2020 Fall Statistics PhD Application Evaluation

[VirtualStatistician]

Hi all, 

I'm currently a senior planning on applying to PhD programs in statistics this winter. I'm especially interested in machine learning theory.

Undergraduate Institution: Cornell University

Majors: Mathematics, Computer Science (Double major)

GPA: 3.83/4.0 (cumulative) 

Type of Student: Domestic Asian Male

Relavant Courses:

Math: 

• Linear Algebra (lower level) (A-)
• Multivariable Calculus (B+)
• Applicable Algebra (A-)
• Numerical Analysis (A)
• Linear Algebra (upper level) (B+)
• Proofs (A+)
• Honors Analysis (Currently taking)

Computer Science:

• Discrete Math (A-)
• Objected Oriented Programming and Data Structures (A)
• Functional Programming (A-)
• Systems Programming (A-)
• Operating Systems (A-)
• Algorithms (A)
• Machine Learning (A)
• Artificial Intelligence (A)
• Natural Language Processing (Currently taking)

Other:

• Statistics and Probability (Economics) (A)
• Data and Systems Analysis (Operations Research) (A-)
• Learning with Big, Messy Data (Operations Research) (Currently taking)

GRE General: Have yet to take

GRE Math Subject: Have yet to take, not sure if it's worth it

Research Experience:

• Started to do research this semester in machine learning fairness. There's not much time, but I'll put a lot of time into it and try my best to accomplish something.
• Not sure if this counts, but the proofs class had a term paper assignment where we came up with an interesting math problem to solve, and wrote a pretty detailed paper exploring it throughout the semester (~8 pages)

Work Experience:

• Two internships at large companies. One for software development, one for actuarial science. 

Letters of Recommendation:

• Have not asked yet, but planning to get one from the professor (from the operations research department) supervising the research I started this semester.
• One from the professor from the proofs class.
• One from the professor from a sociology class I took. The class was focused on discussing data science papers that tackled issues in social science. Because it was discussion based, and it was a small class, the professor knows a decent amount about me.
• Alternatively, I may be able to ask for a letter from the numerical analysis professor, as I went to his office hours a lot. However, I took his course 2 years ago.

Schools I'm Considering, Ranking from US News

• University of Massachusetts–Amherst (#74)
• University of Virginia (#69)
• University of California–Santa Barbara (#67)
• Northwestern University (#55)
• University of Connecticut (#44)
• Colorado State University (#44)
• Ohio State University (#37)
• University of Illinois–Urbana-Champaign (#37, Math GRE recommended ?) 
• University of California--Davis (#31)
• Cornell University (#20)
• University of Wisconsin–Madison (#16)
• University of Michigan–Ann Arbor (#12)

As you can see, I don't have that many math classes, and I didn't do exceptional in them. I consider my work in computer science to be stronger, but in order to do research in machine learning, I figured it would still be easier to get into a PhD program for statistics than for computer science, because there is less of an expectation to have undergraduate research. Would switching Natural Language Processing for grad measure theory help? Or perhaps would taking the Math GRE help more? Both of these would require a significant time commitment, which might impact the quality of my research (along with the recommendation letter).

Please let me know if my list is too ambitious, or if I could do better. I'm worried about not getting in anywhere though, as even the #69 University of Virginia typically accepts 3/80 applicants (according to their website). Also, some of these schools accept applications from January onward, which could be helpful to have more time to do research and improve my application. Should I be only looking at those? Thanks for your input!

 

 

[bayessays]

You're aiming too low, with a great GPA and good math/CS background from an Ivy. I'm going to assume you'll get at least a 165 on your GREQ and do decently in your analysis class.  I'd get rid of a few of those lower schools. I think you should add a few more schools in the 10-30 range, think big state schools like NCSU, Minnesota, PSU, TAMU, Wisconsin, Iowa State and a few reaches like Duke or CMU if you want to go for it.  I'd be surprised if you didn't get into schools like OSU and don't think you need to apply to any schools below Colorado State unless for personal reasons.

[Geococcyx]

Yeah, you'll get in somewhere -- as I recall, Virginia's just a small program, so they're harder to get into (just like NYU, Northwestern, Brown biostat, etc.) because of their small size.  

I got into Illinois last year without taking the Math GRE, with roughly the same math background from a worse university.  There's no reason to take the Math GRE unless you're applying to Stanford.  Maybe it helps at Columbia, UChicago, and UC Berkeley, but I know the first two accepted people last year who didn't take it, so it's not a big deal.  

Bayessays posted while I was writing, and they are correct about your chances.  I think you should apply a little above Duke, and maybe a little above CMU, but I do usually suggest people apply higher than bayessays does -- that's your choice, and ultimately is a factor of how much money you have lying around to throw at applications.  In any case, you're definitely aiming too low right now.

Edited August 31, 2019 by Geococcyx
Redundancy

[VirtualStatistician]

Hi guys, thank you so much for the responses! I can rest a little easier at night then!

From what I gather, selectivity is dependent on both the school's ranking, and how big the program is, then. @bayessays, thanks for the suggestions! I'll probably apply to at least Minnesota and Penn State out of those. 

@Geococcyx, thanks for also including your results! I get the feeling that grad apps are very volatile, so it's reassuring that you were still accepted into 7 programs! My dream school is University of Washington, so I might try the math GRE just for a shot, but I won't stress too much about it. 

 

[bayessays]

I don't think Washington requires the math GRE, just highly recommends it like Chicago if I remember correctly. You should definitely apply there if you'd like - I don't think any application besides Stanford would be throwing money down the drain.

[VirtualStatistician]

Hi guys, now that the deadlines are approaching, I'm attempting to finalize a list. Here's an update:

• For the GRE, I got 163V/167Q/4.5W.
• I ended up switching from Honors Analysis to Regular Analysis. I'm doing quite well in Regular Analysis though (I think I'll probably get at least an A; I actually got a perfect score on the midterm)
• I ended up making good progress in my research, so my recommendations should be a bit stronger than I was originally expecting. 

Reach:

• University of Washington (#8)
• University of Michigan–Ann Arbor (#12)
• Duke University (#12)

Target:

• University of Wisconsin–Madison (#16)
• University of North Carolina–Chapel Hill (#19)
• Pennsylvania State University (#20)
• University of California–Davis (#31)
• University of California–Los Angeles (#31)
• Johns Hopkins University (#31)
• University of Illinois–Urbana Champaign (#37)
• Ohio State University (#37)
• University of California–Irvine (#50)

Safety: 

• University of Connecticut (#44)
• Colorado State University (#44)

What do you think of my list? A few more questions I have:

• I sort of assumed that UC Davis and UCLA would be easier to get into than schools like Texas A&M and U Minnesota because they're lower ranked, but I could be wrong. If they're actually harder to get into, I'll probably swap them out. 
• How many schools should I apply to at the most? I'm ok with the application fees, but I don't want to inconvenience my professors too much. 
• Next semester, I'm planning on taking a lot of math classes so my application will be stronger next year if I don't get into any this year. Would schools prefer seeing an undergraduate math class (E.g. Stochastic Processes), or a graduate operations research class (Probability Theory)? (Or, perhaps, a CS Machine Learning Theory class)

[cyberwulf]

I think your list is still a little bottom-heavy. You'll likely get into your bottom 7-8 places; do you really need that many safety options? I would cut a few schools in the 30-50 range and add a handful more top 15's.

[VirtualStatistician]

Hi, thanks for the advice. I'll think I'll swap out JHU for CMU. It actually really pains me to do this, just because I'm (irrationally) paranoid of not getting in anywhere, but I think I'll go with it.