interesting piece of analysis by orion weiner using ucsf data link here. it is fantastic that someone was able to do this, and it confirmed some of my suspicions that i've had from sitting on the stanford bio dept grad admission committee. i would posit another way of interpreting the results. it isn't that any of the scores of the admitted students (except number of years of research experience and subject gres) are useless, but that at the top programs, like ucsf, the students scores are above where they start being predictive. another case is nih grant scores - if i recall correctly, there is no correlation with grant score and productivity (sum of impact factor points or some such metric), up to a point, estimated around 30% or so. after the 30th percentile, things get predictably worse though... i suspect something similar is going on with our students, where beyond the 30th percentile things would get predictably worse. so, all in all, we may be doing as good a job as we can on the committee. yeah! perhaps we can do it faster next year...
Posted