Our first regular Higher Ed Data Central post gets a bit wonky (but don’t worry they won’t all get this deep into the weeds).
Previously, we posted a graph showing the number of highly compensated administrators per 1,000 students as compared with tuition. Over at the Pearson blog, Kristen DiCerbo suggested using a logarithmic curve instead. In fact, she even rewrote the code to add the curve, yielding the chart below. As she suspected, the logarithmic curve provides a better fit as the universities on the far right no longer appear to be outliers.
As an aside, we ran into a bit of a puzzle in R. When we used the scatterplot function in the car package, changing boxplots=’xy’ to boxplots=FALSE (to make room for some labeling) would change the intercept of the curve when it was plotted, even though the curve command didn’t change. I ended up using the plot function to get around this problem, but if anyone out there knows how to avoid this issue in the scatterplot function, do get in touch.
On a separate note, Twitter user @dvppraxis requested the names of the universities in our previous chart showing the relationship between tuition and U.S. News & World Report rankings. We can put the names in easily enough, but as you see in the chart below, they are too bunched to be able to read most of them.
Separating the publics and privates mostly solves this problem for the public universities, as you can see below.
But tuition varies too much at the privates for this to work, so I had to cut off the schools with low tuition (the blue schools towards the bottom in the second graph above – Brigham Young U, Howard U, Sanford U, etc.)
Don’t forget to send in ideas for what you’d like to see us investigate in future posts by leaving a comment here or sending us an email.