Want to apply for a PhD in stats but not sure where I should apply. Advice?

[clurp]

Hi, 

 

I know it's way early to ask about "chances" for fall 2014, but I have to make a big decision soon and having a better idea about this would make that easier. 

 

I'm a US citizen. My undergrad is in math at a top 20 university (according to usnews). I had a 3.78 gpa and took a lot of math courses:

 

• a couple semesters of graduate analysis (I took the undergrad version first) 
• two semesters of algebra, one of which only included grad students
• representation theory
• combinatorics/graph theory
• differential geometry
• a few research seminar style courses
• honors linear algebra 
• honors calc courses 
• an intro to applied and computational math (basically a matlab course) 
• a course on bioinformatics which basically introduced some machine learning methods with python 
• basic stats (which I took my last semester, sort of blew off and got a B+ in ) 

 

and probably some others I'm forgetting. I was pretty ambitious in my course selection, but didn't always get A's. There was a mix of A's, A-'s and some B+'s, maybe a few B's. 

 

During undergrad, I did two summer research experience things: 

 

1) one, in a math department, resulted in a publication with many co-authors in a journal which publishes undergraduate research 

2) the other was in computational neuroscience (solving systems of differential equations with matlab) 

 

I also partook in the budapest semesters program and took intro to algebra (which I've counted above), functional analysis, and topology. 

 

After graduating, I applied and got a fellowship to a masters program abroad. After the first year of that program I had a 3.72 GPA. There, I took additional courses in 

• analysis 
• an algebra sequence 
• functional analysis / differential equations 
• stats/prob 
• a course on algorithms / theory of computing 
• 2 courses on statistical learning / non parametric statistics 

and maybe some I'm forgetting. 

 

I really enjoyed the courses on statistical learning / nonparametric statistics and did well in them (kind of blew off the intro prob/stats course again though, and got a B+ :/ ), so the summer after that, I returned to the US to do an internship as a "data scientist" and continued on beyond the summer, taking a leave of absence from the master's program. 

 

At this job, I work to solve statistics/machine learning problems, mostly with R. Doing this has forced me to face a lot of problems which are central to statistics and machine learning and has probably contributed more to my understanding and passion for the subject than my previous academic work in those subjects. It's also made me a lot more knowledgable about computers.

 

Both in order to do a better job and out of curiosity, I've been doing as much reading and MOOC participating as I can in statistics/machine learning topics. In the process, I've come to realize that I would love to do a phd in statistics, particularly something related to statistical machine learning, bayesian methods, and/or nonparametric statistics. 

 

If I went back to the master's program, I could work with the well-known professor of my statistical learning / nonparametric statistics courses on a master's thesis and possibly get a good rec from him (but by the time it would have to be written I would just be starting in earnest on the thesis and i've heard international professors are much more reserved in their praise than american professors), but i'd rather not because I like my job quite a bit and living abroad is more difficult for me now than it was when I made the decision at the end of undergrad. 

 

If I don't return to the master's program and instead continue working, I'll probably have a good recommendation from an undergraduate professor with whom I did one of the research projects (though it was quite long ago) and recommendation(s?) from phds I work with. I'm not sure how recommendations from outside academia will be viewed.

 

Does it seem like returning and completing the program would have a big effect on my chances? Would I be likely to get into some program with researchers in my areas of interest regardless? What schools would be good targets for me?

 

[SenhorJose]

Some good statistics departments for machine learning seem to be Stanford, Berkeley, Washington, Michigan, Duke, Chicago, and CMU. Stanford and CMU are reputed to be more applied, Chicago more theoretical, Duke obviously Bayesian. U Washington has a new "Big Data" type initiative (http://escience.washington.edu/) and they seem to be hiring many good faculty in ML. Berkeley has some really good people who seem to work more on the CS side of things. Not sure about Michigan...

 

But take all this with a grain of salt -- these are only my impressions, so the list of schools is definitely incomplete, and the way I characterize them may be imperfect/incorrect.

Edited May 2, 2013 by SenhorJose

[ssk08]

As a backup you could apply to Adaptive Signal Processing research groups in the Electrical Engineering departments. The stuff you do there are pretty similar to Machine Learning/Applied Stats

[Funkoverload]

Cornell, Columbia, Duke, USC would suit your background.

[clurp]

I think I will be applying to Columbia and Duke. Cornell's program looks good, but I don't think I'd like to live in Ithaca. I've looked at USC a bit, but had a hard time figuring out which department would be best to apply to. It seems as though they have the relevant faculty scattered through a number of departments. 

[Funkoverload]

How do you think of UC Irvine? It would be good for machine learning. USC (marshall, stat) is for high dimensional stat. Contact prof. Jinchi Lv if you like it.