Stat Assistant Professor

Fwiw, UFlorida has just hired a bunch of new faculty and is making a concerted effort to diversify its expertise. The new chair of the department is a biostatistics guy, and we just hired new faculty in the areas of computational statistics, spatial statistics, and machine learning. Also, only one person graduating from UF this year wrote their thesis on MCMC. The rest of the theses are on a variety of topics, including vector autoregressive models, graphical models, network models, multiple testing, Bayesian nonparametrics, machine learning, and high-dimensional statistics. I'm not sure if you are considering UF, but it is a very Bayesian department (some frequentists), and I would say, within your reach given your profile. It has traditionally been quite theoretical but seems to be diversifying so it has strong faculty in more applied areas as well.

University of Chicago is one of the best schools in the world, definitely on par with the Ivy League. For the field of Statistics specifically, I would also place UChicago in the very, very top-tier (in the top 3 programs in the country, along with Stanford and Berkeley). So as far as education goes, I vote for UChicago. That said, for industry hiring at the Masters level, location is probably a concern for companies -- particularly for international students. Companies based in NYC may prefer to hire local MS candidates, so I would factor in location as well (although I personally love the city of Chicago more than NYC). If you plan to get a PhD, this is less of an issue.

Out of curiosity: would you say that a prospective biostatistics PhD student should have at least *some* interest in research problems that are motivated by real data sets and problems from public health/medicine/biology (something which is not always the case for Statistics PhD programs)? This doesn't seem to be an issue for the OP, I am just curious. And I do see that many good biostatistics departments require their PhD students to take year-long sequences in some advanced topics from probability and statistical theory (e.g. UCLA has such a year-long sequence in inference, as well as a year-long sequence in linear models/generalized linear models). So mathematical rigor and training in theory are not in question. It's more the "day-to-day" research I am curious about. I actually applied to a lot of biostat postdocs and postdocs in more interdisciplinary environments (like the University of Chicago Booth School, where I could also work with econometricians), because I wanted the opportunities to work on real data and branch out beyond hardcore theory research. But for thesis work, am I wrong in thinking that a Biostatistics dissertation and "day-to-day" research would be much more likely to be motivated by methodology on real data sets and public health/medicine problems than on proving new theorems and the like? That's just the impression that I got from browsing the dissertation titles at various institutions (e.g. the ones at Harvard or at my current institution), but I could be wrong.

Excellent points by gc2012. I would make the decision based more on research fit. Both Statistics and Biostatistics programs at top schools will be mathematically rigorous and provide you with training in both statistics theory and applied statistics (computing, regression, etc.), so the training will be excellent regardless. And both are likely to lead to careers where the compensation will enable you to live comfortably, so I wouldn't make the decision based on money. Above all, it is most important that you are personally fulfilled by what you're doing/researching. You would be able to develop new methodology for things like high-dimensional statistics, causal inference, etc. in either a Statistics and Biostatistics departments. However, a Statistics dept is likely to be slightly more theoretical (as in, the mathematical properties and foundations of your methods -- e.g. convergence rates, tail/asymptotic behavior, properties under different loss functions, etc. -- are of greater interest than their specific application... whereas in biostats, the methodological contributions and their application to stuff like genetics or electronic health records tend to be the emphasis). A dissertation coming from a Stat department is more likely to be primarily a theoretical contribution rather than an applied/methodological one. So really, it depends on what type of research you would be most fulfilled doing . It's a matter of personal taste and preference.

One thing to consider is that if you become especially interested in biostatistics/public health applications, you can always complete a postdoc in a Biostatistics Department. My PhD will be in (theoretical) statistics, but I have applied to several biostatistics postdocs. These postdocs are still in my general area of research (high-dimensional statistics), but with a particular focus on developing methods for problems like analysis of electronic health records (EHR), missing data, and causal inference. And even then, I still have the opportunity to collaborate with faculty members on theoretical foundations of these methods (mostly in the schools' Statistics/Math departments), while learning more about the methodology in Biostats.

My Masters degree is in Applied Mathematics. For the most part, it was perfectly fine. It definitely helped to have some experience with proof-based math before starting the MS program. For Statistics MS programs, I think taking Calc I-III and linear algebra and some upper division math classes should definitely be helpful. Masters programs are also less competitive and have many "non"-tradititional students who did not study math or stat (there are a lot of econ, business, biology and life sciences majors who enroll in Statistics or Biostatistics MS programs after they have completed the mathematics prerequisites).

I was in a similar position when I was applying to Masters programs. My mathematics background from my undergrad included Calc I-Calc III and Linear Algebra but not much beyond that (I majored in Economics), so I had to take a few more math classes to be qualified for admission to an MS program. So in order to get into a Masters program in Applied Mathematics, I had to take a few more upper division college classes to show that I was capable of handling it. In total, I took four math classes and a computer programming class (all while working full-time). If you can afford it, I would recommend taking the courses at local universities. Luckily, I could get some tuition reimbursement from my job, so that covered some of the cost, and I lived in the Boston area where there are a lot of universities with afternoon/evening classes. If possible, you could also arrange "flex hours" with your job if you must take classes that meet during the morning or early afternoon. Re: rec letters. It shouldn't be a problem securing a strong letter of recommendation from professors who only taught you for one semester. All of my rec letters for my Masters program were from professors whom I had taken only one semester of mathematics with. For PhD programs, it might be advantageous to get one or two rec letters from professors you've known for at least two semesters, but for MS programs, it's not as big of a deal. Good luck! As someone who made a similar career change, I say it's definitely worth it.

No problem. The "typical" day does tend to vary from day to day (some days of mine were spent preparing manuscripts, other days only doing simulations/data analysis or working out new theorems and proofs, and many more days spent just reading/rereading papers). I do have a set "routine" though to spend a certain amount of time per week on research activities, but the actual day-to-day tends to vary.

Typically, the first two years will be spent taking courses and preparing for qualifying exams. At some elite programs, students will begin reading statistical literature in their first or second year in order to identify potential research areas of interest and potential advisors. Starting in your third year, you'll finish up remaining coursework (maybe 1-2 classes a semester), but the research phase of your program is what consumes much of your time. At this point, your time will be a lot less structured. You do need to set your own schedule, and it requires a bit of self-discipline (you can work as much as or as little as you want -- nobody is going to "force" you to do anything, so if you took a couple of weeks off and didn't accomplish much, probably nobody would notice). I started off the first few months of research just reading papers and a book and familiarizing myself with work that had previously been done and trying to find open problems. Then once you identify problems that show promise, you try to work on these open problems. This involves a lot of: rereading papers, browsing new ones for inspiration, playing around with formulae and computer simulations, and later on, writing, revising, and preparing manuscripts for submission to journals. In short, it varies from day to day, but in general, expect to spend the first 2 years primarily on courses, and starting in your third year, be prepared to spend at least a couple of hours everyday on research, plus anywhere from 5-15 hours a week on RA or TA (if you aren't supported by a fellowship).

1) The job market for (pure) mathematics is definitely tougher than it is for statistics. It is still possible to secure a tenure-track job in a Statistics department without doing a postdoc (although it's becoming a lot tougher now), but this is virtually unheard of for mathematics. As a reference, even Mathematics Professor Daniel Kane -- who publishes a gazillion papers in both pure math and theoretical computer science -- did a postdoc. 2) I am not sure about the competition for a top mathematics program vs. a top statistics program, but it is still fairly competitive for Stat. In the Statistics program I'm graduating from in August, they accept around 15-20 students a year out of roughly 200 applicants. Other schools have similar statistics (e.g. on the Stanford FAQ page, it indicates that they admit 10-12 students out of 120 applications, and it's probably similar at other top-tier statistics departments). 3) For securing a good postdoc, the PhD advisor matters far more than the pedigree (although the latter is also taken into account). I don't think that any math OR statistics department is going to frown upon someone who had Persi Diaconis (in the Stanford Statistics department) as their advisor -- someone who had Diaconis as their advisor can definitely get a good postdoc in a mathematics department. Also, there certainly is a strong correlation between pedigree and success on the job market, but I suspect it's because there are more choices to pick from in a highly ranked department. But in the end, the most important thing is that your advisor is someone who is active in the field, well-connected, and most importantly, someone whom you can get along with (I think this last point is pretty crucial and something that a lot of students who are "starstruck" don't consider nearly enough). 4) Re: culture. You will find people interested in a variety of topics in any math or stat department. My work is mainly theoretical (as in theorem-proof type stuff). There are other students in my department who are predominantly interested in theory as well. For example, the students who work on Markov chain Monte Carlo theory in my department are doing some very abstract stuff with functional analysis. The same would be true in a mathematics department, where you'd probably have both applied mathematicians and pure mathematicians. I think you should visit the places that you are accepted to and then you can better assess culture and personal fit.

As long as your PhD advisor is a reputable name in the probability theory community and matches your interests, then I don't think it matters so much whether it's in a Math or a Stats Dept. There are some Statistics (as well as Computer Science and Operations Research & Engineering) departments that are very strong in probability theory. If your interests are primarily in probability, then you are almost certainly going to need to do a postdoc, so the choice of advisor is absolutely crucial for securing a good postdoc after you graduate (whereas in Stat/biostat, you can sometimes get away with not doing one if you have at least one article accepted in a top journal as a grad student). Moreover, the top schools in Statistics (like UC Berkeley) have a specific plan of required coursework for students who are interested in probability theory, so you are certain to take the necessary classes needed to conduct research in a stat department (e.g. UC Berkeley Statistics would require you to take the 2 semesters of analysis/measure theory and some other relevant classes from the math dept). The main difference would probably be the coursework requirements. In math departments, you'd be required to take abstract algebra and topology (maybe multiple semesters of these), but not necessarily any core stat classes like theoretical statistics, regression, linear models, categorical data analysis, etc. That's not really a problem from a research perspective -- most of the learning you do in grad school is teaching yourself through reading papers, conducting your own research, etc. But the coursework requirements for your personal knowledge/edification are something to consider.

In mathematical statistics, one of the hottest areas of research currently is high-dimensional data analysis and "big data" (especially the case where p >> n, p = # of covariates, n = sample size). Some top journals in statistics flat-out reject any submissions that deal with regression and/or variable selection unless the p>>n scenario is examined. I would also say that unsupervised learning (particularly clustering and deep learning/neural networks), graphical models, and ensemble methods are hot right now.

While there are not any current machine learning courses offered in the Statistics Department at University of Florida, there is active research being done in the area of machine learning -- namely in the areas of network analysis, feature/variable selection, graphical models, and Bayesian nonparametrics, A lot of it is theoretical machine learning though, so if that's your thing, then UF is a good choice. Additionally, the Statistics Department at UF is hiring a lot of new professors in the coming years who are affiliated with the newly opened Informatics Institute. So I expect there will be even more opportunities for ML at UF in the upcoming years. Just my two cents.

Research experience can be helpful, but a lot of people get into graduate programs in stats or biostats without any research experience. It's just not possible to do a lot of PhD level research in stat/biostat without having taking graduate level courses in statistics, so I don't think it will carry a whole lot of weight beyond demonstrating that you have some idea about research and know what it entails. I think your grades (especially in upper division math courses) and recommendation letters will make the most difference, so make sure that those are stellar.

I think you would be very competitive for Statistics as well... you could certainly crack the top 10 with this profile, assuming you have decent to stellar letters of recommendation.

Do these programs have any famous professors in the field of geometry/topology? A good way to tell is by searching number of journal citations and amount of NSF grant money. If your hope is to become a mathematician, your chances would be maximized by working with a famous advisor who has a successful track record of advising PhD students and placing them in desirable postdocs and TT academic jobs (you should look up these alumni too on Google, LinkedIn, etc. to see where they have ended up). This increases your chances of getting good postdocs, references, etc. For reference: NSF Awards Search: http://www.nsf.gov/awardsearch/advancedSearchResult? Math Genealogy: http://www.genealogy.ams.org/

The Statistics Graduate School rankings (though this list does not separate Statistics and Biostatistics rankings, so you'll need to separate them). I agree that the programs ranked in the top 15ish by USNWR are indeed the best, but below that, I am not sure how precise the rankings really are (for e.g. I think some programs like Yale and Cornell are underrated, and I think perhaps they are penalized for being smaller programs). PS - use the "Multiquote" and "Quote" buttons to reply to previous posts directly.

Assuming cost is not an issue, I would definitely pick Stanford, given its proximity to Silicon Valley. You could get a great internship, and that's a great location to be for start-ups. I could also be wrong, but I think a MS in Statistics would be a bit more flexible (in terms of opening doors to different careers) and marketable than degrees in Business Analytics, but I'm not too familiar with what the curriculum in Analytics programs looks like.

It's not out of the question, but given the lower undergrad GPA AND the fact that you're an international student, you may want to target schools mostly in the range of USNWR rankings of Ohio State and below (to be honest, your best chances may be at places like UMissouri and Virginia Tech, but you might be able to get into places like University of Florida and Ohio State). I don't think your math GRE subject score is a problem; you should not have to retake it.

I think NCSU's proximity to the Research Triangle is why many who go there opt for industry jobs rather than academic ones. According to the most recent PhD placement data I could find, over 40 percent of PhD graduates from Duke's Statistical Science department also went into industry, and Duke is certainly a very reputable program. Additionally, it seems as though Harvard sends the majority of their stats PhD graduates in industry, http://www.stat.harvard.edu/alumni/PhD.html I think it really depends on the culture of the department too. Some departments may be more-or-less indifferent to where their graduates end up, while others are much more "academically-oriented" and do all they can to place their graduates in academic positions (my current program is one that is invested in success in academia). However, WRT USNWR rankings: based on their methodology (http://www.usnews.com/education/best-graduate-schools/articles/science-schools-methodology), they seem to measure mostly perceived reputation, which tends to be very slow to change. While I do not think these rankings should be discounted completely, I do think that prospective students should also do their own research and look at things like placement of PhD graduates (some of the "lower ranked" schools place very well for academia), NSF grants/awards conferred, and research productivity/number of citations by current faculty (to get a sense of the actual current impact of faculty's research). It is also best to look at the rankings as clusters (so IMO, Stanford-Harvard would be the top cluster, Washington-Wisconsin would be the second one, etc.).

Since your application is for Masters programs and you attend a fairly prestigious university, I think you should have no problem being admitted to some Statistics MS program (though a few MS programs can be quite competitive, e.g. Duke, Yale, Stanford have Masters programs are actually competitive). The overall GPA is kind of low but not unacceptably terrible, and if your GPA in your last two years and your GPA in your mathematics courses are solid and you can get good letters of recommendation, then those can mitigate the overall GPA. However, you should definitely apply to more schools than just the big-name ones you listed. Those might be reaches.

IMO, Statistics doesn't have the same pedigree issues that a lot of other academic disciplines do (although certainly it can't be denied that someone with a doctorate from Stanford, Berkeley, Chicago, Wharton, Harvard, etc. might have an edge when applying to jobs within these same top-tier schools). A well-known advisor, publications in good journals, and a prestigious post-doc can definitely go a long way in getting a job at an R1 institution, if not at one of the aforementioned top-tier schools.

I did my Bachelors in social sciences and did not take any math past linear algebra and Calculus III. After graduating college, I took several math and programming courses at local colleges while I was working full-time, so I was able to get into an Applied Math MS program. I did well there and now am in a Statistics PhD program.

What are your career goals? That might be helpful in determining whether an MPH or an MS would be more helpful to you. If you have adequate mathematics skills (i.e. you did well in Calculus I-III and linear algebra), then you can probably get into a Statistics MS program even without an extensive math/science background. If you don't have these prerequisites, you would either have to take these before you begin the program, or in some cases, a program might allow you to take the prerequisites before enrolling in the Master's level classes.

Speaking as an American at a heavily international department, I can say that the international students typically have good English, so interacting with cohort is not usually a problem. Social life and dating should not be a problem either, even at a department that is very heavily international students. I hang out mostly with the few Americans, plus some of the international students who are not from China or India (I get along with the students from China and India fine, but do not socialize much with them outside of school). But there are many opportunities to meet people outside of your cohort if you are active about it (e.g. beach volleyball, pickup soccer, online dating, etc.).