Stat Assistant Professor

Thanks for the clarification! Based on your description, the 209ABC sequence is definitely a PhD-level real analysis sequence (you need to know concepts from measure theory like product measure and Lebesgue integrability in order to study/understand Fubini and Fatou Lemma), and it is not strictly necessary for the OP to take this -- of course, if they did and performed well in this PhD level sequence, it would certainly strengthen the application and they may even be able to try a school like Duke or University of Washington. Even if they don't take this sequence, they should still be a competitive candidate for schools like TAMU, PSU, UMN, and Purdue if they have a strong GPA. OP, it's up to you if you want to take the 209ABC sequence, but the 151ABC sequence would suffice for getting into a Statistics PhD program (the Stat PhD program would teach you Fubini, Lebesgue dominated convergence theorem, and Borel sets as part of their PhD curriculum).

1) Is the "Advanced Calculus" series the same as a baby real analysis class? And what is the content of the Math 209 ABC sequence? I know some math graduate programs insist on reteaching their first year grad students the "accelerated" undergrad analysis sequence at the level of the Rudin textbook unless they can test out of it. If the Math 209 ABC sequence is stuff like convergence of sequences and series, pointwise vs. uniform continuity, metric spaces, Riemann integral, etc., then I absolutely recommend that you take this sequence. You need to be fairly comfortable with doing mathematical proofs at this level if you're going to be doing even harder ones in a Stat PhD program. 2) If you can keep your GPA above a 3.7, I would say you can actually aim higher than ASU and FSU. The "sweet spot" for you would probably be TAMU, Penn State, and UMinnesota, but you could certainly afford to try a school or two like NCSU or University of Washington if your GPA is good. 3) I think the summer REU is a good thing to have on your CV. Doing an undergrad thesis might also help, but this isn't strictly necessary. 4) I think your list of classes is good. It may be helpful to take a proof-intensive linear algebra class as well (not sure if Math 131 contains proofs... if so, I think you are good). 5) I don't think they will care that your lower division coursework was at community college. Just do well in your upper division classes at UCR. 6) You probably don't stand a chance at a school like Harvard, UPenn Wharton, Chicago, or Stanford because your degree will be from UCR, but I think you should be fine for schools at the level of NCSU and below (and may even be able to get into a school like University of Washington, which does accept some students from schools like Utah State, etc.). UCR is not a super-low ranked or unranked school (it's still in the top 100 according to USNWR), and it's not an unknown regional school or tiny, nonprestigious SLAC. And since you are a domestic student, even if you were from a regional school or a tiny SLAC, you could still get into schools at the level of FSU if you had a really high GPA. 7) You need to take the general GRE and ideally score 160 or higher on the Quantitative section, but I would not recommend taking the Math Subject GRE unless your math GPA is hovering around 3.5 or lower (then taking the Subject test and scoring well on it might alleviate some potential concerns about your ability to handle graduate-level courses in stats).

I think you have a very good shot at some of the schools you listed, but it's a very top-heavy list (for both statistics and math -- NYU Courant and UCLA are basically two of the best applied math schools in the world, and NYU and Princeton are even more competitive to get into than Stanford of UCB Statistics). Competition for international applicants tends to be particularly fierce, so it might be a gamble to apply to *only* those schools. I would suggest applying to a few "safer bets." WIth your profile, you probably don't need to apply much lower than NCSU or UMich for Statistics or University of Maryland-College Park for Applied Math. As for your concerns: 1) If you're finishing in the high first class at Oxbridge/ICL, I think you're golden. 2) It is more important that the letters be strong and say things like how you were one of the best students they've taught in years than that the letter writers be well-known. Well-known recommenders who have connections may matter more in fields like computer science or physical sciences where students are recommended for acceptance by a PI (so obviously, a letter from a collaborator of the PI would look great). But in Statistics and Mathematics, admissions decisions are made by the department, and PhD students are not accepted into PI's labs. 3) I think adcoms are generally neutral about industry experience (doesn't tend to hurt or help the application). There are PhD alumni and students at UC Berkeley who have started their PhD over 10 years after they graduated undergrad. 4) See first paragraph.

I don't think the W's will be a big deal, as long as your other grades in lower-division math courses were acceptable. You should still be able to get into a Statistics MS/MA program. I could be wrong, but I also don't think you need real analysis for quantitative psychology. You certainly need familiarity with things like SEM, factor analysis, multivariate analysis, etc., but it seems like overkill for psychology students to study the mathematical theory and foundations behind them.

It sounds like you should follow your instinct and attend JHU for all the reasons you mentioned. For PhD admissions in biostat, you need real analysis at the minimum, and advanced proof-intensive linear algebra is also helpful. Research experience among PhD applicants is becoming increasingly common as well, and based on your description, it seems like you would have better opportunities to take more math and serve as an RA at JHU. These things will strengthen your PhD applications.

A "W" will look better than a C for either of those grad programs. As long as you are not planning to get a PhD in Statistics, it should be fine to withdraw from this class and not take Real Analysis period (for most Statistics Masters programs, you only need to have completed Calculus I-III and non-proof based linear algebra).

Yes. 1) The first two years, focus on coursework but don't obsess too much about grades. Focus on learning and making sure that you know the material well enough to pass your qualifying exams. That's really all that matters -- that you pass the qualifying exams. Whereas many undergrads are distressed by an A-/B+ vs. an A, this is really not the point of PhD education. I got two B's in my PhD program and not one of the postdoc positions I interviewed for asked about grades. They were only interested in my research. 2) Once you are mostly done with coursework (you may have a couple of electives still to take in your third and/or fourth year), you need to make the transition from "student" to "researcher." This is the roughest adjustment period for most students because there is no real structure like courses and exams, but you just have to be persistent. If you don't work well with lack of structure, then you should try to "force" yourself to spend [x] amount of time on research weekly (allowing for mental breaks and time off if you really need it). You will likely understand zilch of the papers you are reading at first, but just be diligent, ask questions, re-read several times until things "gradually" start to sink in, and know that it gets *much* easier to read dense, heavily technical papers with more experience. Also, learn to be okay with failure and rejection (many inexperienced grad students will have their first paper(s) rejected by journals). Know that it is okay to fail and to be rejected at this stage (most research is "hand-waviness" and trial and error until something works), as long as you take it as a learning experience and seek to improve yourself.

It may be a good idea to spend some of your free time reviewing: - basics from Calculus I-III (specifically: the derivative rules like product rule and chain rule, integration and rules for integrals like u-substitution and integration by parts; partial derivatives, double/triple integrals and changes of variables/Jacobian matrix, and sequences and series... you don't really need to review things like cross products or torque) - linear algebra (both basic and proof-intensive) - basic real analysis (at the level of Abbott's textbook). It is natural for a lot of people to forget things from these classes if they have not encountered/consistently worked with those tools in awhile, but it is a good idea to refresh your memory before starting. I find the MIT OpernCourseWare (OCW) to have useful resources for reviewing these things.

No, they should take advice from someone who is working in academic statistics and is intimately familiar with the admissions process for statistics, like a stat/biostat professor, postdoc, or current Stat PhD student. The admissions process in Statistics is completely different from that of Computer Science. My advice echos that of other professors from the departments I have been affiliated with and other professors on this forum, and many posters on this board would tell you that my predictions of their chances were spot-on (some even messaging me personally to thank me and tell me where they were going to attend). My advice is totally objective, not borne of being a "hater." If the OP had indicated that they attended ISI for undergrad, I would have recommended that they aim for Harvard and Penn Wharton Statistics. Given that their alma mater was not ISI, this would ultimately be a waste of money. Even at my mid-tiered ranked stats PhD program, many international applicants from China and India were automatically rejected if the adcom had never heard of the undergrad institution. However, this could be overcome with a Masters degree from a respectable program in the US or Canada. At top-tier programs, however, it is basically impossible for international applicants who did not complete their undergrad at the top schools in their home countries to get admitted.

Computer Science admissions is very different from Statistics. In CS, a publication or two in a top conference and reference letters from professors who have strong connections to the lab PI's indicated in the application/statement of purpose can greatly mitigate a lower GPA. But this is not the case at all for Statistics. In Statistics, prestige of undergrad institution and undergrad GPA play a critical role for international applicants, and a 3.1 undergrad GPA would make it impossible for *any* applicant (international OR domestic) to get into the top-tier Statistics PhD programs, even with a Masters degree (this ugrad GPA would not preclude them from getting into a school in the range of 20-50 though, given strong performance in a Masters program). I think someone who is in the actual field of Statistics would know better than you.

Lol, $3000 for a consultant? I would not listen to this bad advice, OP. Most PhD programs in Statistics in the USA are swamped with international applicants. The top tier programs like Penn, Stanford, and Harvard can take their pick and typically do not consider any applicants who are not from the top 5% of their graduating class at ISI, Peking, Tsinghua, etc. The OP sounds like a strong candidate but the reality for international applicants is they need to be the top students from the most elite schools in their home country in order to have a reasonable shot at these schools (even if they get a Masters degree from a school like Duke in the U.S., that is still not likely enough to overcome the undergrad alma mater issue because graduate school grades are usually more lenient). We provided the OP with a realistic range of schools they should apply to and even encouraged them to try a few "reach" schools like NCSU (Harvard and Penn Stats would almost certainly be a waste of money though).

Also, you do not need to retake the GRE or the math subject GRE. The subject test is only absolutely necessary for Stanford Statistics (as far as I know) and strongly recommended for UChicago Statistics and Columbia Statistics. I think those schools are most likely outside your reach anyway. The highest I would aim for in your case is NCSU Stat and Michigan Biostat (though if you have the funds and wanted to try UNC Biostat or UWashington Biostat, I suppose it couldn't hurt).

Those are less of a reach than Harvard or UPenn Wharton but it is hard to ascertain your chances only because the competition among international applicants is very fierce and your undergrad institution was not ISI or IIT. Your paper submissions and your MS from Duke are certainly pluses though, and I think it is worth a shot to try a few "reach" schools like NCSU.

Your list is very top-heavy. I don't think you have a chance at Harvard Statistics or UPenn Statistics because your undergrad institution was not ISI or an IIT. UPenn Wharton and Harvard are amazing schools for causal inference, but these schools are very tough to crack (Penn Wharton cohort is only about 4-5 students every year) and only seem to accept those from elite universities. Your MS from Duke and your MS research will probably help though. I would trim down the list of top-tier schools and maybe only apply to 2-3 of them. I could see you getting into UPenn Biostatistics (not Wharton) or UCLA Biostatistics. For Statistics, you likely have a good chance for schools in the range of UFlorida and UIUC.

Xiaofeng Shao's work is impressive but he is not a Bayesian. And his recent work doesn't seem to be focused at all on spatial statistics. Bo Li seems to do the type of research you're interested in though. It seems like of the schools you listed, Purdue has the biggest group of professors who work on spatial statistics: http://www.stat.purdue.edu/research/spatial_statistics.html There are also a few very strong faculty who do work in Bayesian statistics at Purdue as well, including Faming Liang and Guang Cheng, although I am not sure how much spatial statistics research they do.

Boston University doesn't have a separate statistics department; its Mathematics & Statistics PhD programs are both under the Mathematics department. So the rankings for BU would probably be under Mathematics rather than Statistics. To the OP: what are your interests? Are you more interested in mathematical/theoretical stuff or more applied (e.g. in clinical trials, medical imaging, meta-analysis, etc.)? If the former, BU would probably be a better fit than UMN, although I am sure you could find at least a couple faculty at UMN Biostatistics who do some research on more mathematical statistics as well. But if your interests are mainly applied, then UMN is better probably.

Outside of a few super-star researchers who are constantly publishing in top journals and conference venues (e.g. Tony Cai, Michael Jordan, Jianqing Fan, etc.), most potential PhD advisors will probably not be publishing more than one article a year in JASA/Annals/Biometrika/JRSS-B (one every 1-3 years seems more common). It's probably important that they *have* published in these venues in the past, so they know what level of quality/novelty is expected for publication in these venues. But I wouldn't say it is essential to have a PhD advisor who has 1-2 publications in top journals *every* year. To land a *really* good postdoc, I would say that nearing graduation, the PhD student should have either: a) 1 publication accepted/in revision in a top-tier journal like the ones you mentioned or a top-tier conference (ICML, NIPS, AISTATS), or b) 2 publications accepted/in revision in relatively good journals (say, top 10 like Statistica Sinica, Journal of Multivariate Analysis, etc.). And there should be at least one other work in preparation on top of that. Any postdoc or PhD student who has 2 publications in JASA/Annals/Biometrika/JRSS/Biometrics will most likely be able to land an R1 job at a department in the USNWR top 80 (say). Of course, if you want to get a job at Stanford, Berkeley, Harvard, Columbia, etc., you'll probably need more than that, with the most competitive job candidates having like, 4-5 in top venues. The people who tend to get the jobs at the most prestigious departments also tend to be graduates of similarly ranked prestigious programs, so it seems difficult to "move up" significantly if your PhD is from a lower ranked institution (although a former PhD student of my advisor was able to get a faculty position at Duke, and our program is ranked a couple tiers below Duke -- in this case, the postdoc at a top 10 program most likely helped). I myself am personally okay with going to a department in the top 80, but not necessarily the very top-tier, so I am working on getting two publications in very top venues at the moment.

It looks like UNC STOR does have some joint faculty who are in UNC Biostatistics: https://stat-or.unc.edu/people/adjunct-and-joint-faculty Some of the joint faculty from Biostat are pretty well-known (e.g. Ibrahim and Kosorok) and whose work appears in top statistics journals like JASA and Biometrika (not just journals like BMC Bioinformatics, Biostatistics or Statistics in Medicine, which also contain good content but may not be as helpful in getting a job in a Statistics department). So it would probably be fine to go to UNC STOR and have one of those Biostatistics professors as a co-advisor for your thesis. Besides interacting with those with joint appointments, I'm not sure how much overlap there is between UNC STOR and UNC Biostat there is though -- the coursework is separate, and as a whole, the research emphasis seems to be vastly different between the two departments.

I would go with the cheapest option which sounds like UMinnesota. Minneapolis is not a bad place to live, and you only need to be there for two years if you find you don't really care for it. Living in NYC is really expensive, and it doesn't seem like UMN grads are at any disadvantage getting jobs outside of the greater Minneapolis region (that might be an issue if it was a no-name MS program but UMN is quite well-regarded for its biostat program).

UNC-STOR is a very solid department as well. I admit I'm somewhat less familiar with the work of their department than NC State (I follow the work of several profs from NC State). The coursework and research at UNC do seem to lean more heavily theoretical (e.g. they require two semesters of measure-theoretic probability, and probabillity theory and stochastic processes seem to be two of the department's main strengths). I think one would also get very solid training at UNC and be well-positioned for a good postdoc with a degree from there. I'm sure it depends a lot on the PhD advisor too. For example, even if one goes to a mid-ranked school like Rutgers or UIUC (which are lower ranked then UNC), a student who has Cun-Hui Zhang as their advisor (say) and publishes one or two good papers should be in excellent position to get a postdoc at a top 10 department.

Congratulations on your acceptances at excellent schools! I would rank University of Michigan, Duke, and NC State as top 10 programs. I wouldn't say that there is any discernible difference in rigor or coursework preparation (although Duke's coursework seems to be more heavily geared towards preparing students for research in Bayesian statistics). It seems as though Ann Arbor is your preferred geographical location of those three, so it's probably best to follow your instinct and go with UMich. You would be in good shape to get a good postdoc with a PhD from any of these schools (of course, that will also depend on your publications as a PhD student and the reputation of your PhD advisor). The other schools you've gotten offers from are also good, but UM, Duke, and NC State are better IMO. The USNWR rankings are fairly accurate, IMO, though you could make the case for moving a few of the schools up or down (for example, I would probably personally rank Yale and UT-Austin higher than their current positions).

You can turn down offers and accept one at any time before/until April 15. It shouldn't cause any harm to your image (the only things that would seriously damage an applicant's reputation would be stuff like finding out they falsified information on their application or engaged in research misconduct like plagiarism). Adcoms will most likely not even remember your application next year when they have another few hundred they need to go through.

UT Austin is a solid program with some good academic placements (they have placed some PhD graduates at postdocs at UC Berkeley and Princeton). However, Duke is also probably the best school in the world specifically for Bayesian statistics. If you really want to go to Duke, I would just wait it out until early April. You don't need to officially accept any offer until April 15.

I vote Minnesota, if it's really half the cost of Emory. UMN has a very reputable Biostatistics program.

I can't comment much on the programs, but re: location. It should be noted that Providence, RI is about a 50-minute train ride and a 50-60 minute drive from Boston. And probably way cheaper to live in Providence too. So you would still have fairly close proximity to Boston/Cambridge and its hundreds of biotechnology companies if you attended Brown (a lot of people make the commute between the two cities every weekday).