wine in coffee cups

UW's curriculum is not Bayesian but it doesn't lack for faculty and students doing research using Bayesian methods, particularly in the statistics department. And combining both of OP's research interests, Wakefield has supervised a bunch of students from both departments in Bayesian epidemiology.

^ Good find, 845 applicants sounds much more plausible. As another sanity check number, Harvard econ PhD says over 700 applications each year, and Yale should have a similar volume. I bet either Yale accidentally added a zero in the letter they sent to applicants or they meant 10,400 applications to the entire graduate school of arts and sciences (not a useful number). Bringing this back to statistics, by altering the URL for Yale's econ PhD data, there's info about Yale's statistics PhD program, which I hadn't seen before: http://gsas.yale.edu/sites/default/files/department-files/statistics_0.pdfThere were 85 applicants to the Fall 2014 class and 4 enrolled students. Only 16 students registered in the entire program.

Geographic preferences are a fair starting point, though you've omitted a number of programs that logically should be there based on the regions you included (e.g. UNC, Chicago). I would prioritize them based on your research interests, which you didn't mention, but you must have some stuff in mind if you're coming from engineering and making a switch now. There's not like a central site that lays it all out for you unfortunately, you have to figure this out by talking to statistics faculty and looking at department webpages and lists of publications. As Bayesic said, some of the programs you listed are quite small and idiosyncratic in their admissions (Yale, Northwestern, NYU Stern), so I'd only apply to those if there actually is somebody there you'd want to work with. Add some bigger departments to your list like NC State or Michigan.

Congratulations on your success with the biostats and OR applications. Where did you end up? somehow I don't think a lot of applicants will be comforted seeing someone with a 3.96 GPA from an Ivy League etc. getting rejections from statistics programs! Nitpick: you're off by an order of magnitude on the Yale econ PhD application totals. Peterson's says 776 using what is probably old data, so around 1000 last year is plausible, but not 10,000.

Sounds like you're at NC State. You should talk to the faculty there who have helped undergraduates apply for top PhD programs to see what they think. I'd generally emulate whatever has been successful for previous students, which might include research with faculty over the summer. I wanted to plug the Budapest Semesters in Mathematics program as an alternative study abroad that can be actively helpful in math/stats graduate admissions and need not conflict with your last undergrad summer. They offer a lot of math classes (most recent semester syllabi), it's as easy or hard as you make it with your course selections, and the BSM people are used to dealing with American PhD applications for their alumni. I did BSM at the same time as a couple of NCSU kids, many students in the statistics and math PhD programs at my university have gone, and I would enthusiastically recommend the experience.

Just do your thing, enjoy your new university, worry about what's realistic later once you have more information. For graduate school, extracurriculars don't matter, except perhaps to the extent that they have faculty advisors who get to know you better through them and will write nice things about you later. Participate in clubs/societies if you're genuinely interested, but don't expect them to be a way to make up for poor grades.

I think you will be competitive for those and other biostatistics master's programs, but who can say anything for sure. For context, Berkeley biostat MA reports a 25% admit rate with a mean GPA of 3.62. Based on what you've written here, your main weaknesses appear to be downward grade trend (A-'s and B's) in more advanced math classes and a lack of a mathematical statistics course (take this and aim for an A). ginagirl's advice is good. I'm surprised a 3.78 GPA places you in the top 2% of your class. For undergrads at my (also top 20) public university, 3.75 is top 10%, 3.86 is top 3.5%. It sounds like you might be coming from a university with less grade inflation, so you'll want at least one of your letters of reference to put your grades in clear context (maybe that real analysis II class was harshly graded).

These are strange criteria for picking a grad program. The overall enrollment won't have much bearing on the graduate experience (though department and program size are things to care about). You're going to be spending all your on-campus time either in classes, in your office/lab, around department common areas, or in a library or coffee shop, so I'd re-think how much the campus vs. integrated urban setting criterion matters to you if you'll be in the same one or two buildings all day. It's nothing like the undergraduate experience. The first two subfields are a focus of applied math departments. Data analysis is the focus of statistics and biostatistics departments. Data visualization is more of a niche topic addressed by some CS and statistics departments as well as in various programs related to HCI and informatics. Based on your comments about structured programs not being interested in independent research, I'd recommend against applying to PhD programs. A PhD is years spent lacking structure banging your head against a wall doing research in preparation for a career to continue doing that. I think you're going to be happier with a master's. Funded master's programs will involve TA or research assistant work to earn your keep, though, and often require a thesis involving original research. If you are really research averse you might prefer a coursework-only master's. Lack of funding is not so bad, though, since the kinds of careers you get after applied math/statistics/CS degrees pay well enough. It's usually a lot cheaper to go to a public university at in-state rates than to a private one, too.

Nothing should stop you from applying to places like CMU or Duke unless you wouldn't actually be interested in going there if admitted. (CMU might not be a great choice if you're interested in biostatistics though.) You could spend a bunch of time mucking around department website hell (and eventually you will when you apply), but I think talking to faculty at the SIBS you're doing this summer is the best way for you to learn about research areas and program/faculty strengths. That's pretty much what SIBS is for.

Keep the grades up, try to get to know a few faculty in the next year and a half (work on research or a thesis with them, do reading credits under someone, go to office hours, ask for advice about your plans), you'll be fine. I think you have a good chance at more selective departments than the ones you've listed, so I'd apply based more on your research interests and location preferences when the time comes. Realistically, many of those are going to be safeties for you if you remain around the top 5% of your class at a good undergrad, so maybe prune some of them out and add some more reaches. Rice is a small department, but I think it's a actually good match for someone who is interested in biostatistics. They have a few faculty there working on statistical genetics and genomics (off the top of my head, Vannucci and Allen), and they have lots of affiliate faculty from MD Anderson, which is a major cancer research center with a biostatistics department.

janh: what citizenship do you have, where did you earn your bachelor's degree, and in which country are you working? What were your grades like? From your writing and the "advance mathematics" course description, I am guessing you didn't do undergrad in the USA. This would make way more of a difference for your admissions chances than how you did on the GRE general. Your coming from a foreign university would mean evaluating your academic preparation is going to be challenging for the faculty on admissions committees at American universities, unless you were a math or statistics major at a top university that has a history of sending students to American statistics PhD programs (e.g. Peking U, Tsinghua U). On top of that, some funding lines are restricted to US citizens/permanent residents only, so space for international students is even more limited, leaving hundreds of foreign applicants competing for a few spots. For both of these reasons, getting into a good biostatistics PhD as an international applicant is very tough -- see this thread for discussion.

No, I think it's the opposite actually: they expect it's reasonably likely you won't return if you take a leave of absence. And I think they'd rather have the chance to address some of your grievances (if you have anything specific) and keep you rather than have you drop out without warning. Are you ruling out the possibility of making your current program workable? I'm curious why.

Oh, I assumed you had something more concrete planned for time in-between. If what you're looking for is a way to cultivate research interests, why not join a working group or start an RAship with faculty in your department? Like, normal grad student exploring topics stuff. Nothing like jumping right into an area to to help you figure out whether it's interesting to you or not, and you're not committing yourself to an advisor or dissertation topic. If you truly feel that you need to work away from your university, I would see if you can arrange an extended internship (e.g. a 6 month term) rather than taking a full-time position, and go on-leave rather than dropping out. A company that actually does enough serious statistics-related work to potentially steer you towards a research topic (e.g. some of the big tech companies) might be amenable to this. I imagine most places would prefer to employ you on a defined term rather than have you suddenly up and quit to go to school somewhere else, which is another bridge you risk burning under your original plan. On-leave status would be viewed more favorably by your program than outright dropping out. I think the discussions you need to have with your advisor and the grad chair about how you are doing in the program and your future will be much more productive if you go on-leave to do something in particular rather than quit, too. Maybe they will even agree that the program isn't working for you and can be actively helpful in resolving your issues or smoothing the way to transfer, who knows? Dropping out is the nuclear option and it doesn't sound like you are certain enough about anything for that to be a good idea.

The catch is that your recommenders say will matter 10x more, doesn't matter how much you downplay your previous program. Those letters are the most important part of your application! Members of the admissions committee will be extra keen to hear what the faculty at your current program think about how you stack up relative to other students and whether you are a good candidate for a PhD. You have little control over whether and how your letter writers frame your leaving their program beyond whatever reasons you share with them. It will not look good if what they write is in any way at odds with what you write. There is some relevant discussion in an old thread about transfers a couple of biostatistics faculty who post, and a little more about recommendation letters from the perspectives of those same faculty starting think to succeed in your plan, you need a firm handle on exactly why you are leaving and why you will be happier in a different stats PhD program. Then whatever those reasons are, you need to figure out how to make them sensible and palatable to the faculty in your program who will write your letters so that they can support your move. This means handling your withdrawal this semester and managing those relationships very, very carefully. You just aren't going to have great recommendations if they don't think you left for a good reason or know what you want. Also, what are you doing in the year in between leaving and enrolling somewhere else? Is it going to be that much better for you than completing your second year in your current program?

Yeah, TakeruK is totally right that this will depend on the PhD program. For an extreme example, the UW biostatistics and statistics PhD programs handle previous master's degrees completely differently -- even though these programs have the same required PhD courses! Someone coming into UW statistics with a background that included a Casella & Berger-equivalent course (even if they don't have a master's!) and previous courses in experimental design and applied regression would be able to start the PhD theory and methods sequences right away. There's some details about minimum GPAs you have to achieve in the PhD theory courses, but it basically shaves a year off of coursework requirements for people who took some grad-level statistics before enrolling. People with master's degrees from other universities can finish the statistics PhD in as little as 3 years (which I don't think is a good idea because you're more competitive on the job market with more time -> more publications, but that's a different story). The UW biostatistics department is way more anal about their requirements. If you have a previous master's degree in statistics or biostatistics, you still need to take the master's qualifying exam to be able to skip the first year courses (here is last year's). Further, you need to pass that exam at a higher level than that required of the PhD students who took the exam the previous spring after finishing the first year courses. For context, less than 2/3 of the current UW students who take the exam score at that level, and most spend several hours a week in the quarter prior preparing for it, so it's not the kind of thing you can just show up to and expect to ace. I've never heard of anyone successfully skipping all the first year coursework in biostat because of this onerous exam requirement, it's rare. So a previous master's doesn't take *any* time off the UW biostat PhD relative to people coming straight from undergrad, still 5+ years for just about everyone.

I would take the fully funded option. Where did you get the impression that you would be able to do "meaningful research" in the UW master's program? It's a coursework-only degree with limited opportunities for research.

This part is really important. You can leave your current department for whatever reason you like, but then you're going to have to make the case in applying to other programs why it'll be different there. I think it's a big problem if you can't articulate why some other institution is the place for you while your current one isn't, both from an admissions perspective and from a, like, just being satisfied with your life choices perspective. If the problem is that a PhD in general isn't right for you and not specifically that the department isn't right for you, then you're still going to be unhappy elsewhere. I have a few friends who left their original programs and went to different PhDs, but all involved a change in research focus that made their previous program not suitable. And no, I don't think hiding that you were a PhD student in that previous program is prudent, or even necessarily doable. It's likely that at least one of your recommendation letters will say "I had m.cyrax in my X course while he was still a PhD student here in the Y Department of Statistics". Frankly, a letter that didn't mention this fundamental fact would have to lack context in other ways so as to be not a very good letter, I think. There's also probably going to be a Google breadcrumb trail indicating you were a PhD student there, and if you're a serious candidate for admissions somewhere, there's a decent chance you will get Googled.

Here's my contrarian vote for the advanced linear algebra course, assuming that's the theoretical and proof-based version of the more mechanical lower-level course. Special matrix decompositions (SVD, QR), eigenvalues and eigenvectors, orthogonality, changing bases, projections, block matrix properties, etc. are sooooo important and useful in statistics. I did well in the lower-level linear algebra class but didn't feel like I really understood a lot of this until I took the proof-based upper-level version.

If your goal is to prepare for PhD applications, in a choice between thesis vs. non-thesis programs, I personally would choose the thesis option. I assume your thesis is supervised by faculty in the Chicago statistics department. If so, that means you'll have at least one person who knows your research interests and capabilities pretty well who could write a specific and strong letter of recommendation (assuming you actually are good). Doing well in your coursework is going to be more or less expected, so I think good performance in a research-oriented theoretical statistics program will be more informative about your potential for PhD work than good performance in a coursework-only more applied biostatistics master's. Also, cost of living in Chicago is much less than in Cambridge/Boston, which is something to consider since that will add to the cost of your degree. There are few cities in the country that match Boston in terms of hellishness and stress in finding (even barely affordable) housing in commuting distance of HSPH.

I don't think the MAS degree will be better for jobs than leaving a PhD program with an MS. If you're going to be taking a good deal of PhD-level or PhD-preparatory coursework during your MAS, then you're not getting more practical applied statistics experience than a PhD student elsewhere will. A PhD student who starts working on research and then leaves will probably be more employable than a MAS student who did lots of theoretical courses (but no research) because they'll at least have some experience working on a bigger project, solving a lot of annoying technical problems that don't come up with clean homework problems, etc. Interviewers love to hear concrete examples of things you've done, so for employment, being able to take more applied classes where you have projects is what you'd want to do. As I see it, the main difference employment-wise is if it's just easier to find jobs out of Michigan than the PhD programs you're considering for geographic or recruiting connection reasons. Michigan certainly carries some national name cachet. Well, if Michigan does admit PhD students out of its MAS cohorts, then going there would be a good move for you to try to get into that particular program. I don't know if that is the case. You did say you got into PhD programs ranked around 15-30, though, and rank ~15 is pretty high and not far off from Michigan! Do you have a sense of how realistic it is to actually be enough of a standout student in the Michigan MAS cohort to get into top-ranked PhD programs? Did you find out about students who have gone on from there to PhDs? At least in my department, most of the master's students are mathematically strong students from foreign universities who do well in theory courses. Especially if you're taking mostly theoretical courses to beef up your background, keep in mind there could be non-negligible competition for good grades and recommendations. I personally lean towards recommending you accept one of your funded PhD offers, but if it doesn't feel right then so be it.

If you think those programs have an inflated reputation, which currently lower-ranked programs would you bump over them? You could try to collect data on where faculty at each PhD-granting department obtained their PhDs and implement academic hiring network rankings of the type discussed you think that's what matters.

If you're interested in "statistical genomics/bioinformatics/machine learning/large-scale inference of high-dimensional data", Stanford bioinformatics seems like the best choice. Hastie and Tibshirani from the stat department are affiliated with that program and as you know they are huge leaders in these areas. In terms of competitiveness for academic jobs, I don't think you'd be more competitive for academic jobs in stat/biostat departments coming from Penn biostat over Stanford bioinfo if you had a strong statistical component to your research. Looks like the bioinfo program has had a number of placements in top statistics and biostatistics departments like Harvard biostat, Duke stat, UNC biostat. Horvitz and Heckerman now at Microsoft Research, too, were graduates from an earlier incarnation of the program a couple decades ago. As an anecdatum, my statistics department has a recently tenured faculty member who did his PhD in biomath. If you work with someone who is a BFD in the fields you apply for positions in -- and there are numerous such people at Stanford -- it wouldn't seem to matter the exact name of your program. Also, bear in mind the growth in these umbrella "data science" research groups, like what you're seeing the Stanford bioinfo department morph into. When you're on the academic job market in ~5 years, I would guess that quite a few of the opportunities you'll be looking at will be linked to newly formed programs mixing statistics, CS, informatics, computation, and applications. A bioinfo PhD from Stanford would be no less competitive for those kinds of positions than a biostat PhD from Penn.

Admissions for the those master's programs are all very selective: Stanford's MS program has an 11% acceptance rate and average incoming undergrad GPA of 3.9 (source). Berkeley's professional MA program reports application numbers implying an acceptance range of 13-28% (source). UW's professional MS program had a 21% acceptance rate for domestic applicants and 15% for international applicants in 2014 (source). UCLA doesn't break out admissions for its PhD and MS programs separately, but the overall acceptance rate for the past few years is 16% (source). For context, that's just a bit higher than UW's combined acceptance rate (about 13%). I'd ballpark MS acceptance rates at UCLA at 20-25%. With the scary figures out of the way, go ahead and apply to some of these schools, but consider other master's programs in the UC and Cal State system. There is no or extremely limited funding for master's students at all four programs you named, which is another reason to broaden your scope. Doing directed reading with a professor, working on an honor's thesis next fall under someone's supervision, or taking a small upper-level special topics course could help a faculty recommendation from someone who knows you well. Also: don't flub the GRE general when you take it. A low score on the quantitative can get you filtered out.

Transferring departments is difficult, sure. But I don't think it's harder than getting a tenure-track job in a statistics department from a school that has little academic placement, or even getting into a PhD program much better than UCR coming from an unknown MS program. It's just none of the paths you have open to a tenure-track faculty position in a statistics department are straightforward so transferring is an option I would entertain. If you did have offers at MS programs that regularly send graduates into stats PhD programs that have the kind of academic placement you hope for, I would definitely recommend those instead. What about your current MS program offers, do those ever place graduates into good stats PhDs? If they do then I would take those more seriously as an alternative to UCR.

Thanks for being willing to share some more information. From what I gather about the UCR program, it is an *applied* statistics PhD with explicit emphasis on finding a substantive field of application as part of your dissertation. My guess is what you're hearing from them talking up their industry placement is because people who are considering applied statistics PhDs are not usually as interested in academic jobs from the get-go. I think you are probably better off going to the applied stat PhD program at UCR than unranked MS programs at even less stat-oriented departments. Also, it's not common, but you could apply to transfer into another statistics PhD program after a year if UCR ends up being too industry-oriented for you (assuming you don't burn bridges when getting recommendations, you do very well in your coursework, and you retake that GRE general).