Stat Assistant Professor

I mean... if you are aiming for an academic job at a place like Stanford, Columbia, or UChicago specifically, then having a PhD from a similar institution is probably much more essential. If you are ok with other less prestigious R1's, R2's, or SLAC's, I think you will find academic placement to be more strongly correlated with PhD advisor or postdoc advisor than PhD granting institution. My PhD alma mater hired two new faculty last year with PhDs from UCSC and UCincinatti (however, their postdocs were at Duke which probably helped a lot).

I think it definitely is the case that Stanford/UC Berkeley PhD students are more likely to have a paper or two in a top venue like Annals of Statistics, JASA, or NIPS by the time they graduate than are students from other programs. Thus, they do tend to be better-positioned for the academic job market than most others. This makes a lot of sense to me, at least at Stanford, since they not only recruit very talented students, but they also: a) have a plethora of prestigious, productive faculty to choose from as PhD advisors, and b) they get their students started on research and reading papers right away (I believe that all Stanford first-year PhD students have to take a reading course every quarter). But a lot of future success is also dependent on the PhD advisor and the person themselves (e.g. how driven and self-motivated they are to explore research areas). For example, Michael Jordan of Berkeley doesn't even have a PhD in Statistics/Math, and there are other prominent researchers I can think of who have PhDs from respectable but not "top" programs (like David Dunson who has a PhD from Emory Biostatistics or R. Dennis Cook -- of the famous Cook's distance -- who got his PhD from Kansas State). So PhD grads from lower ranked programs should not fret their chances of success, especially if they can secure a good postdoc.

I don't think a "P" will mean much to the adcoms, because it doesn't give any indication whether you did exceptionally or whether you just did the bare minimum to pass. Also, graduate school grades are somewhat inflated, so do not take this class for a letter grade if you are not confident you can get an "A" in it. If you are very interested in the subject, you could ask the professor to audit the class. In my experience, simply auditing is not a good way to learn though -- to really learn measure theory (and most subjects within math), you need to actually practice and do the problem sets. If you don't think you will have time, then I would just not bother with it and wait until you are in a Statistics PhD program to learn it.

If you think you are more likely to get an A in the multivariate class, then I would take that. PhD programs will teach you the measure theory that you need to know in your measure-theoretic probability theory class, so it isn't strictly necessary to have learned it before enrolling. Plus, the programs ranked in the 11-30 range are more willing to accept applicants from non-stat/non-math backgrounds (e.g. students who majored in economics, engineering, or physics), and I would guess that almost none of those applicants who were accepted have taken measure theory previously.

It isn't necessary to have, but if you think you can get an A in measure theory, then that should definitely help your PhD application. It might actually give you a bit of a "headstart" in the measure-theoretic probability class you'll have to take in a PhD program, because you will already know what measure spaces, Lebesgue measures, counting measures, and product measures are and how to read notation involving the Lebesgue integral. And you will have already seen stuff like Fubini's Theorem or Fatou's lemma. What kind of PhD programs are you targeting? It would probably help you the most for some places like Stanford, University of Chicago, or UPenn Wharton. For other programs, it's probably more important to have straight A's than to have measure theory on your transcript.

Nowadays, stat departments are also becoming more receptive to conference proceedings (probably because there are many on machine learning now and this field moves so quickly). The top conferences for statisticians are NeurIPs, ICML, and AIStats, which seem to lean theoretical. There's methodology in these papers, but you typically can't just state a new method -- you also need to prove some theoretical guarantees. And some papers in these conferences are purely theory (e.g. a new error or tail bound, a new convergence rate, etc.).

UPenn Wharton is an excellent department and a top 5 one in my personal opinion. They have some great faculty like Tony Cai and Dylan Small. Their main strengths are casual inference, nonparametric/semiparametric regression, and large-scale multiple testing. Their cohort sizes are also small (each year has only 4-5 students), so you'll definitely get to know your professors and form some close friendships in the department. Also, even though the Statistics Department is housed in the business school, it is actually a primarily theoretical/mathematical department (some very good faculty in pure probability theory, fwiw), and most of the faculty do not do research on applications to business.

The distinction is very vague, to be sure. I would look at which journals they are publishing in. If a lot of their work is in places like Annals of Statistics, Annals of Probability, or Bernoulli, it is probably mainly theoretical. If in JASA-Theory and Methods, Biometrika, or JRSS-B, it is probably a mix of the theory and methodology. If in JASA-Applications and Case Studies, Annals of Applied Statistics, Biometrics, Journal of Computational and Graphical Statistics, or JRSS-C, it is mostly methodological/applied (these may have one theorem but usually not more than two). Some other journals like Statistica Sinica, Journal of Multivariate Analysis, Journal of Machine Learning Research, Bayesian Analysis contain both heavily applied and heavily theoretical articles (and everything in between), so to gauge this, you will need to read the abstract of the article and scroll through it to see how many theorems there are.

If you are interested in Statistics, it would make more sense to me to go to a Statistics department. But there are PhD's in other subjects like Computer Science, Electrical Engineering, and Mathematics/Applied Math who eventually become faculty in Statistics departments -- some pretty renowned ones as well, like Michael Jordan (PhD in Cognitive Science) and John Lafferty (PhD in pure mathematics, focusing on geometry). Depends mainly on your research area and what journals you can publish in. If you publish in good statistics journals and related journals/conferences like the IEEE pubs or NIPS, the degree field itself is not so important.

It should be the same in Biostatistics as in Statistics -- or any other department where acceptances are made at the departmental level rather than into a PI's lab. When my father was working as a professor, his department aimed for an incoming class of around 8 PhD students, and they usually sent acceptances to around 12 applicants annually and had a short waiting list. But in order to get to the waiting list, at least 5 would have to decline the offer.

UW is particularly strong for social science statistics and analysis of network/spatial data. It also has a few outstanding researchers in the machine learning (both theoretical foundations and methodology/applications).

Sorry to be a buzzkill but I would not count on getting off the waitlist at a reputable PhD program. It happens sometimes, but PhD adcoms always extend offers to more than their target incoming class (so for example, if Stanford is targeting an incoming class of 8-10 students, they will give offers to 12 applicants with the expectation that at least 2 will turn them down). At a bigger program where they expect 50-60% matriculation, it is likely that they have extended offers to 10 or more applicants than their target number.

No, courses and grades do not matter one bit in hiring. In order of importance for faculty hiring for an R1 or R2 institution: 1) publication record, 2) recommendation letters, 3) teaching experience (not always necessary to have, especially not in Biostatistics). For hiring at liberal arts colleges and regional schools, that order will be flipped. I am just saying that JHU and UW Biostatistics resemble "traditional" statistics departments in that they expect their students to have a firm foundation in theoretical probability and mathematical statistics so that they can conduct research in theoretical statistics if they choose to. In contrast, lower ranked Biostatistics programs might not require things like measure theory, because their students are probably going to conduct research exclusively in more applied areas and target publications in journals like Biometrics, Statistics in Medicine, or BMC Bioinformatics, which don't need much (if any) theory to get published.

Have you checked the professors who could act as potential advisors and their job placements? I'm not familiar with any renowned faculty from TAMU in the area of Markov chain/stochastic process theory, but UCLA has Qing Zhou.

It seems like the top tier Biostatistics programs place more emphasis on theory. UW Biostat requires a semester of measure theory and JHU Biostatist requires a full year of measure-theoretic probability. One of my postdoc PI's got his PhD in Biostatistics from Hopkins and did his PhD thesis on large sample theory for boundary problems in multiple hypothesis testing. It also depends on what your research is, though. If you did genomics/statistical genetics, precision medicine, or something like that at UW or JHU, it might not position you as well for a faculty position in a Statistics department (as opposed to Biostat) than if you did something like causal inference or something less specifically related to public health/biology like high-dimensional machine learning.

UW Biostatistics is very strong in both theory and applications. I've seen some faculty like Daniela Witten and Ali Shojaie consistently publishing in top statistics (not just biostat) journals like JASA, JRSS-B, and Biometrika. OSU has some strong faculty; it seems to be more of a Bayesian department and strong in particular areas like spatial statistics.

That said, it probably is harder to go from Biostatistics to a Statistics department if your work was all applied and in stuff like clinical trials, statistical genetics, or medical imaging. These are very important areas of research, no doubt, but Ithese graduates will have a much easier time getting employed by a Biostat department than a Statistics department.

Your publication record is ultimately what matters more than what your degree is in. For example, Veronika Rockova at University of Chicago and David Dunson at Duke University both have PhD's in Biostatistics, but they are now consistently publishing in top journals like JASA, Annals of Statistics, etc. So it doesn't really matter if your degree is in Biostatistics or Statistics as long as your work is good and well-respected. Most Biostatistics departments are happy to hire both Biostatistics and Statistics postdocs or faculty, but at least one paper in a venue like Biometrics, BMC Bioinformatics, or JASA Case Studies and Applications is needed in order to be a competitive job candidate. The very top-tier Biostatistics departments in the country may also want to see publications in Biometrika, JRSS-B, or JASA Theory and Methods (which tend to fall more on the theoretical side). Statistics departments will also hire those with Biostatistics PhDs, but some may only consider job candidates with Biostatistics PhD's if they have published one or two papers in a more theoretical journal like JASA Theory and Methods, Annals of Statistics, Annals of Probability, Biometrika. I think University of Washington Biostatistics PhD students and faculty have a very good track record of publishing in both 'applied' and theoretical venues, so there are a handful of UW Biostatistics PhD's who end up as faculty in Statistics departments. .

It would make sense for you to obtain a Masters in Statistics or Mathematics first in order to alleviate concerns about your mathematical ability and your lighter math background -- at the very least, you need to take real analysis and some upper division/graduate-level math classes to demonstrate that you can do proof-based mathematics. Assuming that you perform well in your Masters program, I imagine you have a very good shot at schools in the range of Ohio State, UIUC, and University of Florida. I think that those would be your best shot, though you might also be able to get into a school like University of Minnesota or Texas A&M. There are graduates from top universities in Latin America who get into top-tier PhD statistics programs in the USA (e.g. Jose Zubizarreta is from Chile and got his PhD at UPenn Wharton and is now a professor at Harvard, and Carlos Carvalho who got his PhD at Duke is from Brazil). So you finishing in the top 3% of your graduating class at the top university in Brazil will certainly work in your favor, and I think you can get into a respectable PhD program in the U.S.

I typically met with him once a week, unless he was traveling (then once every 2-4 weeks). He traveled more during the summer and winter breaks when I was also traveling a bit though, so it's understandable that I met with him less during those times than during the fall and spring semesters. Once a week is pretty standard -- I don't think there is much value in meeting more than that, since less than a week doesn't give the advisee much time to make noticeable progress (even if the progress is as minor as rereading a paper and understanding the contents more clearly). If you are looking to submit some papers to computer science/stat conferences which have hard deadlines for paper submissions, then you might need to meet with the advisor more frequently right before the submission deadline.

UT Austin is a great place to be a Bayesian. TAMU also has a few very strong faculty in Bayesian statistics as well, including Johnson, Mallick, Bhattacharya, and Pati. These professors consistently publish in top journals and have placed PhD graduates in top postdocs like Harvard and Duke. Overall, I'd say that these are a few great schools you have to choose from. I personally love the city of Austin, so I would be biased towards UT Austin. But Rice and TAMU are great schools too.

My PhD advisor didn't have an overwhelming number of students (my department was smaller when I was a student there) but he was a Distinguished Professor who traveled a lot, had a lot of collaborations, and sat on a lot of PhD committees (nearly every PhD committee in our department and the external committee member for a fair number of students in other departments, e.g. CS and Biostatistics). Nevertheless, I still completed my PhD in just under 4 years. If you want to work with a very busy/distinguished professor, my suggestion would be to make sure to be persistent about meeting with them and to make sure that *every* meeting you have with him/her is productive and moving you forward. I personally would type up a "progress report" in Latex before every single meeting I had with him outlining what papers I had read, what questions I had, and what my thoughts and ideas were -- even if I hadn't made any progress on my research topic or even if I had to completely abandon an unpromising research direction, I still wrote up what I had done for that week. We would then go through that every meeting and decide on the next steps. Once I began generating my own ideas more successfully, I would write summaries of them, write out the technical details thoroughly, and type up any theorems and proofs for him to review. When I began writing my own papers, I would email him revisions/new drafts at least four days before our scheduled meeting so he would have time to read the draft and give me feedback in our next meeting. In my experience, this helped a lot. Finally, don't be afraid to chat with the other PhD students of your advisor. Chances are that you will learn just as much or more from talking with them about research than from completing a problem set or trying to make sense of a very difficult paper. Sometimes you have a general, vague idea of what is going on, but you just need that "a-ha!" moment, and chatting with more senior students can give you exactly that to make things click.

Yes, if you are very motivated and driven, you can teach yourself. That's what a PhD program and postdoctoral training are, for the most part... learning how to teach yourself new things. The coursework is useful for gaining a basic foundation, but after you're done with classes, it's all about teaching yourself (with guidance from your PhD advisor, of course).

Any reputable Statistics PhD program is very mathematical, even those ranked in the mid to lower tier. I meant, there isn't much relevance of stuff like abstract algebra, number theory, complex analysis, or topology (subjects that are tested on the Subject GRE) to statistics research. In addition, graduate coursework in mathematics and high grades earned in those courses -- especially at a school like Harvard or Duke -- should be good enough evidence that one has the mathematical ability needed to succeed in a Statistics PhD program. Just my opinion, of course. I of course think very highly of Stanford Statistics, but if I were on an admissions committee, I personally would not put too much stock in the math subject GRE if there is strong evidence of ability from advanced coursework.

If you are interested in a PhD, the reputation of your Masters institution will play a big role. There is no way around it... more reputable schools will be known for being rigorous, whereas adcoms who are not familiar with a more regional program might have more concerns about rigor. Students from regional schools also will be at a disadvantage if it comes to getting a job outside that particular region in comparison to grads from elite private schools or flagship public schools. If money is an issue, is there any reason why you could not just bypass the Masters and apply to PhD programs in a year or two?