Tuesday, May 1, 2007

Meta-h index rankings

A couple of us have been trying to make a more objective (note that does not mean more accurate!) ranking of CS research departments. We're using the google scholar data combined with the h-index. Still lots of mistakes I am sure. Interesting that Greenberg and Hanrahan are in a dead heat for the highest rating in graphics. Not worth drawing too much conclusion from, but the data is interesting. I'd like to add more universities and research groups, so feel free to send me updates and suggestions.

The data is here

5 comments:

anonymous said...

The h-ranking seems reasonable for individual researchers, but I think it quickly loses meaning at the meta (i.e., university) level.

For one, it throws out information that could be used to break ties -- MIT and CMU are both ranked #1, but if you look at the mean and median h index for researchers at each school, MIT should come out ahead (MIT:23.2662/21, CMU:21.7239/18).

And then there are schools like Princeton whose meta-h index rank isn't very high (#12), but whose mean/median h-index bests even the #1 schools (Princeton:27.4286/26). In general the meta-h index rank seems to favor schools who have a lot of superstars over those with good quality overall.

The problems mentioned so far are fairly subjective, but here's one that's hard to ignore: the size bottlenck. A school can never have a higher meta-h index than its population. Take a look at Caltech (#46) and Rice (#35), which have meta-h indices of 13 and 14, resp., but whose populations are only 14 and 16, resp. Again looking at researchers' average h-indices we have Caltech:23.7857/20 and (more amazingly) Rice:41.75/46.5!!

I realize that the h-index is a measure of quantity as well as quality. However, quantity is clearly less meaningful for universities (among whom size varies) than for individuals (among whom size is constant... well, more or less ;).

Peter Shirley said...

I do not disagree with any of these points, with the possible exception of the h-ranking being more meaningful for individual researchers. It rewards "co-author weasels" as well as the trend-followers. It also would punish departments full of youngsters which would likely be much better than past-their-primes such as myself. And there are more problems besides. I do think the main merit of the meta-index is that non-US schools can send me their list and get immediately included. As these strong departments evolve outside the US, they can get ranked by something besides "USNews".

Note that both Rice and Princeton do very well on our "participation" measure. But they do raise a philosphical issue. Should a department that just has Pat Hanrahan be ranked the best in the world? Perhaps. Would one with 20 people of that caliber be better than one with 10? No easy answer on where the diminishing return of size behaves.

Thanks for your comments.

Peter Shirley said...

Also note. If some enterprising people from departments not listed would send me a file for their department in the format listed on the page, I would be happy to include them.

Unknown said...

A-index

After h-index is calculated multiplay it by the average number of citation of the papers contributing to h-index, and than take square root. This gives area based(A-index) of the h-index citation field. This removes problem that h-index of N can not distiquish is it N-times N-citations (the weakest posible case)or N-times of citations that are much larger of N. For instance h-index of 5 one can get by citations: 5 5 5 5 5, but also by citations: 50 5 5 5 5. A-indeks in this two cases is 5 and 8.4 respectively. Such an idex will put science genius (mailstone discovery) back to the top.

Nenad Juranic, Ph.D

Unknown said...

A-index

After h-index is calculated multiplay it by the average number of citation of the papers contributing to h-index, and than take square root. This gives area based(A-index) of the h-index citation field. This removes problem that h-index of N can not distiquish is it N-times N-citations (the weakest posible case)or N-times of citations that are much larger of N. For instance h-index of 5 one can get by citations: 5 5 5 5 5, but also by citations: 50 5 5 5 5. A-indeks in this two cases is 5 and 8.4 respectively. Such an idex will put science genius (milestone discovery) back to the top.

Nenad Juranic, Ph.D