using the first measurement of party membership and, in parentheses, the increase (positive) or decrease (negative) in the size of the party. The black dots reflect the positions of the deputies specified at the line below each party.
The dashed line, as you probably have guessed, plots the estimated median of the legislature.
In the next post I will discuss some of the patterns that can be found in the figure that are related to Desposato’s arguments about party switching.