Carolyn's NRC-PGR data

I find Carolyn's post on the 2006 NRC data quite fascinating. (Remember to click on her diagrams to bring up a bigger version.)

I take it that the S-ranking reflects actual performance (however that might be measured) with regard to factors that are explicitly valued. And the R-ranking reflects not exactly reputation, but actual performance in factors that are found to influence reputation. So to some extent, R is appearance (actually predicted appearance: high performance in what makes for reputation) and S is reality. Of course, the measurements and analysis may be rubbish, so one has to take the whole thing with a grain of salt.

Still and all: there are some strange results . . .

For instance, assuming that the numbers on the two axes are ordinal rankings, then it looks as if there are big discrepancies between reality and  (idealized) appearance. Boston University is 10^th in reputation-explaining factors but 45^th in the explicitly-valued factors. I am not entirely sure what to make of these measures, but that is sure a big discrepancy. On the other hand MIT is also 10th in reputation making factors but 5th in explicitly-valued factors. That's a huge difference between MIT and BU. And there are other such discrepancies that I find pretty interesting.  Ohio State: worse than 60^th in explicit value, but a little worse than 30^th in reputation making values. CUNY worse than 80^th in explicit value; more or less 50^th in R.

What’s just as interesting is the discrepancy between what explains reputation and actual reputation (as measured by PGR). For example: CUNY (I think) which is 50^th in reputation making factors turns in at 23^rd in actual PGR rank. (And don't forget it is 80^th in actually valued factors.) UCLA, which comes 30^th in reputation-predicting factors, earns a 7^th rank finish in PGR. USC, predicted at worse than 50^th, actually comes in at 16^th.  The four departments all predicted to come in 10^th (or so), actually come in at 5^th, 7^th, 16^th, and 50^th. Same prediction, very different results.

I would not mind seeing the R rank plotted against PGR rank to see how good the fit is. But to repeat, there are certainly some big discrepancies.