Stat Assistant Professor

Well.. PhD admissions committees in Statistics would certainly view more mathematics courses (not necessarily more statistics) as a positive, but they also admit a lot of students with degrees in Statistics. How much funding are we talking about? If you are able to take the Casella-Berger sequence and maybe one additional stat course in the Mathematics MS program to familiarize yourself with statistical concepts, then this is not a bad option. On the other hand, if the Statistics MS programs give you enough flexibility to take math courses, then that is also a good option.

It is good advice for the most part. I think a good goal for most PhD students who are aiming for academia is two first-author papers completed and one project that is still "in the works." If you plan to go into industry, then the publications aren't going to be as crucial, but in that case, I still think at least one first-author project fully completed is a good target (you may not have much say in this matter anyway -- usually, the PhD advisor will stipulate their own requirements to sign off on a student's graduation, and for many, that is going to be two finished projects). I think many PhD students should understand that with a few exceptions (e.g. medical crises, family emergencies, or other extenuating circumstances), the amount of time to completion IS actually very much in their control (that is, completing within 5 years vs. taking longer). I have seen some students take 7-9 years to finish the Statistics PhD, but I don't think they really needed to languish that long if they had been more proactive and done *more* than what their PhD advisor requested of them from the get-go (i.e. read papers on their own, try different methods on their own and then discuss these with their advisor). If all you do is what your PhD advisor asks you to do and if you never ask questions because you're afraid of "looking dumb," then don't be surprised if it takes longer to finish. The PhD advisor is there to GUIDE you and can point you to certain papers to read or certain new things to try, but ultimately, it is YOUR work that you must own and be proactive about.

I don't think so, but you may not want to put it in your application that you have *no* intention of going into academia. The PhD degree is one that trains you to conduct academic research, but most PhD advisors are are understanding if you choose a different path outside the academe.

I think Rutgers is a great department, with lots of great faculty (Zhang, Tan, Strawderman, etc.)! I interviewed with them for a job, and they do indeed have a lot of great faculty working on high-dimensional statistics. Some of the more nascent areas of research these days are high-dimensional generalized additive models, and constructing asymptotically valid confidence intervals for high-dimensional problems (when the 'classical' limit theory fails), and those were some of the projects they would have had me working on if I had accepted their offer. Rutgers strikes me as a more theoretical dept than OSU. If you are certain that you are *very* interested in statistical theory for high-dimensional stats (particularly of the frequentist flavor), then Rutgers is the better option, IMO. I asked Rutgers about their PhD graduates, and it seems as though most of their graduates go into industry in the NJ/NY metro area in data science or big pharma.

I would agree with the UMN program description that if you want to get a PhD in Statistics, your best bet is to take graduate level math classes... possibly Real Analysis II, a measure theory class, an advanced linear algebra class, complex analysis, or a partial differential equations class. As a Masters student, I wouldn't bother taking an upper division PhD-level theoretical stats class (like Advanced Statistical Inference/decision theory, theory for Generalized Linear Models, or large sample theory), since you would just need to take these courses again in the PhD program. And some schools seem to want to teach these to you *their* way -- a lot of it is the same at different schools, but the qualifying exams are likely to cover slightly different material. Showing mathematical maturity and getting great letters of recommendation which speak to your potential to excel as a researcher are the most critical components of your PhD application.

Interestingly, I just looked at the USNWR rankings in mathematics, and it gives separate rankings by specialty (e.g. Analysis, Applied Math, Geometry, etc.), but no such breakdown exists for the Statistics rankings. https://www.usnews.com/best-graduate-schools/top-science-schools/mathematics-rankings I think that would be interesting to see a breakdown by specialty (e.g. methodology, financial mathematics, probability, Bayesian statistics, etc.), as there are some programs that are not ranked as high overall but are very strong in particular specialties (e.g. in financial mathematics, UCSB and Rice Statistics depts are pretty strong).

Very true. But I think this is just one piece of the overall puzzle. I imagine that a PhD graduate from a more regional university that is not well-known may have a slightly harder time getting a job. But most flagship public universities seem to be fine. I looked up the job placement at University of Kentucky out of curiosity, and see some of their alumni working at Intel, Deloitte, etc.

It matters a little bit, but not as much as programming/hacking skills (for industry) and publications/recommendation letters (for academia). For academic jobs, the job talk during your on-campus interview is also very important. There are some exceptions (e.g. some hedge funds may screen out anyone who didn't get a PhD from an Ivy League school), but for the most part, if you can get the job done for industry, then most companies aren't going to pay much attention to where you got your PhD. For academia, the most important thing is to have quality publications and show evidence of independence and productivity (i.e. a number of first-author publications, stand-out recommendation letters, etc.). As long as these are clear in your application packet, then no academic search committee is going to be like, "Oh, this person DIDN'T go to Berkeley/Stanford/Harvard? They're no longer on our shortlist!" For reference, my department hired a new faculty member this year who got a PhD from University of Cincinnati. I haven't heard much about the Statistics program at Cincinnati, but this incoming faculty member did a very prestigious postdoc at a top school and he had first-author papers in top journals. In his job talk, he outlined his seven-year plan which included being co-investigator on some federal grants and culminating in aiming for a NSF CAREER award. This job candidate ultimately got picked over others who may have had more "prestigious" pedigree. Judging from my personal conversations with faculty who have served on search committees and who have been hiring postdocs, the place you got your PhD is of far lower importance (in statistics and biostatistics, anyway) than your publication record and your potential to sustain a productive research career, obtain grant money, etc..

Your case sounds like one of the exceptions. If your work is resulting in publications and you're intimately involved with the whole process, then it definitely helps your application. What you have is quite rare though. A lot of PhD students in statistics enter with Bachelor's degrees in mathematics where they did not do any research of note.

I don't think undergrad research *hurts* per se, but it usually isn't given much attention by adcoms. When reviewing applications, many adcoms are most concerned that you are able to handle/complete the coursework. So recommendations, grades, and breadth of math classes taken tend to receive the most attention. There are exceptions, of course, but the truth of the matter is that most undergrad research in statistics -- and even Masters projects and theses -- bears little resemblance to completing a PhD dissertation. By construction, undergrad and Masters projects/theses need to be very limited in scope because they have to to be completed in a very limited amount of time. But with PhD projects, the time frame for finishing them is much more open-ended (in fact, research is never really "finished" because there are always new problems to work on and new extensions to be made!), and it takes considerably more time to get a publishable result. Sometimes the projects you start working on in your PhD end up being a dead-end and you have to give up and completely start over, but undergrad/MS projects don't tend to be that way. The papers you write for your PhD thesis also need to be of a certain quality.

That's interesting that this is the first time it has appeared in the rankings... not sure why MD Anderson wasn't previously ranked. I think it is definitely top-notch in the area of Bayesian methodology. I applied for a postdoc there and did not get an offer from them (that's okay though, I got other offers). But I think very highly of this place. I think that may be true, but these rankings are mostly accurate even after accounting for size. I tend to think of it more as "tiers" rather than specific rank. Stanford, Berkeley, Harvard, and Chicago are certainly the top schools in statistics, but you could easily interchange the schools in 5-9 (I would personally rank Michigan and UPenn Wharton right below the four aforementioned schools, but a case could be made for the current USNWR rankings too). Once you get to 10-20, you could make the case for interchanging a bunch of the programs in the ranks; similarly for 20-30, etc. etc. Lastly, I feel that it has already been said, but it bears repeating. Rankings are useful, but not the most important thing for getting a job after. For industry, the *most* important thing is to be good at hacking/programming (my current program has PhD grads who are currently working at Amazon, Google, JPMorgan, etc.). For academic jobs, the *most* important thing is to have good publications and excellent recommendation letters. While my current program is not the highest ranked, it also has very strong academic placements.

Those are both solid programs. As long as both have the typical Casella-Berger mathematical statistics sequence and an applied sequence (usually it's on regression/design/categorical data), I would go with the option that is cheaper. For PhD programs in Stat/biostat, research experience is not a big factor in admissions. Letters of recommendation and grades are the most important.

The new rankings look about right to me, though I would personally rank Yale Statistics higher. It's also worth noting that some of the top statistics scholars in the world are not in Statistics depts but in Operations Research and Financial Engineering, Computer Science, and Mathematics departments (e.g. Jianqing Fan of Princeton and Daniel Kane of UCSD come to mind). I don't see that reflected in the rankings.

There are a few other current PhD students who frequent this forum. I've visited it on and off over the years, but I have not seen many posts from current PhD students about their experiences. I thought this may be of interest to potential applicants, so I decided to write about what I have learned (I am about to graduate, finishing my final defense and thesis in May). I am happy to report that my PhD experience was largely positive. 1) A PhD program is fundamentally a research degree, and research is nothing like taking classes. I think some Stat/Biostatistics programs do a great job of involving students in research early on through rotations with different professors or through reading courses to familiarize students with statistical literature. But there are a lot of programs where students do not start research until the end of their second year. And I have seen many PhD students who were very, very bright (acing all their classes, 4.0 GPA, etc.) but who really struggled with transitioning from being a student to becoming a researcher. I definitely think you should work hard in your classes so you can pass your written qualifying exams and so you can developed a solid foundational understanding, but once you get to the research stage of the program, you really do have to teach yourself a whole new area. Moreover, research is about discovering something new and pushing the boundary of your field. There is just no way of knowing if some "open problems" can be solved or not! It's not like solving a problem on a homework set where there is generally one correct solution/approach. If you do a theoretical topic for your dissertation, you need to prove new theorems that have never been established before, not just "show” something that already has a known solution. And even if you start working on a problem, you may get stuck for long periods of time (or need to cut your losses and give up), or you may end up somewhere completely different from where you started. Unlike problem sets and exams, there are no concrete solutions. For example, for the first paper that I wrote, I was stuck on a proof for my main theorem for three whole months. Nothing I tried seemed to work! But my PhD advisor pushed me to keep trying, and eventually I found the technique that worked. Phew! 2) A lot of the learning in grad school happens outside the classroom, and you need to ask questions. This comes from talking with your peers, meeting with your advisor, attending departmental seminars, and reading papers. Here is the thing: when most people start research, they do not yet have the skills to really excel at it. A small number of people are able to excel right from the get-go, but for most people, it takes a bit of adjustment, and that's okay! It is important to reach out for help if you need it. If I didn't understand an author's proof or a new concept that I had never encountered before, I would ask my advisor to help me. I didn't have much experience with high-performance computing or running simulations on multiprocessing systems, so I asked my more experienced classmates to help show me how to navigate it. 3) Everybody thinks about quitting at some point. This is perfectly normal. A PhD can be a very demoralizing, frustrating experience. Plus, things can happen in your personal life that can derail you. It's just part of life. When I felt like quitting, I just took some time off... maybe 2-4 days of not doing any work to recuperate and assess why I was putting myself through the PhD. After some time off (not too much time off), I could reason to myself why I wanted to get a PhD, and I got right back to work. So if this happens to you, accept your feelings, take a breather, and then really question your own motivations for pursuing a PhD. If you can answer this question to yourself, "Why do I want a PhD? Am I willing to 'tough' it out when I'm feeling frustrated?", then you will be able to pick up right where you left off. 4) Just about EVERYBODY gets their papers rejected, even Distinguished Professors and Nobel Prize winners. My PhD advisor has co-authored over 250 papers and is quite smart, and he still has papers rejected. Professors at all levels get their papers rejected, some multiple times before they are finally published. It’s part of the process. It also happened to me for the first paper I ever submitted. Rejection always stings, but I say if it happens, take a deep breath and cool off a bit. Once you’ve acknowledged the disappointment and cooled off, read the referee reports and comments from the Associate Editor very carefully. Peer review is inherently a subjective process, but for the most part, paper referees take their jobs very seriously, and there will be valid concerns and comments for improving your manuscript (even if some might not be the most diplomatic when letting you know the faults they find with it!). It may be that the journal you submitted to just might not be the most appropriate venue for your work. Or there may be more substantive changes that are needed to make your manuscript more acceptable for publication. After my first paper was rejected, I spent a lot of time with my advisor revising it. We eventually re-worked the whole paper (e.g. cutting down the length of the literature review to the most essential points), we proved a new lemma and a new theorem that showed our new estimator’s improvement over previous estimators, and we performed several new simulation studies that showed quite interesting results. We just resubmitted this paper, making appropriate changes suggested by the peer reviewers who had rejected the manuscript, and I have to say my paper was way better than before. The paper was better off in the long-run. 5) The choice of PhD advisor is critical. It's very important that your PhD advisor is someone whom you can have a great working relationship with, whose research is interesting to you personally, and who is actively publishing in respectable journals. I think the last two are more important than anything else, especially for academic jobs. You basically need to have quality papers and excellent recommendation letters if you want to get a good postdoc or faculty position. Some PhD students are hesitant to work with Assistant Professors and are "star-struck" but there's really no point working with a world-renowned professor if their mentorship style and their research does not align with your personal working style/interests. Plus, an Assistant Professor who is actively publishing their work in top journals can still help you develop your career. Some people need a bit more guidance and an advisor who gently “pushes” them, while others can operate fairly independently and do not need to meet their advisor very frequently. The working style of you and your advisor should mesh well if you hope to be productive. 6) The fields of statistics and biostatistics change very rapidly, so it's more important that you do research that "comes from the heart" than try to keep up with a "hot area." I would not recommend researching a topic that is so archaic and obscure that only a tiny number of people in the world are still working on it. But I also think that you should prioritize your personal interests above what's currently "hot." It can be very difficult to predict what will be "hot" years from now. For example, Dirichlet processes were not very popular when the concept was first introduced, but decades later, Bayesian nonparametrics have exploded in the field of machine learning. It used to be that SVMs were very popular and neural networks lost some of their popularity, but currently, it is the opposite. There is an explosion of interest in neural networks/deep learning and not as much in SVM. The fields of statistics and biostatistics are constantly evolving and changing, so trying to "time" your thesis to a "hot area" can be tricky. But most importantly, a PhD is a very time-consuming commitment (at least 2 years of research). So you do not want to be miserable the whole time you are doing it. So make sure to pick a thesis topic that you find interesting. You probably won’t be able to do that yourself at first, but to that end, your advisor will help you hone in on some interesting open problems to work on. Do not do a topic that you have no personal interest in! Sure, some people might be more impressed if you do (what they perceive to be) a more "difficult" topic, but at the end of the day, you're the one who has to live with yourself and your career choices. And if your heart just isn't into it, it will make finishing the PhD much more excruciating. 7) Do not assume that your PhD thesis topic is the only thing you will work on for the rest of your career. To tie in with my previous point, you can always change gears and switch to a “hot” research area after you are done with your PhD. Finishing the PhD is the start of your career and certainly not where you want to peak. A PhD dissertation is usually on a specific, narrow topic or set of topics. Some people are lucky and can milk their research area for the rest of their career, but many people aren't that lucky. Even if you want to go into industry, an employer of PhD graduates is going to expect that you can teach yourself new things (new software, new models, etc.) on the fly, even if you've never seen/used these things before. In fact, it is this creativity and ability to learn new things quickly that makes hiring a PhD graduate more appealing than hiring someone with juts a Masters. Likewise in academia, professors are teaching themselves new things and moving into new areas all the time. My own PhD advisor began his career doing frequentist nonparametric statistics, but now he has research in a variety of areas of Bayesian statistics. The postdocs I am currently considering are in entirely new areas that I haven't learned before. By the end of the PhD, you should ideally have enough maturity and initiative to teach yourself different areas of statistics.

Just one point... for industry jobs, the most important thing for PhD level jobs is to be good at computer hacking/programming. Theoretical or applied thesis topic is mostly irrelevant. There are tons of PhD grads in pure math, theoretical physics, theoretical stats, etc. making bank on Wall Street, in Silicon Valley, etc. I just wanted to chime in, because there might be people who are super interested in theory or pure math, but who have no desire to become faculty,. And the PhD is their best opportunity to really delve deeply into theory before they go off to work in industry.

That is not to say that people should write off Assistant Professors as potential PhD advisors, though. Although they may not be as famous or established yet, it's often the Assistant Profs that are doing the more innovative, "current" research. And success in publishing papers and getting good academic jobs has a lot to do with the topics you are researching -- i.e. whether it's something where not as much work has been done before, vs. doing research in areas that have been saturated to the point that publishing on these topics in top journals is difficult unless it's something *very* different from the current literature. I would say to look at the publication records of faculty members and to go with the PhD advisor whom you get along with, whose research is most interesting to you, and who is actively publishing their work in respectable journals. An Assistant Professor who is very productive and publishes a lot in JASA, AoS, JRSS-B, Biometrika, Biometrics, etc. can be a great choice for advisor.

Reputation of department certainly does carry weight, but I would say reputation of your PhD advisor and the quality of your papers matter far more. My program is solid in rankings but not a top 10 program. However, my advisor is world-renowned (i.e. gets invited to speak at conferences all over the world), and I got a paper accepted in a fairly selective journal (~20 percent acceptance rate) and another paper that won a paper award at JSM. As a result, I managed to get postdoc interviews at really good schools.

That's true. If feasible, I would recommend visiting the programs and learning more about them firsthand. Or barring that, reaching out to current PhD students and asking about their experiences. I think completion rate and job placements are very pertinent information to consider as well (usually job placements data will also list the PhD advisor). Institutional reputation and rankings ARE also important, but not as important as doing good research and getting a good postdoc (if you want to go into academia). So you need to make sure you are at a place that is conducive to getting research done.

I would say to all prospective applicants to ultimately choose the program that you feel is the best fit for you personally. This includes factors like program size (some people thrive better in a smaller dept), location, and "culture" of the department. These are highly personal and it's perfectly fine to prioritize one thing over another (like location, for example). I would also tell prospective applicants to not be overly concerned with rankings. It would be wrong to say that some people won't make a snap judgment based on your pedigree. But if you want to go into industry, it probably isn't that important. If you want to go into academia, on the other hand, the PhD advisor matters the most, and the most recent (postdoc/VAP) appointment and publication record are what receive the most consideration in hiring. Nobody is going to hire someone who got their PhD from a top school, but who has no publications. For reference, my department (a reputable stats dept) has hired two new faculty this year who got their PhDs from University of California-Santa Cruz (UCSC) and University of Cincinnati. These are excellent schools, but their statistics PhD programs are not considered "top tier" (I'm not sure if they are even ranked by USNWR). However, these new incoming Assistant Professors DID do prestigious postdocs and were quite productive during their time there (publishing their postdoc work in top journals), which gave them an edge in the hiring process. In the current job market, it's very difficult to get a tenure-track job without doing a postdoc (it does happen but the vast majority of aspiring academics in stat/biostat need to do a postdoc now).

Some of the reasons I can think of off the top of my head: 1) Bayesian models allow you to incorporate prior information/beliefs about the parameters of interest into your model. If you have little or no information, you can still perform Bayesian inference by choosing appropriately "weakly informative" or "non-informative" priors. 2) In some ways, the Bayesian interpretation of uncertainty quantification is easier and more "natural." For example, when you talk about confidence intervals in the frequentist sense, you're not talking about probabilities, you're talking about "long-term" coverage (i.e. if you construct many intervals, 95% of them will contain the true parameter). Meanwhile, the Bayesian credible interval gives you a 95% probability that the true parameter is contained in said interval. I am not sure about specific applications in social science, but it is sometimes a philosophical difference as to whether it is appropriate to describe your parameters of interest probabilistically (i.e. whether the parameters should be treated as random variables, rather than as fixed values).

You should definitely ask the Graduate Coordinators for this information!

Also,if you're interested in faculty jobs, you can always change your research area when you do a postdoc or when you become a professor. Many people don't stay in the same area as their PhD thesis topic. For example, one of my advisor's former students wrote a thesis on small area estimation, but she abandoned it shortly after and now is an assistant professor working on computational social science and record linkage problems (at a top school, I might add). I myself have had postdoc interviews in fields as diverse as causal inference, electronic health records, pseudolikelihood methods, nonparametrics, etc., and I did not write my thesis on ANY of these topics. By the time you get to the postdoc or faculty stage, most people will assume that you have enough maturity to teach yourself a new field and conduct independent research in new areas. So this just suggests to me that your thesis research should be on something that personally interests you above all else -- since you can always switch to a different area later, if you want.

I think it is worth noting that doing a PhD is about YOU and your personal interests. It is a bit difficult to predict what will or will not be "hot" years from now. For example, Dirichlet processes did not explode in the field of machine learning/computer science until about a decade after they were first introduced. SVM used to be a very hot topic and neural networks lost some of their appeal, but now it's totally the opposite: currently there is a lot of renewed interest in neural networks/deep learning, but not so much in SVM. So trying to go into what's most "fashionable" at the moment can be a tricky game, because the fields of stat and biostat are constantly changing and evolving. I wouldn't recommend doing a PhD on a topic that is extremely archaic to the point that only a tiny number of old profs are still conducting research in it. But I also would prioritize personal interests above anything else. A PhD is a time-consuming commitment, and you don't want to be miserable the whole time you're doing research. So skim a few papers by faculty members and find something that you are truly passionate about!

A number of famous faculty members retired, died, or left to go to other schools (e.g. Casella, Young, Agresti, ...), so it lost some of its "elite"-ness this past decade. In U.S. News and World Report, UF Statistics was ranked #9 in 2008, but has since fallen in the ranks. However, I would not concern myself too much with rankings. Nowadays, a postdoc is practically a requirement for an academic job, unless you have at least one paper in a top journal or went to Stanford or something like that. What's most important is that you do good research, preferably with an advisor who is active in your field and who is well connected in the bigger statistics community.

Yes, and feel free to PM me if you have any questions. As mentioned above, UF is world-renowned in the area of MCMC, but many students do not focus on that (myself included). With the new chair, the dept is definitely diversifying, and there are likely to be lots of opportunities and changes in the future (e.g. there is talk of restructuring the program and the curriculum). I myself got postdoc interviews at some top schools, so if you play your cards right, you can definitely succeed here.