Deciding between UChicago, Columbia, Duke, and UW for stats PhD

[conflitz]

I'm very fortunate to have been admitted to the statistics PhD programs at UChicago, Columbia, Duke, and the University of Washington. Now I'm having a hard time choosing between them! I'm mostly interested in statistical machine learning, causal inference, and Bayesian statistics (particularly as it relates to the first two topics), but my interests are definitely not that specific yet. I'm also not married to Bayesian statistics -- I find it interesting, but I think I would also be fine if I didn't end up doing research in that area. I enjoy both theory and applications in the sciences and social sciences. I'm not entirely sure where I stand on academia/industry at the moment.

UChicago pros: tons of advisor options between statistics/CAM/TTIC/Booth, big city, lovely campus, highly prestigious, high stipend relative to cost of living.
UChicago cons: far from family, very cold winters, less safe than the other options (at least according to the grad students I've spoken with), the 9 week quarter system seems intense and stressful.

Columbia pros: good number of faculty with compatible interests, big city, nice campus, highly prestigious, close to family, great location for tech and finance internships.
Columbia cons: low stipend relative to cost of living, cold winters, maybe a little less prestigious than Chicago stats.

Duke pros: good research fit, warm weather, very high stipend relative to cost of living, great campus, no written qualifying exam.
Duke cons: not a ton of advisor options, far from family, not in a big city, probably less prestigious than the above 2.

UW pros: decent research fit, lovely city, great campus, great location for tech internships, the students seem very happy.
UW cons: very far from family, cold and wet winters, low stipend relative to cost of living (but better than Columbia's), probably less prestigious than Chicago and Columbia.

I'm also on the waitlist for UPenn Wharton, CMU, and Yale. Would love to hear any thoughts from people more knowledgeable than I am!

[kidbayes]

I'm currently a graduate student at UChicago. I don't think safety is enough of an issue to prevent one from going here. Feel free to pm with any questions

[M1ller]

Duke will brainwash you to be Bayesian. If you are really into Bayesian, Dunson and Tokdar will make you a rising star.

If you want to do machine learning, definitely CMU; do the joint ML PhD program. Columbia is the next, if you can work with Blei. UW is also very good.  I would not recommend UChicago for ML.

Columbia is notorious for its huge master's program. UChicago also runs a big master's program, too. Master's students in these programs are ambitious, and many of them are advised by faculty members and do research. This means that PhD students have to share advisor's time and resources with them.

If you want to graduate early, I have heard that Duke is very flexible; many students graduate in 3-4 years. 

 

Edited March 28, 2023 by M1ller

[Econometrician]

10 hours ago, M1ller said:

Duke will brainwash you to be Bayesian. If you are really into Bayesian, Dunson and Tokdar will make you a rising star.

If you want to do machine learning, definitely CMU; do the joint ML PhD program. Columbia is the next, if you can work with Blei. UW is also very good.  I would not recommend UChicago for ML.

Columbia is notorious for its huge master's program. UChicago also runs a big master's program, too. Master's students in these programs are ambitious, and many of them are advised by faculty members and do research. This means that PhD students have to share advisor's time and resources with them.

If you want to graduate early, I have heard that Duke is very flexible; many students graduate in 3-4 years. 

 

Faculty at Columbia do not advise or work with Master's students at all. I don't know about Chicago.