Ivy-equivalents (ranking based on alumni outcomes) take 2.1

With respect to the numbers in post 87, can we pull in a few more universities, and also consistently use rates rather than raw numbers to compare both doctorates and awards?

If you looked at the top 10 universities, or top 50, by percentage of students who earn PhDs, you would not find that half of them are public universities. It would be more like none in the top 10 and maybe 3 or 4 near the bottom of the top 50. Example: http://www.nsf.gov/statistics/infbrief/nsf13323/, see table 4.
Berkeley (#43), CO Mines (#47), and William and Mary (#50) are the only public research universities in the top 50 (by “institution yield ratio”, for STEM PhDs only, for 2002-11 only ).

Now maybe we should only be comparing only schools of similar size (undergraduate enrollment size?) or schools that exceed some output threshold. Alumni from 9 research universities listed in that NSF table earned >1000 STEM doctorates from 2002-11. Here’s the rank by “institutional yield ratio”:

MIT 16%
Princeton 10.1%
Harvard 10%
Stanford 8%
Brown 8%
Yale 7.8%
Cornell 7.7%
Duke 7.1%
Berkeley 5.9%

With the possible exception of Cornell, the top 8 are all more selective than Berkeley. If we wanted to compare only research universities of similar selectivity, maybe Alexandre’s list in #87 is a good comparison. But have a look at Washington Monthly’s “Bachelor’s to PhD Rank” (http://www.washingtonmonthly.com/college_guide/rankings-2015/national-universities-research.php). There, the highest-ranking public university (Berkeley again) is 17th and Michigan is 28th (very close to their USNWR rankings). UCLA is 32nd, Wisconsin 33rd. T30 schools include the Ivies, Caltech, MIT, Stanford, Chicago, and JHU … but also Rochester (13th), CMU (15th), Case Western (16th), and Brandeis (18th).

Do award and scholarship distributions tell a very different story than the PhD rates?
To compare, I think again we’d want to use rates not absolute numbers, for an aggregated critical mass of awards/scholarship data representing a similar time period. Isn’t that what PurpleTitan already has done for us?