Stat Assistant Professor

My sense is that the attrition is definitely lower at top programs. But top programs are more likely to accept more advanced students (i.e. those who already have taken Casella & Berger Masters level statistics, measure theory and tons of other advanced math, etc.), so they will be able to hit the ground running on research faster. Top programs also do more to ensure that their students can focus on their research and complete their degrees in a timely manner (for example, giving the students more reasonable teaching loads and higher stipends). The pros of Master'ing out of a PhD program are: lower opportunity costs (4-6 years is a long time to be out of the "regular" job market), you can make a decent amount of money right away, and if you really aren't that interested in an academic career or a career where a PhD is needed/preferred, then you will have less sunk cost. The con is that if you do that, certain careers (e.g. academia, leading your own R&D group in big pharma, etc.) will probably be closed off to you and your pay ceiling may not be as high as if you had a PhD. For example, Distinguished Professors often make more than their industry counterparts. There are also more cons if you're not an American citizen/permanent resident (it's much easier for foreign nationals to get an H1-B work visa if they have a PhD, particularly in STEM).

Your math background is already more than sufficient to get into just about any Masters program in Statistics. In fact, many Stat/Biostat MS students have only taken the bare minimum of Calculus I-III and Linear Algebra. You don't need measure theory or advanced probability theory to be admitted to any MS program in Stats, nor are these required courses for MS students (unless you feel like taking them just for the hell of it). I would just apply to programs that interest you and review Calculus (e.g. how to do integration by parts, change of variables, that kind of thing) before enrolling. Then you should be good to go!

I would also consider UM to be a bit better than NCSU in terms of overall quality, but NCSU is also a very strong program, and I would not consider it outrageous for someone to strongly prefer NCSU over UM. NCSU also has famous faculty and outstanding researchers who can help their students get good postdocs/TT jobs. The PhD/postdoc supervisor(s), the quality of your letters of recommendation, and your publication record matter above all else if you're interested in academic jobs. I'm from a program ranked in the 40s in USNWR, and one of my PhD advisor's former students got job offers from Johns Hopkins Biostatistics and Duke University when they were on the market (they also got interviews at Cornell and UC Berkeley). I also got multiple interviews this year, and so did one of my PhD classmates when he was on the market last year (he is now at University of Minnesota), because our job application packets were competitive at the time we applied.

The Econometrics/Statistics Chicago Booth is a solid program with some great, renowned statisticians, including ones who work in the area of high-dimensional statistics and ML. For example, Veronika Rockova and Nicholas Polson have worked a lot recently on deep learning and regression tree methods. They also publish in top machine learning conferences like NeurIPS, not just statistics journals. I think if you were to attend Chicago Booth and worked with one of their statistics faculty, you would be in excellent shape. I would rate University of Michigan Statistics as slightly better, but UNC-STOR is still a great program. It seemed like in the past, UNC-STOR focused very heavily on probability theory, but they have been expanding their statistics group a lot in recent years and have hired some great professors in the areas of high-dimensional statistics and machine learning. There are also some Biostatistics faculty at UNC-CH who hold joint appointments in Statistics and who would be excellent supervisors for Statistics PhD students (e.g. Dr. Michael Kosorok has placed UNC Stat PhD graduates in top faculty positions and top postdocs like UPenn Wharton). If you're comparing between UMich and UNC-STOR, I would focus on more than just rankings (e.g. job placements, location, qualifying exams, etc.), as the difference in rankings/program quality is not that material, IMO.

I would ask the graduate coordinators. But my hunch is that you most likely won't be able to start in the spring, since most first-year graduate courses in Statistics follow two-semester sequences (e.g. Statistics Theory I in the fall and then Statistics Theory II in the spring), or consecutive-quarter sequences if the school is on a quarter system. If it is a serious issue, I'm sure the graduate director can arrange for you and other international students to do a portion of the coursework online/remotely. During the epidemic, many universities have moved to completely online instruction for both undergrad and grad courses.

I think University of Michigan is quite well-regarded, and their graduates seem to do well in the job market. In the 2019-2020 academic job market cycle for Statistics, there were two folks from Michigan (one fifth year PhD student and one who got their PhD there and is now a postdoc at UC Berkeley) who were getting a ton of campus interviews, including interviews at Ivy League schools, Carnegie Mellon, University of Washington, etc.

If you can obtain an external grant that would partially pay for your PhD studies, I think that would definitely boost your PhD application. I think you should have a good chance at some great programs. I would be surprised if you were not admitted to at least one of Minnesota or NCSU. You may have a chance at Duke or UW too.

As far as academic placements go, I would consider Minnesota and NCSU to have stronger placements than Penn State. There are several UMN alumni who are faculty at my PhD alma mater, and they hired one more faculty this past year who also came from UMN. I also know of several alum from UMN at other solid places like UIUC, FSU, etc. And at the place where I am a postdoc (an Ivy League school), there is an Assistant Professor who has a PhD from NCSU. I also know of several PhD alumni from NCSU at places like UC-Irvine and Rice University. EDIT: I am not aware of any publicly available list of placements for UMN or NCSU, but I am aware that several prominent professors (e.g. R. Dennis Cook, Hui Zou, Brian Reich, Subhashis Ghosal) place very well for academia. You can check out their CV's/websites to see where their former students have gone.

If the NCSU graduate coordinator has not already done so, you can ask for a list of job placements from recent years. They should be able to provide it. Duke's placements are readily available here: https://stat.duke.edu/people/phd-alumni I think NCSU would be fine, personally. I am currently collaborating with a postdoc at University of Chicago who got their PhD from NCSU. Grad students who are interested in academic careers should be focused on publishing papers in statistics journals and machine learning conferences like NeurIpS, ICML, and AISTATS as early as possible (and getting teaching experience as instructor of record if they are specifically interested in pursuing careers at teaching-focused institutions). Most students won't really do much or any research while they're completing coursework (the first 1-2 years of the program), but you should be thinking about it early if you want to pursue academia. I think a nice thing about Duke is that they require the first-year students to take a Readings in Statistical Science seminar each semester of their first year, so they can become acquainted with the most recent statistics literature earlier. But even without this, there are many other things grad students can do to get a "head start" (e.g. attend seminars, do independent study, get involved early with interdisciplinary research, etc.).

OP: Another reason why getting the Masters might be necessary is to compensate for the B earned in Theoretical Stats II and the C in Linear Models. If you can get A's in graduate stats theory and linear models, you can highlight in your PhD application that while you didn't do as well the first time, you aced the classes the second time you took them. You may need to explain this discrepancy in your statement of purpose (but without making excuses or blaming the teacher -- how you frame it is very important). Agreed with the above that you may want to look primarily at lower ranked programs -- but I would actually widen the range from 40-70 even more and consider the entire USNWR list.

Mike Daniels still supervises both PhD students and postdocs. I think he will stay at UF for awhile -- he had left for UT Austin previously, but UF gave him a lot of money (>$200k salary) to come back and be the Department Chair. He has placed some students in great academic positions -- he has former students at UT Austin SDS, University of Louisville Biostatistics, and Boston University Biostatistics. He is also more applied, so if you are not that interested in theoretical stats, he is a good choice to work with (UF has also hired other more applied statisticians in recent years). I think most of the Assistant Professors would probably be good advisors too, as they are well-connected (one of them was co-supervised by two big names at Harvard Biostats -- Ivy League pedigree helps in that regard!), and they are pretty productive. It is always a bit riskier to work with an Assistant Prof since they're not tenured, but I don't think the ones they hired from Harvard or Penn will have any difficulty getting tenure, TBH.

I attended University of Florida Statistics. Feel free to PM me if you have any specific questions about the program or Gainesville. Are you interested in academia or industry? If the latter, I don't think it will make much difference, and you should weigh things like location, size of stipend, etc. If you are interested in academia, I'm also not sure that one would give you an inherent advantage over the other -- hiring would be determined a lot by your PhD advisor, the letters of recommendation, and your CV (mostly how strong your publication record is). I would caution that the USNWR rankings do not necessarily reflect academic placements. For example, UFlorida and Michigan State have better academic placements than some top 20 schools like Penn State. In particular, there are UF alum working in TT positions at places like Duke, University of Minnesota, and UT-Austin. This might partly be self-selection and not necessarily causation (i.e. maybe PSU students are more inclined to join industry). But it also shows that below the top 15 or so schools, the rankings are not necessarily correlated with academic placement (i.e. higher rank =/= better placements).

I would ask the Graduate Coordinator of McGill Biostatistics for a list of their most recent placements. I think any of the top 5 universities in Canada (University of Toronto, UBC, McGill, McMaster, and University of Montreal) are quite solid, and schools like UBC and Toronto should be comparable to programs in the U.S. in the rough tier as NCSU-Minnesota (i.e. right below Michigan and Duke). I am not sure about McGill though. But I have seen several faculty with PhDs from these five schools do very well in academia (e.g. Yves Atchades who was at University of Michigan and is now at BU got his PhD from University of Montreal). Academic hiring in Statistics/Biostatistics is also not just about where you did your PhD though, but involves a lot of moving parts (e.g. what your publication record looks like, who your advisors are, what your research is on -- sometimes a department may just not be into your research or is prioritizing hiring someone in another specific sub-area, etc.). Do good work where you feel you will be most productive and in a good position to land good postdocs, and you should be fine.

I would probably not count on being admitted to an Ivy League school, Stanford, UC Berkeley, or UChicago, but the thing that would stand out in your application from others is your first-author publications. This is exceptionally rare, especially for a domestic applicant. A few B's are not a big deal as long as you have consistently gotten A's in classes that are most relevant (e.g. mathematical statistics, linear algebra, and real analysis). The admissions process can be quite noisy at the top schools, so I would recommend applying to two or three "reach" schools like Carnegie Mellon or University of Washington, maybe UC Berkeley, and then applying a bit more broadly among the top 50 programs ranked for Statistics in USNWR. I think you have a good chance at a school like Minnesota or NCSU and you might get lucky with a top school as well. Make sure your letters of recommendation all highlight your research experience and your strong performance in most math/stat courses.

Which large state school did you attend? Your chances at top schools could be very different depending on whether you went to University of Virginia vs. Oklahoma State. If you could let us know roughly where your university ranks (in USNWR), that would give a better gauge. That said, you do have some impressive research experience -- two first-author publications in statistics (even in "average" journals) is not very typical for most applicants to Stat PhD programs, not even at elite PhD programs. You also performed pretty well in both your ugrad and grad. So regardless of what your undergrad is, I wouldn't be surprised if you got into some top 20 PhD programs such as Minnesota or NCSU.

Letters from famous/distinguished professors usually do carry a lot of weight. Working with an Associate Professor is usually okay as well -- a lot of job candidates who get academic positions were supervised by Associate Professors. One of my postdoc PI's is an Associate Professor, and one of his PhD students got interviews at all the top schools in Biostats this spring (Harvard, Johns Hopkins, Michigan, Columbia, you name it) -- and got a job offer from Johns Hopkins. If you want to work with an Assistant Professor and are interested in academia, I might suggest being co-advised by an Associate Professor or a Full Professor, as they are more likely to be well-connected. Also, I wanted to add: while on the job market this past spring, I saw that there were PhD students and graduates from schools like Rice University and UC Davis getting campus interviews at Cornell, UPenn Wharton, and the like. So even at Ivy League schools, they're not *only* inviting Stanford, Berkeley, Harvard grads to interview for TT positions. It is certainly the case that a higher percentage of Stanford PhD's (and PhD's from schools of similar tier) will be viable academic job market candidates than those from lower ranked schools. But a lot of PhD's from lower ranked schools probably aren't interested in academia anyway. And if the grads from the lower ranked schools have strong CV's and strong letters, then they can certainly compete with the best of them.

I'm not so sure about VT, but MSU seems to have some pretty good academic placements. I've seen/met some of their PhD alumni working at decent Math & Statistics departments like Arizona State University and University of Florida. I would recommend sending an e-mail to the Graduate Coordinators of these schools asking about job placements. Any department that doesn't readily provide such information should be a red flag, IMO.

As noted above by other posters, including a Biostatistics faculty member, it will also depend on how familiar the admissions committee members are with the rigor of the institution, not just its ranking (though institutional ranking/prestige often IS a good proxy for assessing that). For example, Reed College will probably be known for its grade deflation and academic rigor. The other schools you have listed, like Oberlin, Bryn Mawr, Mt. Holyoke, etc. are also very well-regarded nationally. That is also why it is exceptionally difficult for international applicants who aren't from the top universities in their home countries to be admitted to PhD programs in Statistics in the USA. Accepting a student from ISI, Peking, Tsinghua, or SNU is usually a safe bet (i.e. in high likelihood, the student will be able to handle the coursework and finish the PhD within a reasonable amount of time), whereas it is a bigger risk to go with a student from a more obscure college. Actually, even at my mid-tier PhD program (ranked 40 in the most recent USNWR rankings), the grad coordinator said that they automatically rejected any applications from international applicants if they had never heard of the university -- because they were already getting a lot of qualified applicants from the top universities in China, India, South Korea, etc.

Many of the schools you mentioned are in the same league as Amherst, Pomona, Williams (e.g. Swarthmore, Weslyan, Middlebury, Bowdoin) or are otherwise very top-tier and nationally well-regarded. I can't imagine that a math major with a 3.9 GPA from Smith or Vassar (say) would struggle in PhD admissions for Statistics or Biostatistics -- and indeed, I have seen PhD students/grads from these SLAC's at top-tier Stat programs like UC Berkeley. If you're talking about a very obscure liberal arts college that isn't well-known outside of the region, on the other hand, then these students will indeed face an uphill battle in PhD admissions (to the very top PhD programs, anyway -- they could still be competitive for mid-tier programs like UFlorida or Rutgers, maybe even fairly good schools like Minnesota, NCSU, or TAMU). Said students actually need to have very, very high GPAs and test scores and have taken a lot of mathematics just to compete with the applicants from UChicago, MIT, or Ivy school who might have "only" a 3.6-3.7.

To the best of my knowledge, my PhD program still requires a year-long sequence in measure theoretic probability and a course on advanced inference based on Lehman & Casella. It's just that the second semester of the Casella & Berger Mathematical Statistics class now deviates a bit from the textbook, since the instructor doesn't want to emphasize some of the material in the later chapters. So while things like MLE, asymptotic distribution of the MLE, and likelihood ratio test (LRT) are still covered, other sections of the book are abridged or skipped entirely. There is a Statistical Computing class as well.

Even if you completed your studies in the U.S., you're not an American citizen or PR, so your application would be compared to that of other international applicants. Pedigree matters a lot in PhD admissions -- especially for international applicants, in the sense that the less prestigious your school is, the more "perfect" your grades need to be and the more you need to demonstrate mathematical ability. Finally, for the top programs in Statistics and the Ivies, those with heavier math backgrounds will typically be favored. You can find a few exceptions for sure, but in general, those with more math will be viewed as stronger applicants who are more likely to succeed in the program and pass qualifying exams. If you are insistent about wanting to get a PhD in Statistics, I would recommend first enrolling in a Mathematics or Statistics Masters program and taking Real Analysis I-II in your first year to demonstrate facility with mathematical proofs. Then I would reapply -- but forget about applying to the Ivies, Stanford, Chicago, Duke, etc., and apply to a wider range of schools. You could also target large state schools like Texas A&M or Purdue, but I would consider these a reach for your profile.

I observed this even in the case for the first-year Masters-level Mathematical Statistics sequence at my PhD program. The first semester, based on chapters 1-5 of Casella & Berger, is more-or-less the same, but the second semester now deviates from Casella & Berger quite a bit. They used to spend a ton of time on things like UMVUE, Neyman-Pearson, and Karlin-Rudin, but now, they either skip it or abridge it considerably, and instead, focus on topics like the EM algorithm, lasso and ridge regression, etc. By now, things like EM algorithm and lasso are not that "new," but they're certainly not relatively archaic like UMVUE or UMP tests, and they will probably be standard tools used for awhile. I think it's a good thing. But then again, when I started to do research, I was basically learning everything on my own (I could go to my advisor for help and questions). So I can't say that most of the classes were really directly useful for research, but it didn't end up mattering in the end anyway.

What are the job placements like for the schools you mentioned? For industry, it probably makes no difference. For academia, having to take these courses may be helpful in that they allow you to sharpen your proof skills, and you pick up on certain techniques from them that you can use repeatedly in your research (splitting the expectation ftw). But if you read enough papers carefully, you can probably also pick up on "standard" proof techniques. For academic hiring at research universities, it's most important that your *research* is prolific and at least some of it is cutting-edge (i.e. getting published in the top journals or top machine learning conferences), not the content or grades of your coursework. Anyway, my two cents: Lehman and Casella is a very classical text but a lot of the material in it may not be very relevant to most modern statistics research (for example, L&C gives a *very* rigorous treatment of UMP tests, admissible estimators/tests, etc., which isn't a popular research topic now). I guess it's nice in that L&C has a lot of material on things like James-Stein estimation that was one of the earliest shrinkage methods (before lasso and all the sparse regression methods). But is it really necessary to know the risk/minimaxity properties of these kinds of estimators in great detail? I'm not sure. As for probability theory, I definitely think it's good to be able to understand notation for the Lebesgue integral and know basic inequalities (e.g. union bound), but if you're a statistician and not a probabilitist, you may be able to get away with only the basics. I believe that at UC Berkeley, PhD students in Statistics do not even need to take measure-theoretic probability (they can instead take only the Applied Statistics and Theoretical Statistics sequence), and their PhD graduates seem to get along just fine.

This is field-dependent. For things like finance, investment banking, and management consulting, I can attest that there is a clear preference for graduates from Ivy League, Stanford, MIT, etc. for many entry-level jobs (I attended an Ivy for undergrad and about half of my graduating class went into finance or consulting). But in Biostatistics, UPitt would be considered quite good, and there is much less obsession with "elite" names in entry-level hiring anyway. In Biostat, the strongest programs are also not uniformly at the Ivies -- apart from Harvard and JHU, the top schools include a bunch of solid state schools like UW, UNC, Michigan.

University of Pittsburgh Biostatistics is a solid, highly respected program and should provide ample job opportunities posts-graduation. If it will cost much less to attend, I would go there, no question.