NRC rankings are getting closer...

The methodology report for the forthcoming NRC rankings is out. It's an interesting approach to ranking programs, but it's complicated. For those without a statistics background, I'm sure it's conceptually pretty weird.

In the short term, the NRC rankings have two weaknesses: first, they're quite late, and second, the algorithm takes a fair amount of work to wrap your head around (and I say this as a math PhD with a reasonable stats background). Most of the reactions I've seen to date harp on precisely those two things.

With a longer view, however, things may be different. In the best case scenario, people will eventually figure out the mechanism - maybe it will even serve to educate people about the many benefits of resampling. If that happens, US News-style rankings may come out looking like gross oversimplifications. Also, a lot of the reason for the delays is that this type of approach has never been used before. In a repeat of the process, the NRC won't have to go through the exercise of generating importance weights, which surely was much of the work involved. We shall see.

In the meantime, this quote from the Chronicle stood out:

In an interview this afternoon, Jeremiah P. Ostriker, a professor of astrophysical sciences at Princeton University and the chair of the committee that oversees the doctoral-assessment project, said that he does not expect the report to be released during the next several weeks, but that he would be surprised if it is not released by the end of 2009.

Ouch!