Jump to content

Statistics PhD - 2019 Application Cycle


Recommended Posts

 I will be applying to Statistics PhD programs this fall. I had a few questions with respect to forming a prospective school list, which I will include at the end. Below is some information to give some context while still maintaining anonymity (I hope). 
Undergrad Institution: Top 10 US Private University 
Major(s): Mathematics and Statistics
GPA: ~3.75 cumulative, ~3.88 Math/Stat
Type of Student: Domestic Male

GRE General Test: 166 Q, 164 V (W pending)
GRE Subject Test in Mathematics: 9/15 score pending (I found it tricky lol)
Programs Applying: Statistics PhD
Research Experience: ~2 years. Preparing to submit a paper this fall. Field is one of {computational biology, chemistry, physics, economics}, i.e. applied area with a tradition of quantitative/computational methods (intentionally vague to maintain anonymity)

Awards/Honors/Recognitions: None really
Pertinent Activities or Jobs: Grader for various math courses. Software engineering internships for two summers.
Letters of Recommendation: 1 "very, very strong" letter from research PI (words from the horse's mouth). Planning to ask two other profs from math/stat courses, which should be solid.
Math/Statistics Grades: 
Multivariable Calculus (A)
Intro to Proofs (B+)
Real Analysis I, II, III (all of Rudin; A, A, A-)
Abstract Linear Algebra (A)
Abstract Algebra I, II, III (Groups, Rings, Galois Theory from Dummit and Foote; A, A, A-)
Complex Analysis (A)
Mathematical Statistics I (A)
Intro to Probability Models (essentially stochastic processes; A)
Measure Theory (A)
Point-Set Topology (B+)
Optimization (A)
Modern Inference (A)
Functional Analysis (A)
i) My research interests skew towards theory (maybe Probability Theory, ML Theory, Bayesian Nonparametrics, or Monte Carlo methods, but I'm pretty open). I like doing mathematics and proving theorems. However, the phenomenon I'm interested in are not in the realm of classical mathematics; I'm interested in the things that statisticians study. 
ii) The ultimate goal is a career in academia doing research. I understand it is quite difficult to secure a tenure-track position, and theory work seems to be only done in universities. However, I am trying to keep my expectations realistic, and although I wasn't totally thrilled about my software industry internships, it was honest work with its own interesting, technical challenges. A PhD will also lead to more of the interesting work done industry, so this is an upside if I have to leave academia. Nevertheless,  I am currently aiming at an academic career. 
Schools: Applying to most of the "Top 15" (by US News), although I've added a few "safety" schools (is there even such a thing at the grad level?). For sake of clarity, list is below.
Stanford, Berkeley, Chicago, CMU, Harvard, Washington, Michigan, Duke, UPenn, Columbia, Yale, UNC, Wisconsin, Cornell, UIUC, Northwestern, Rice. 
1) What are a few safety schools to which I have a pretty solid chance of being admitted?  
2) Am I a competitive applicant for the schools listed above?
3) I am aware that the academic job market for Mathematics is probably the worst job market in existence. It is my impression that the situation is better in Statistics, but still not great. It seems like at least one postdoc is needed before becoming a tenure-track assistant prof. Is this impression accurate? Do your prospects change with research area? (i.e. theory positions worse than more applied ones?)
Thanks to everyone for taking a look. This forum has been really informative!
Link to comment
Share on other sites

1 & 2) Your profile looks very strong, and I believe that you will get into a top PhD program. However, it may be a good idea to add a few larger programs like NCSU, Penn State, or Purdue. However, I anticipate that you will get accepted to at least one school on your list at the level of CMU or Duke, possibly higher.

3) The academic job market is much better in Statistics than in Mathematics, but still competitive. It is possible to get a tenure-track job without doing a postdoc, but usually, this requires a getting a publication or two in one of the top Statistics journals (Annals of Statistics, JRSS-B, JASA, or Biometrika) as a PhD student. Even if a PhD student manages this feat, a lot of PhD graduates interested in academia still choose to do postdocs, since a productive postdoc can give them more employment options if it goes well and gives them a chance to focus solely on research -- thereby easing the transition from PhD student to Assistant Professor, where you have to spend a lot of time preparing for teaching. My PhD advisor actually graduated a student the year before me who had gotten a TT job offer at University of Arkansas, but this graduate turned down the job in favor of a 2-year postdoc in Statistics at Columbia University.

Doing a postdoc seems to be the norm, however, even for graduates of elite programs like Berkeley, UPenn, and Harvard. The research area does play a role, but not so much w.r.t. theory vs. applied (unless the department looking to hire a new Prof is tilted in one direction or the other). Hiring decisions are much more based on whether or not the research area is relevant and of current interest. My PhD granting institution has been on a hiring spree in recent years (looking to hire three more Assistant Professors this year), and the job ads explicitly states a preference for candidates with expertise in machine learning, network analysis, and data science. Within these areas, one can work on either solely applications, hardcore theory, or a mix of the two. Of the four professors my PhD school hired this past year, two were in applied areas and two were in theory.

Edited by Stat PhD Now Postdoc
Link to comment
Share on other sites

Thanks for your answers to my questions. Wow, I wasn't aware the academic statistics job market was good to the extent that students could afford to forgo a TT job and still find success later on the market. Maybe those students who are able to earn such a job straight out of a PhD would be able to do so whenever they please, but their confidence certainly indicates that the job market is healthy. In any case, it seems that high impact work is really the key for success, rather than pedigree, which is reassuring. 

It is also interesting to hear that the theory/applied distinction is not really relevant to hiring decisions. It seems to me that theoretical work is just done on a much longer timescale than applied work. For example, applied ML results are iterated at breathtaking speeds due to the conference format, whereas the reviewing processes in journals is longer. How exactly does a department choose to hire an ML theorist over a ML practitioner specializing in a certain application, or vice versa? 

Additionally, is there a real pressure to secure grants in the statistics world? Clearly grants are the lifeblood of the experimental sciences and a very large part of a prof's time/efforts is to write and submit grant proposals, dragging the prof away from actually doing science. I have an impression that such pressure doesn't really exist in mathematics. Does the same hold true in statistics?

Thanks again for answering my questions. The "business model of academia" is certainly a subject in its own right and it's probably too early for me to concern myself about such things, but it's kind of cool to think about how all of the cogs fit together.




Link to comment
Share on other sites

1) Indeed, the most important thing for getting a job in academic statistics is to publish in good journals and conferences. In statistics departments, articles appearing in top journals (JASA, Biometrika, Biometrics, JRSS-B, AoS) and top-tier conferences (NIPS, AISTATS, etc.) carry a huge amount of weight both towards getting hired and your tenure case. Pedigree is helpful, but the publication record matters more. 

2) Hiring decisions in math, CS, and statistics will often be made based on filling a particular "niche" that the department is looking to expand or that it currently lacks. My PhD department did not have any faculty specializing in Spatial/Environmental Statistics, so they hired a new Assistant Prof whose research was on that this past year (they even put out a job ad saying they were looking to hire a new Assistant Professor in the area of spatial statistics). Right now, they are looking to expand the number of faculty in Data Science and Statistical Network models. As to whether to hire a theoretician or a more applied statistician, it may also come down to whether the hiring Department thinks you will fit into their "culture." Some departments are very theoretical and plan to stay that way. My Department has traditionally been rather theoretical, but right now, it seems to want to branch out a bit. 

In any case, there is a job search committee, and the hiring decisions are made collectively by the committee (so some faculty may prefer the theoreticians, but if enough faculty prefer the more applied statistician, then they will extend the offer to the applied statistician). Typically, there is a "shortlist" of 5 candidates, with the top 3 invited to campus to give job talks as departmental seminars. Then the top 3 are ranked, and if the top choice declines, then they extend the offer to the second choice, the the third if necessary. If all three decline, then they may choose to invite choices #4 to campus, and #5 if necessary. If *no one* accepts the job offer, then it is declared as a failed search.

3) Securing external grants certainly helps and is counted towards your tenure. Additionally, in Math and Statistics, the salaries are for 9-month contracts. If you want to get paid in the summer, you need a grant. Plus, if you get grant money to pay part of your salary, you can "buy out" of teaching for one semester or two, or use it to fund more of your students and your conference trips. So most Statistics professors will apply for grants. But I would say the pressure to bring in grant money is less than in other STEM areas.

Edited by Stat PhD Now Postdoc
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Create New...

Important Information

By using this site, you agree to our Terms of Use and Privacy Policy.