wine in coffee cups

I think learning how to ace the quantitative GRE section will help you in the long run for statistics programs. One thing I was surprised by in my program is how exam-focused the core coursework is. Lots of memorization, doing integrals very quickly, taking many exams that count for a large % of your grade in a short period. All of this is to prepare you to take do-or-die master's and PhD qualifying exams, which are a required component of many statistics and biostatistics programs. For an anxious test taker, preparing to pass these exams is going to involve doing a lot of timed practice questions, identifying the types of things you are stumbling on, learning how to triage tests to get the parts you can nail out of the way quickly and save the other stuff for last -- all skills you will get better at by trying to do the same for the GRE. You can do this.

It's not as if the only piece of information admissions committees will have available to them is a list of your publications. You can (and probably should) talk specifically about your research experiences in your statement of purpose. I would also hope that any letter writers with whom you've worked would convey details about your contributions to their projects. People making decisions don't need to infer the extent of your involvement solely based on authorship order.

My undergrad gave me an ASA membership for free, I think there is a deal where departments can sponsor some number of students. Ask around.

You'll probably do okay unless there are other well-qualified people applying from your university to the same departments at the same time whose recommendations are all more excited about them than you. That's an really good GRE math subject score, really good GPA, really good transcript, good Putnam placement, so your pure math credentials are truly much stronger than the majority of American statistics applicants (and competitive with serious applicants to good pure math programs). Maybe you'll end up on a lot of waitlists, but that's not so bad. I would not submit a letter from a humanities professor unless you were really desperate or already had three letters from more relevant faculty. I don't think statistics departments will know what to make of it because it won't provide the comparative information they're looking for about you vs. other statistics/math students. I would aim for a letter from someone you took a statistics class with, a letter from someone you took an analysis class with, and then a letter from some other math or statistics professor. Faculty are surprisingly used to writing letters for students they weren't particularly close with. Just don't be an asshole, give them a short explanation of how your interests have moved towards statistics, ask nicely for advice from the faculty who have sent other students to stats PhD programs, consider sending drafts of your statements of purpose to help them understand you and your motivations better if they agree to write a letter.

Not defined. Few to none are coming in with anything even resembling a research proposal, it's just not how things work in statistics. (Unlike other sciences, so pipe down, non-statistics people!) I think it can be helpful if you have interests in certain areas or applications to discuss them in your statement of purpose, since faculty working in that area may have research assistantship funding and you might be given a slight preference in admissions if you're a strong applicant who could fill that slot. This was the case for me: I ended up on an RAship with my advisor coming right in because I wanted to do work related to a topic that he had a grant for, and I was entering with relevant background and an excitement for the area that I don't think many other applicants had. It's not an especially common situation in our discipline, though, because many new students are fresh out of undergrad and haven't had exposure to either the methods or applications they might work on. It's probably more important to come across as mathematically prepared, intellectually curious, and an overall interesting person who someone would want to advise. The particulars of your research interests don't matter too much beyond establishing that you could fit into the department. Students also change their minds substantially, so it's hard to take lists of interests too seriously unless you actually have some experience to suggest a commitment. Off the top of my head, I can think of statistics students who shifted from genomics to demography to network models, from topic modeling to genetics, from stochastic processes to nonparametric theory, from spatial methods to algebraic statistics, from dynamic modeling to empirical processes, and from social networks to ecology.

My guess is that your weakness isn't so much uncertainty about the rigor of your undergrad program -- hopefully your references will be able to put your grades and accomplishments in context -- but just that the process is really competitive, especially for international students, who face single digit acceptance rates at top programs. (cyberwulf, a biostat professor of Canuck origin, has commented a bit on Canadian students applying to US programs and perhaps elsewhere.) The four programs you've identified are all particularly selective. Columbia and Cornell are both pretty good statistics departments that attract a disproportionate number of applicants because they are in fancy Ivy League universities. Columbia additionally has its NYC location (almost all the PhD students I talked to there cited that as their main reason for applying/going) and Andrew Gelman as maybe the most public-facing academic statistician in the US to draw in even more applicants. Berkeley is like the #2 statistics department in the world, and even if that weren't true, it would still draw a lot of applicants because of its location and overall university reputation. The best applicants from the US and around the world are all going to be sending applications to Berkeley, Columbia, and to a lesser extent, Cornell, but these schools can only admit so many people. As for NYU, the statistics PhD at Stern appears to have literally just TWO students. Who can begin to guess what separates those individuals from everyone else who applied? My point is, your profile sounds pretty good, but even if you had gone to the best university in Canada, you're still going to be one of many well-qualified international students who want to study statistics and finance applying to these popular programs. It sure would be nice if one of those four worked out for you and I think that's absolutely a possibility, but it's very good that you have Toronto as a backup.

I don't know anything about Indiana either, but it looks like a small department with a few focused research areas. A couple of new assistant faculty appear to be interested in social networks so that might be an area they are hoping to build/maintain strength in. It's possible that the graduate program is relatively new as there are only a handful of PhD students. Shouldn't be a top choice over somewhere like CMU or UW, but if you were serious about network methods, this could be a worthwhile backup/safety program. It can't hurt to look into it more.

Not a safety school, but Berkeley stat/CS is a must-apply for you.

Not to be a Deborah Downer, but I wouldn't characterize that as being "a lot of mathematics background" for biostatistics PhD admissions. A full calculus sequence and linear algebra are barely minimum requirements at many programs. Quite a few places will want to see a proof-based analysis class too. I agree with the suggestions to consider applying to master's programs first.

CMU stat is rather heavy on social networks: Fienberg, Rinaldo, Shalizi, Thomas, Choi (in the Heinz school). U Washington stats too: Morris, Hoff, McCormick. Some other social network researchers I can think of with stat department affiliations: Handcock (UCLA), Hunter (Penn State), Wasserman (Indiana), Airoldi (Harvard), Neville (Purdue). Hopefully that gives you some ideas about places to look.

You apparently go to UMN. UMN's biostat PhD program is one of the better ones, so let's take a look at exactly which math courses their program requires its students to take since they know the material and level of rigor. Their curriculum page says that incoming students without a master's degree in statistics or biostatistics need to take Math 5615H or Math 4603 their first semester. Math 5615H: Honors: Introduction to Analysis I 4.0 cr; Prereq-[[2243 or 2373], [2263 or 2374], [2283 or 3283]] or 2574; fall, every year. Axiomatic treatment of real/complex number systems. Introduction to metric spaces: convergence, connectedness, compactness. Convergence of sequences/series of real/complex numbers, Cauchy criterion, root/ratio tests. Continuity in metric spaces. Rigorous treatment of differentiation of single-variable functions, Taylor's Theorem. Math 4603: Advanced Calculus I 4.0 cr; Prereq-[2243 or 2373], [2263 or 2374] or 2574 or # ; fall, spring, summer every year. Axioms for the real numbers. Techniques of proof for limits, continuity, uniform convergence. Rigorous treatment of differential/integral calculus for single-variable functions From this it is clear that the math UMN biostat needs its students to acquire during the PhD program is advanced calculus, not real analysis. In terms of applying, coming in with advanced calculus already completed ought to be good enough.

I would take the advanced calculus sequence rather than the real analysis sequence. You want to get as good a grade as possible in whichever sequence you take, especially if you're making up for poor grades in lower-level classes. If you've never done proof-based math before and have only self-prepared, you will probably have a tough time keeping up with the rest of the students in a graduate-level analysis class because a lot of mathematical sophistication will already be assumed. They'll be seeing harder versions of things they already know, too, while you'll be learning all the material (as well as just how to write proofs) for the first time. Every biostat PhD program that requires analysis prior to enrolling will be happy with material at the level of advanced calculus. Graduate-level analysis is overkill. You just want to take enough analysis to be able to handle future math requirements in whatever program you end up in (e.g. a measure theory class).

Good on you for doing such a nice job consolidating this advice and sharing your experiences. I would have found this useful while I was applying. I particularly liked the bits about the "debate" on statements of purpose (though not really a debate so much as cyberwulf describing how his department uses them, nobody else having much of a clue beyond a reasonable suspicion of heterogeneity) and the advice to use applying for fellowships as an opportunity to think about your motivations and research interests. All of your advice seems pretty reasonable to me except that specific 70th percentile math subject GRE recommendation. I think that's an apocryphal tidbit that someone here once posted, apparently confidently enough to convince others to buy into it and parrot it even if they didn't take it themselves! Is there a reputable source underlying this (e.g. some professor at a top department flat out saying so)? If not, I don't think anybody here knows what makes a good enough score on that test for the statistics programs that encourage or require sending it. Stanford publicly posts an 82nd percentile average on their admissions FAQ, but could conceivably have a very wide distribution of scores of admitted applicants. And I would not be surprised if whatever bars that exist for the subject GRE at Stanford or elsewhere are lower for domestic than international applicants, might be lower for women or underrepresented minorities, might be ignored or waived altogether for a sufficiently interesting applicant from a non-math major who didn't have the breadth of pure math background to contemplate taking it and doing okay (well, whatever "okay" means).

You don't need 40 hours per week (perhaps not even 40 hours total) and you don't need to spend any money thanks to all the internet resources out there. I had not taken calculus in about 10 years when I started my program and did not review anything before starting, which made for a rough adjustment back to school. Here's a pretty exhaustive list of the high school/early college math I used in master's-level theory and methods coursework: Common differentiation rules (power, product, quotient, chain) Partial derivatives Univariate integration techniques: primarily integration by parts, u-substitution (I don't think I've ever needed trig substitution, actually just stay away from trig functions altogether). 90% of the integrals you evaluate will look like the definite integrals on my fave wikipedia page, if you can figure out how to do some of these then you're probably set. Infinite series representations, Taylor approximations L'Hopital's rule, general comfort evaluating limits, understanding of continuity and piecewise-defined functions From multivariate calculus: Jacobians and change of variables, inverse function theorem, iterated integrals and multiple integrals. You do not need to review line integrals, Green's theorem, all that physics-y vector field stuff. Linear algebra: vectors and matrices in general, multiplication, lengths/norms of vectors, orthonormal bases and rotations, matrix representation of linear transformations, projections, geometric interpretations of determinants, what positive definiteness is and how to test for it, eigenvalues and eigenvectors, inverting 2x2 matrices by hand, general awareness of special matrix decompositions (Cholesky, spectral, singular value) Logarithm and exponentiation properties in general, representation of exp(x) as lim n-> infty (1+x/n)^n Software: Mathematica is useful for checking your work, or just get good at using Wolfram Alpha if your school won't have a Mathematica student license. Knowing enough R to do simple simulations to check answers non-analytically is good too (e.g. how to generate a bunch of realizations from an exponential distribution and transform them, plot them against whatever distribution you're claiming they have)

I can comment a bit on UW's program. It's new as a formal master's program and the first group of seven students is just finishing this spring, so not a lot of outcomes to report on yet. Two of the students are moving on to statistics PhD programs (one staying at UW, other going to Yale). The remaining are job hunting, to my knowledge, mostly looking at data analyst positions, I think at least two have accepted job offers already in the Northwest. The current first year master's cohort is larger, 18 students, not sure what their goals are generally. Most are from China, though there are a few US citizens. Next year, even bigger I would guess? RA funding is unlikely to happen for master's students, as the overall TA/RA situation is already tight for the PhD students and they get priority. I think some are master's students are graders for undergrad classes, don't know of any first years who TA or RA, one of the second year students TA'd. So in terms of funding to offset the cost of the program, in practice this is pretty limited so far. The coursework is theoretical and shares a lot of the requirements with the statistics and biostatistics PhDs. If you are thinking at all about going on to a PhD, it's definitely good preparation, the coursework is undeniably rigorous. The first year theory sequence and the second year methods sequence involve a lot of homework and are particularly time consuming. For the most part, the offerings are traditional statistics classes, which are interesting and challenging but not direct professional training. For machine learning offerings specifically, see here (note the theoretical emphasis). There are some electives that could be useful to someone interested in a data science career who wanted to get a portfolio started, such as the popular new "machine learning for big data" class which demands a lot of programming and a substantial final project. The required first year linear regression class and elective nonparametric regression class also usually have final projects and a poster/presentation, again, perhaps nice for portfolio building and interview fodder. I haven't heard too much about the required master's capstone class, but I think it involved some sitting in on consulting sessions and seemed more practical than most of the other coursework, may be somewhat useful for learning how to bridge the divide between real problem posed by a non-statistician and a solution grounded in statistical theory. The department forwards on job postings it receives to students, seems like a few per week, emphasis on local employers but some things from all around the country, but doesn't maintain any kind of job database or resume service. You can go through general UW career advising of course, job fairs and whatnot. Not sure how that compares to whatever UCLA does for its students. The university has an overall very strong reputation in statistics and computer science, though, and I don't think it would be hard to get employers to take your training seriously coming out of UW. Unfortunately can't tell you if this is worth the premium over a shorter cheaper program at UCLA.

Crime isn't necessarily the main reason to avoid the U District (although the frequent UW alert emails about muggings are not comforting). The U District kind of sucks because not only do you pay a premium for proximity to the UW for units that are often not kept in great condition, but there isn't actually that much to do. If living near fun things matters to you, I would advise against the U District. The Ave on a Friday or Saturday night is surprisingly pretty dead. A couple of places have good beer and decent food (Shultzy's, Big Time), but most of the bars on the Ave are just college bros drinking cheap pitchers of Bud Light, not even divey in a good way. You have to get outside of the U District to Wallingford or Roosevelt to find bars that have more of a neighborhood feel. Very limited places offering outdoor seating for the 5 months out of the year when we have nice weather. No good for cocktails or wine. Certainly a lot of cheap eats, but you get sick of the basic Thai, Vietnamese, teriyaki, or gyros offerings, especially when you have these for both lunch and dinner because you're a busy grad student and that's all that is convenient for you to pick up. For other kinds of food, brunch on the weekends, or really anything even the tiniest bit more upscale, you need to leave the U District. Far too many frozen yogurt and boba places. The concert venues are almost all in Capitol Hill or Ballard, you won't see much live music in the U District and will have to bus/bike. The nearest parks are still outside the U District proper (Cowen/Ravenna to the north, Green Lake to the north west). The main "fun" advantages the U District has over some other places to live are coffeeshops (still not amazing compared with Capitol Hill/Ballard/Fremont, but Trabant and Cafe Allegro are good to work in, Herkimer if you go to the far north end) and the movie theaters (Sundance, Grand Illusion, Varsity, a couple others whose names I don't remember).

Given your desire for a non-academic career, can you say more about what kinds of industry jobs you might be interested in? And are you sure you need a PhD in statistics for those jobs, as opposed to a master's degree or PhD in some related but potentially more relevant area like operations research or financial math?

We've discussed this before, but I think you might be underestimating how many stat/biostat applicants have this profile? I think it's fairly common now for programs that are even a bit less selective than UW and CMU to reject applicants who would meet these criteria. ppham27 looks to be in good company. Just anecdotally from this year on this forum, one could probably find more: look at aridneptune, who is a US citizen with a 4.0 GPA from UNC, math major, still was rejected by Berkeley stat, UW stat, Wharton stat and waitlisted by UW biostat. Or 3.82 GPA in math from an elite liberal arts college, female, US citizen; rejected by UW biostat, NCSU stat, Emory biostat; waitlisted by UMN biostat and Pitt biostat. Unmentioned weaknesses, who knows, but the happy ending is that they still got into other great programs. I did well overall when applying two years ago, but I was still rejected by CMU stat + joint stat/policy and Harvard stat as a female, US citizen, math major, attended high ranked LAC, 3.93 overall GPA, 3.98 math GPA, tons of proof-based math coursework, department routinely sends students to good math and statistics graduate programs, great recommendations. I have to assume there are enough applicants out there now with strong profiles that top places have the luxury of picking among them based on department-specific preferences about research fit, depth of prior research experiences, depth of statistics coursework, computational skills, who knows.

Yes, a fair amount of of open drug use and dealing around Belltown (mostly 3rd and Pine vicinity). Not outright unsafe, but less safe than most other areas in the city. Not a place I like walking around at night coming back from The Crocodile, and even less so as an easy target by myself with a laptop bag and smartphone. You can look at a real time crime map to get the flavor: http://www.seattle.gov/police/crime/onlinecrimemaps.htm, or just Google " 3rd and Pine". I certainly wouldn't want to pay $1600+ a month for an "open one bedroom" (the new euphemism for a studio) in Belltown only to have to deal with limited grocery stores and needing to walk through the heart of sketch twice daily to get to and from the UW-bound buses, which are going to be 30+ minute rides each way. I don't know of any UW students who live around Belltown. I think it's mostly people who work downtown or at Amazon.

43 or 49 buses take about 20 minutes from Broadway/John to campus in the morning, which usually doesn't have too much northbound traffic in the 8-9 am window I travel in. Getting back to Capitol Hill from campus in the evenings generally takes longer because of traffic, maybe 25 minutes. Worst case scenario you are on the 43 and have to wait for the Montlake Bridge to open up and allow boats through, which could set you back another 20 minutes with traffic backups.

Completely agree! I think whether industry placement from a program is good or not is subjective and hard to generalize. I have no intention of working in clinical trials or finance, for instance. Strong placement into those fields is largely irrelevant to me (and perhaps even a drawback if it indicates allocation of resources away from things I care about), but ought to be a selling point for someone who might want to work in one of those areas.

You'll probably have to take some PhD-level courses (e.g. advanced statistical theory) that you don't have an equivalent for, so I would expect to take some new courses in your PhD program. As for skipping/retaking coursework, that very much depends on the program. For example, just within UW, the statistics and biostatistics departments have nearly identical course requirements but completely different approaches to letting their students place out of master's courses. Stat lets you skip all the first year master's coursework with no hassle just by talking with the grad coordinator, while biostat has numerous formal hurdles listed here. The result is that UW stat PhD students coming in with some grad-level stats courses are jumping right to second year PhD work (which has worked out to be about half of incoming students in the past two cohorts), while I only know of one biostat student with a master's degree who was able to skip any courses (and even then, I think only the first year applied classes).

I don't understand how to identify "the best schools for Industry/Government placement". "Best" in what sense? Everyone who took a non-academic job is in a position doing interesting and challenging statistical work? Highly paid positions? Well-known companies? The department is a feeder for particularly industries or agencies? Or focusing on the negative, what would bad industry/government placement look like? Again, I don't really get this exercise, but UW skews academic and there's not a lot of industry or government PhD placements to speak of. That may evolve as the recent glut of students in interested in machine learning graduate.

IMO these are not great reasons, no disrespect to your mentors but when receiving similarly well meaning but not-very-helpful not-very-informed advice from a lot of the non-stats people I worked with, this was just added noise making a decision harder for me. I'm sure you have your own location opinions, but let me just add a counterpoint to the "small college town" argument. If you stay in academia, you're fairly likely to end up in a college town after your PhD for a postdoc or professorship, so you're not really missing your one opportunity to live in a collegiate area if you choose Pittsburgh over Ithaca. I personally see it as the opposite tradeoff: I would want to go to grad school in a city I love and enjoy my (sadly dwindling) youth in an exciting place where I can enjoy the non-academic things that matter to me most, have some friends who have no current affiliation with the local university, etc. As for the "overall a stronger and more interesting institution" argument, I'm unmoved. I think that's just saying that Cornell is Ivy League institution most laypeople have heard of with a full array of departments, while CMU is more specialized, maybe like a poor man's MIT. I think for any job you would reasonably want to use a statistics PhD (or consolation master's?) to attain, though, both departments are going to look great, CMU maybe a little better. I don't have the impression that researchers in Cornell's statistics department have stronger connections to other departments at Cornell than CMU researchers do to other departments at CMU, which would be the main reason something like "interesting"ness should matter.

With an eye towards improving your profile for PhD applications: Agree with everyone else that you should definitely take the advanced calculus class. A course in proof-based analysis is a common prerequisite for PhD statistics programs, and even if you apply to programs that don't require it, it still looks good. You'll also have to put off PhD theory coursework until you've completed that, so you may as well get it over with now. If you are equally interested in the Bayesian stat and time series material, you might take whichever class will give you the opportunity to impress the professor more so that you could hope for a strong recommendation (unless you are already confident about who will write your letters). Recommendation letters are very important and you'll want ones that credibly say you're one of the best students in the program and that you have the intellectual ability + motivation + curiosity to be a successful PhD student. Talk to other students who have taken these classes or had other classes with the instructors. Maybe the class that involves more project work might be preferable to a lecture course that is purely homework and exams, for example, in terms of the professor getting to know your interests a little better. Or maybe one class is taught by a professor who is known to be very accessible and helpful to students. Maybe one is more challenging and theoretical than the other, and doing well in it would be a good sign. In any case, ask around.