Applying to Stats PhD 2021: Am I delusional?

[Lilly187]

Hi everyone!

I'd really like some advice as to whether I stand a chance in applying to some top 20 programs. I will provide as much info as possible. 

I am Canadian and currently completing my undergrad(will be applying to US unis). I am doing a BSc. with a specialization in math and stats (I have more science courses than a math major would since they were required by my program, not sure if this puts me at a disadvantage compared to someone with more math courses). I am extremely interested in Machine Learning, and more particularly, algorithmic fairness and AI safety. As a result, I found some profs whose work just blows me away, but sadly, they are at fantastic universities . I'd like to know how much hope I can realistically hold without being completely delusional. 

Student Type: International (Canada) Female Caucasian

Undergrad: Canadian University

Major: General science w/ specialization in math and stats

GPA: 3.9

Relevant Courses (what I have and will have taken): 

Linear Algebra I and II, differentials, cacl I and II, probability, mathematical statistics, intro to data science, regression, general linear models, real analysis, abstract algebra, time series, complex analysis, computational methods for inference.

SIDE NOTE: in my first semester of fourth year I will be taking Intro to Programming (first year compsci). I have previously taken courses working with R, but never python. I used python extensively to do a deep learning project in my third year, but thought a course would be good to learn a foundational and systematic approach to python. Will schools think it's weird I'm taking this course so late in undergrad?

GRE: Am taking it in September. Will update when I do! Apologies, I hope it is still possible to assess my application barring this info ? 

Research/work:

• Summer after first year: research in a health geomatics lab, worked with Netlogo to make an agent-based model (looking back, it is not an impressive model by any means lol but learned a lot) 
• Summer after second year: was meant to do statistical analysis for a geography lab, but could not get access to a dataset. Ended up doing a lit review (which will hopefully be published this year. Waiting on journal decision). This would be a first-author publication, but in the topic of aging (nothing to do with statistics). Not really sure how this will be perceived. If you're wondering where the math is, I promise it's coming and I really do love statistics, it just took me a minute to find my way lol. 
• Third year: Semester-long application project with supervisor in stats department, used CNNs for image classification
• Summer after third year: Summer research in an ML lab [received NSERC that funded this] (didn't really do a single concrete project with all the craziness going on w/ Covid, it was more numerous little projects)
• 4th year thesis: topic undetermined, likely reinforcement learning or something with a generative text network? Supervisor is in the stats department, and I imagine it will be more of an applied project rather than theoretical.
• TA experience for calc

Letters of Recommendation: 2 letters of recommendation from profs I've worked with in math department, 1 from geography department (summer after second year) is the plan (I figured they could speak to my research ability/work ethic, and also I don't have another person to ask in the stats department)

Where do I want to apply?

(Note: for Canadian schools, plan is to apply to masters and then switch into a PhD (should I meet the criteria)/do a PhD after, for the US I am directly applying to PhD programs)

For all I am applying to masters/PhD in statistics (is the plan as of now)

- U of T (masters)

- UBC (masters)

- McGill (masters)

- Harvard (I feel silly writing it, but there's a prof who is researching algorithmic fairness that I would be so grateful to work with/am super interested in. Any advice is extremely appreciated.)

- UC Berkeley (again, there's a prof researching AI safety that aligns ridiculously well with my interests)

- Stanford (masters) [again, interested in the research, but as far as I understand Stanford only offers a course-based masters so I wouldn't have a supervisor? I am thinking of applying to masters since they require the Math GRE and I don't currently have plans to write it, not sure if this is a mistake but I get overwhelmed trying to figure out how I would fit it in)

If you've read this far, I can't thank you enough for your time. Any advice on how to strengthen my application/if I have a chance is greatly appreciated. If there are any courses you think I need to take to strengthen my statistics/math background, please let me know!! Any discrepancies/weaknesses I might want to address in a statement of purpose? Thank you kindly!!

[StatsG0d]

I don't see Calc 3 on your profile, but since you have other advanced math classes I'll assume that you've taken that (if not, you must).

You have a really strong profile, but being an international student will make it harder than if you were domestic.

I would not bother applying to master's programs. You have a strong enough profile to get into PhD programs directly. There's no guarantee that Stanford's master's program will give you any advantage applying to their PhD program, as for statistics admissions usually master's and PhD's are somewhat separate (there are a few exceptions). My understanding is that you would still need to apply to the PhD program even though you'd be a master's student in the department. It is possible that they have an internal review committee for internal applicants, but I highly suspect you would still need to take the math GRE regardless.

You could maybe consider looking into biostatistics departments as well as there are a lot of faculty working on precision medicine, which you might find interesting. Check out Erica Moodie's work at McGill (there are many more in the US, but she's a leader in the field and you probably have some advantage applying domestically).

[Stat Assistant Professor]

Although OP is an international student, she does attend one of the top four universities in Canada and has a high GPA. And I believe applicants from the top schools in Canada are treated similarly as domestic applicants in the admissions process at a lot of U.S. Statistics programs (Canada's a bit different than elsewhere).

OP: I think you have a good shot at a lot of very good statistics PhD programs (without the Masters), and so I concur with StatsG0d to apply directly to PhD programs. The top PhD programs are difficult for anyone to get into, so I would make sure to have a few "safer" options. But I wouldn't say that schools of the tier of University of Washington or Carnegie Mellon are out of the question for you.

[cyberwulf]

I think the top 5 stat programs are reaches for you, but it's definitely worth applying to a couple of them. I suspect you'll find more success in the 10-25 range. Your math background is solid, though McMaster is probably perceived by most as a little less prestigious than UBC/UofT/McGill/Waterloo. Great research experience, but unfortunately that can be a little hard for admissions committees to evaluate. The primary value of those experiences is that it hopefully allowed you to build strong connections with faculty who will write you glowing letters.

If you're set on going to a top-shelf PhD program, one approach might be to do a Masters at a top Canadian university, then re-apply. 

[Casorati]

I am pretty sure you can get into master's program at Toronto/UBC/McGill. Top programs like Harvard/Stanford/Berkeley are competitive for everyone. You didn't post your grades in real analysis/complex analysis/mathematical statistics ...but if you have strong grades in them, you should be at least in discussion at some of them. You have enough math background but I think courses in measure theory will further strengthen your application, especially if you get an A in it. 

[insert_name_here]

4 hours ago, cyberwulf said:

I think the top 5 stat programs are reaches for you, but it's definitely worth applying to a couple of them. I suspect you'll find more success in the 10-25 range. Your math background is solid, though McMaster is probably perceived by most as a little less prestigious than UBC/UofT/McGill/Waterloo. Great research experience, but unfortunately that can be a little hard for admissions committees to evaluate. The primary value of those experiences is that it hopefully allowed you to build strong connections with faculty who will write you glowing letters.

If you're set on going to a top-shelf PhD program, one approach might be to do a Masters at a top Canadian university, then re-apply. 

This is good advice.

I'd encourage you to apply to 1 or 2 "dream" schools, even if your odds aren't great. Mix that with some Canadian masters (where you'll have no trouble), and 5+ lower ranked schools, and you're set.

My personal approach would be to stay in Canada unless you get into a top (Berkeley/Harvard) program. Admission rates are ~3x lower for international students at top schools in the US, so you're likely to get into a worse school there than in Canada. Places like U of T are great, probably better than a 10-25 ranked PhD program in the US. Plus, America isn't the best place to live at the moment, especially compared to Canada (coming from a Canadian who moved to the US)

[StatsG0d]

15 hours ago, Stat Assistant Professor said:

Although OP is an international student, she does attend one of the top four universities in Canada and has a high GPA. And I believe applicants from the top schools in Canada are treated similarly as domestic applicants in the admissions process at a lot of U.S. Statistics programs (Canada's a bit different than elsewhere).

OP: I think you have a good shot at a lot of very good statistics PhD programs (without the Masters), and so I concur with StatsG0d to apply directly to PhD programs. The top PhD programs are difficult for anyone to get into, so I would make sure to have a few "safer" options. But I wouldn't say that schools of the tier of University of Washington or Carnegie Mellon are out of the question for you.

Yeah, I'll admit that I don't know if Canadians are viewed differently among adcoms as other international students, but I would presume not as I would think that students from Peking, Tsinghua, ISI, etc. are all sure bets (perhaps even more than at Canadian universities), and even some of the top students from these schools have trouble getting into top-10 PhD programs.

[Stat Assistant Professor]

57 minutes ago, StatsG0d said:

Yeah, I'll admit that I don't know if Canadians are viewed differently among adcoms as other international students, but I would presume not as I would think that students from Peking, Tsinghua, ISI, etc. are all sure bets (perhaps even more than at Canadian universities), and even some of the top students from these schools have trouble getting into top-10 PhD programs.

I think it is definitely less competitive for Canadian students than for students from Asia. I think there are overall fewer students from Canada applying to Stat PhD students in the U.S. than there are from China and India (a lot of top Canadian students stay in Canada to get their PhD from UBC, UofT, Waterloo). There are also much fewer visa issues as well (for instance, Canadian citizens don't need F-1 visas to study in the U.S.).

I know former students from McMaster, University of Mannitoba, and Acadia University who were admitted to PhD programs in Statistics at University of Washington, Yale, and Carnegie Mellon. So I imagine that someone with an excellent GPA From McMaster stands a good chance at very good Stat PhD programs in the U.S. (though the top 5 might be hard to break, as @cyberwulf mentioned). 

Edited August 18, 2020 by Stat Assistant Professor

[Lilly187]

10 hours ago, insert_name_here said:

This is good advice.

I'd encourage you to apply to 1 or 2 "dream" schools, even if your odds aren't great. Mix that with some Canadian masters (where you'll have no trouble), and 5+ lower ranked schools, and you're set.

My personal approach would be to stay in Canada unless you get into a top (Berkeley/Harvard) program. Admission rates are ~3x lower for international students at top schools in the US, so you're likely to get into a worse school there than in Canada. Places like U of T are great, probably better than a 10-25 ranked PhD program in the US. Plus, America isn't the best place to live at the moment, especially compared to Canada (coming from a Canadian who moved to the US)

Perfect, thank you so much! This sounds like a great game plan and is probably exactly what I'm going to do  It's a bit hard for me not to get my heart deeply set on things, so I'm going to have to really keep my mindset realistic and grounded lol. Thanks again for the advice.

[Lilly187]

1 hour ago, Stat Assistant Professor said:

I think it is definitely less competitive for Canadian students than for students from Asia. I think there are overall fewer students from Canada applying to Stat PhD students in the U.S. than there are from China and India (a lot of top Canadian students stay in Canada to get their PhD from UBC, UofT, Waterloo). There are also much fewer visa issues as well (for instance, Canadian citizens don't need F-1 visas to study in the U.S.).

I know former students from McMaster, University of Mannitoba, and Acadia University who were admitted to PhD programs in Statistics at University of Washington, Yale, and Carnegie Mellon. So I imagine that someone with an excellent GPA From McMaster stands a good chance at very good Stat PhD programs in the U.S. (though the top 5 might be hard to break, as @cyberwulf mentioned). 

Thank you for all your helpful advice on this thread! It's really nice to know a reasonable approach/expectation when it comes to the top 20 list and applying as a Canadian  Especially when application fees can be a big pill to swallow. I will give the Berkeley/Harvard dream a go with fingers crossed, and then of course apply to good Canadian and US schools (thanks for suggesting University of Washington/Carnegie Mellon, I hadn't looked into them much).

[insert_name_here]

3 hours ago, StatsG0d said:

Yeah, I'll admit that I don't know if Canadians are viewed differently among adcoms as other international students, but I would presume not as I would think that students from Peking, Tsinghua, ISI, etc. are all sure bets (perhaps even more than at Canadian universities), and even some of the top students from these schools have trouble getting into top-10 PhD programs.

Just to add some anecdotes (independently of OP), in my experience top students from Waterloo/Toronto/McGill don't have much trouble getting into top-5 PhD programs. UBC may be in the same tier, idk as many people from there.

[statsnow]

Insert_name  I think your advice is really good but nothing is guaranteed for the top 5 programs.  At that level I think alot is about fit and what they want their group of grad students to look like

[SPIWizard]

I'm taking the path as suggested by cyberwulf, and now doing a masters in Canada at one of the top 4. My two cents on why below, you may find it helpful:

I also had a single NSERC RA, and similar grades from a decent but not top 4 in Canada. I got into all of the Masters programs you listed early in the admissions cycle, and was competitive for the CGS scholarship at each so I think you will be fine getting into those. I didn't send any applications to PhD's since I thought even if I had a chance, I could only increase that chance by doing the Master's first. I don't see the two years as wasted, it gives me more of an idea of the graduate research experience in math/stats (I was a late bloomer, first research right before 4th year), and also a way out to industry if I'm not interested, since most statistician jobs require a Masters here in Canada.

[Lilly187]

11 hours ago, SPIWizard said:

I'm taking the path as suggested by cyberwulf, and now doing a masters in Canada at one of the top 4. My two cents on why below, you may find it helpful:

I also had a single NSERC RA, and similar grades from a decent but not top 4 in Canada. I got into all of the Masters programs you listed early in the admissions cycle, and was competitive for the CGS scholarship at each so I think you will be fine getting into those. I didn't send any applications to PhD's since I thought even if I had a chance, I could only increase that chance by doing the Master's first. I don't see the two years as wasted, it gives me more of an idea of the graduate research experience in math/stats (I was a late bloomer, first research right before 4th year), and also a way out to industry if I'm not interested, since most statistician jobs require a Masters here in Canada.

Thank you for the insight @SPIWizard! That's really good to hear I have a few questions if you don't mind:

1. Did you contact profs at Canadian schools before applying? I used to be under the impression that this was expected, but for stats it seems like it's not?

2. If you don't mind me asking, what school did you attend, and did you enjoy it! My main concerns for grad school are: (1) being in a department which provides a sense of community/support and (2) finding a prof who has some interest in AI safety or at least ML. Do you have any advice on which Canadian schools might meet this criteria? (These qs are a bit long so please don't feel pressured to answer)

Also side note: (this ties in with the above and others might be interested to know) Someone mentioned to me that there's almost no difference between masters -> phd and undergrad -> phd time-wise, since the research experience you gain in masters will make you faster as a phd. So if anyone else is wondering about the difference (time-wise), the advice I got was to not really factor that in, it'll be negligible compared to other differences you should consider to figure out which route is best for you.

Edited August 19, 2020 by Lilly187