Jump to content

Stats PhD Profile Evaluation + Suggestions for schools to apply to


Recommended Posts

Student Type: Domestic Asian Male

Undergrad + Masters: Top 3 University with (a lot of) Grade Inflation

Major: Mathematics (Masters in Statistics)

GPA: 3.85 (3.93 Masters)

Relevant Courses: 

Math: Linear Algebra, Multivariable Calculus, Ordinary Differential Equations,  Applied Linear Algebra , Complex Analysis, Analysis I , Groups and Rings, Galois Theory, Graduate Analysis I, Analysis on Manifolds, Graduate Probability I, Functional Analysis, Analysis II (measure theory)

Stats: Mathematical Statistics, Stochastic Processes I and II, Regression Models and ANOVA, Time Series Analysis, PhD Level Statistical Learning I and II, Applied Linear Models I, II, and III (PhD Level, taking this year). 

GRE: 169 Q, 166 V, 5 W

Research:

  • 2 years in a research lab working with a lesser known professor on evolutionary genomics.
  • 4 months during the summer working on applied statistics research at a gov. lab.
  • Honors thesis in Math in probability theory
  • Various expository final project papers for classes I've taken 

Letters of Recommendation: 2 letters that are hopefully strong from my thesis advisor and professor who I worked with for 2 years. 1 letter from my undergrad advisor who taught an upper division undergrad course that I've taken.

Considering that my university has a lot of grade inflation, I was a bit concerned in the GPA department. I was also a little shaky about the research experience that I've had; I don't know how much is "enough", especially since I've had no real publications that have come out of it! I wanted to get some suggestions on which programs it would be feasible to apply to with my background. Thanks!

 

 

Link to post
Share on other sites

Your profile looks very strong. If you attended a top 3 university, I don't think it matters whether there is "grade inflation" or not. You have also taken a lot of graduate-level courses in both math and stat, and grades in grad school tend to be inflated anyway. Your research experience in evolutionary genetics is also a plus.

I think you should apply to mainly top 15 stats programs (according to the USNWR rankings). I'm sure you will get into several of them.

Edited by Stat Assistant Professor
Link to post
Share on other sites
On 8/4/2020 at 3:25 PM, Stat Assistant Professor said:

Your profile looks very strong. If you attended a top 3 university, I don't think it matters whether there is "grade inflation" or not. You have also taken a lot of graduate-level courses in both math and stat, and grades in grad school tend to be inflated anyway. Your research experience in evolutionary genetics is also a plus.

I think you should apply to mainly top 15 stats programs (according to the USNWR rankings). I'm sure you will get into several of them.

Thanks for the reply. As I was checking out these schools, I was wondering if you could provide insight into what the distinction (if there is any) between the biostats and stats programs?

Link to post
Share on other sites
1 hour ago, krsk said:

Thanks for the reply. As I was checking out these schools, I was wondering if you could provide insight into what the distinction (if there is any) between the biostats and stats programs?

Statistics departments tend to be a bit more theoretical than the majority of biostatistics departments. The top 5 biostatistics departments will have more theoretical coursework as well as more theoretical research (e.g. UW, Hopkins, Harvard, and UNC Biostats all have faculty who publish regularly in journals like Annals of Statistics and JASA-Theory & Methods and who serve on those journals' editorial boards). But for biostatistics, the further down the rankings you go, the more applied the department will be. Additionally, most statistics departments fund their students through teaching assistantships (though some statistics faculty who have a lot of external funding can support their students as research assistants), whereas biostat departments usually fund their students through RAs and external grants. This is in part because there typically aren't undergraduate biostatistics majors, whereas TA's are needed for undergrad statistics courses. 

It's not always so "cut-and-dried" though. You can do a mostly/entirely applied statistics dissertation in a Stat department, even ones that are known to be more theoretical (e.g. I know a few people from UPenn Wharton who wrote applied stat dissertations with little to no theory). And you might be able to do a more theoretical PhD at a biostat department, especially at a top one (one of my postdoc supervisors did his PhD work at Johns Hopkins Biostatistics on statistical theory for generalized estimation equations).  Your profile would be well-suited for either the top statistics or the top biostatistics programs. 

I suppose it depends also on your research interests. If you really enjoyed working on genomics, you could target programs that have strong faculty in statistical genetics (e.g. University of Michigan Biostatistics or UPenn Wharton Statistics -- at Penn, you could work with Nancy Zhang and Hongze Li). If you really enjoyed your theoretical math classes and wanted to keep doing similar stuff like that in statistics, you could focus primarily on Stat programs, with a few of the top Biostat programs thrown in there.

Link to post
Share on other sites

If you're only considering the top stat/biostat programs, I wouldn't say there is much difference between them (stat vs. biostat) in terms of quality, training, or placements. For example, Harvard Biostatistics and Johns Hopkins Biostatistics have both placed exceptionally well not just for Biostat departments but also for Statistics departments. For example, these guys have PhDs from JHU Biostatistics and they publish a ton of work in top methodological and theoretical statistics journals:

https://sites.stat.washington.edu/people/fanghan/

https://yangning.stat.cornell.edu/

Of course, JHU and Harvard Biostat also have very good placements of their graduates as TT faculty in other Biostatistics departments too. 

Edited by Stat Assistant Professor
Link to post
Share on other sites

If you're interested in real theoretical work, the only two real options for biostats are Washington and UNC, with a larger emphasis on the former. While there are faculty doing a lot of theoretical work around the top-5, the training is not as rigorous as the others.

With your background, I'd be shocked if you didn't get into any of the biostats programs.

Link to post
Share on other sites
52 minutes ago, StatsG0d said:

If you're interested in real theoretical work, the only two real options for biostats are Washington and UNC, with a larger emphasis on the former. While there are faculty doing a lot of theoretical work around the top-5, the training is not as rigorous as the others.

With your background, I'd be shocked if you didn't get into any of the biostats programs.

There are some outstanding theoreticians who graduated from Johns Hopkins Biostat and Harvard Biostat. Some I listed above, but there are others as well. I think it probably depends on who you work with as your PhD advisor at these schools.

Link to post
Share on other sites
21 minutes ago, Stat Assistant Professor said:

There are some outstanding theoreticians who graduated from Johns Hopkins Biostat and Harvard Biostat. Some I listed above, but there are others as well. I think it probably depends on who you work with as your PhD advisor at these schools.

Definitely agree that there are theoreticians at JHU and Harvard, and even Michigan biostat (where a grad just got a job at a top 15 statistics department doing theoretical work).

I also struggle with giving advice to prospective students on what it even means for work to be theoretical and whether this distinction is even useful except at the extremes.  Annals of Statistics papers are theoretical (and if you want to publish there, probably go to a statistics department besides a few professors at top biostat places).  Lower-ranked biostat programs, you can do very extremely applied health research and write a dissertation that probably doesn't have a whole lot of innovative stats stuff in it. But unless you absolutely know you want to do Annals-type research (which is rare), you'll probably find everything you would really want in between at a top 60 stats or top 8-10 biostat program.

I also think that the "theoretical" label is often applied to only a very narrow set of what I would consider theoretical research - specifically things on high dimensional stats, asymptotics, measure theory, etc...  That is theoretical statistics according to the statistics community at large.  But when most people not already in the field say theoretical, I suspect they mean "not just applying an R package to a set of data", so the stuff done at any top biostat or stat department fits.  For instance, Tyler VanderWeele at Harvard is one of the biggest names in causal inference, a professor of epidemiology as well as biostats, and his research is described on wikipedia as "applications of causal inference", but I can't imagine reading this paper and thinking it's not a theoretical paper -- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3166439/ it's about creating a framework to solve a class of statistical problems and how to think through them - if that's not theoretical, I don't want to be a theoretician.

Research is usually divided into applied, methodological, and theoretical, which I think are pretty arbitrary, but I think prospective students should think a lot about what they mean when they want to do "theoretical" research.  Do they want to do something that focuses on conceptual ideas of how to solve statistics problems, rather than just analyzing a data set with a known method?  To me this is a more important distinction than "uses complicated enough math to go in Annals of Statistics."

Link to post
Share on other sites
7 hours ago, bayessays said:

Definitely agree that there are theoreticians at JHU and Harvard, and even Michigan biostat (where a grad just got a job at a top 15 statistics department doing theoretical work).

I also struggle with giving advice to prospective students on what it even means for work to be theoretical and whether this distinction is even useful except at the extremes.  Annals of Statistics papers are theoretical (and if you want to publish there, probably go to a statistics department besides a few professors at top biostat places).  Lower-ranked biostat programs, you can do very extremely applied health research and write a dissertation that probably doesn't have a whole lot of innovative stats stuff in it. But unless you absolutely know you want to do Annals-type research (which is rare), you'll probably find everything you would really want in between at a top 60 stats or top 8-10 biostat program.

I also think that the "theoretical" label is often applied to only a very narrow set of what I would consider theoretical research - specifically things on high dimensional stats, asymptotics, measure theory, etc...  That is theoretical statistics according to the statistics community at large.  But when most people not already in the field say theoretical, I suspect they mean "not just applying an R package to a set of data", so the stuff done at any top biostat or stat department fits.  For instance, Tyler VanderWeele at Harvard is one of the biggest names in causal inference, a professor of epidemiology as well as biostats, and his research is described on wikipedia as "applications of causal inference", but I can't imagine reading this paper and thinking it's not a theoretical paper -- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3166439/ it's about creating a framework to solve a class of statistical problems and how to think through them - if that's not theoretical, I don't want to be a theoretician.

Research is usually divided into applied, methodological, and theoretical, which I think are pretty arbitrary, but I think prospective students should think a lot about what they mean when they want to do "theoretical" research.  Do they want to do something that focuses on conceptual ideas of how to solve statistics problems, rather than just analyzing a data set with a known method?  To me this is a more important distinction than "uses complicated enough math to go in Annals of Statistics."

Could you share the web page for the Michigan grad -> Stat prof? 

Link to post
Share on other sites
7 hours ago, bayessays said:

Definitely agree that there are theoreticians at JHU and Harvard, and even Michigan biostat (where a grad just got a job at a top 15 statistics department doing theoretical work).

I also struggle with giving advice to prospective students on what it even means for work to be theoretical and whether this distinction is even useful except at the extremes.  Annals of Statistics papers are theoretical (and if you want to publish there, probably go to a statistics department besides a few professors at top biostat places).  Lower-ranked biostat programs, you can do very extremely applied health research and write a dissertation that probably doesn't have a whole lot of innovative stats stuff in it. But unless you absolutely know you want to do Annals-type research (which is rare), you'll probably find everything you would really want in between at a top 60 stats or top 8-10 biostat program.

I also think that the "theoretical" label is often applied to only a very narrow set of what I would consider theoretical research - specifically things on high dimensional stats, asymptotics, measure theory, etc...  That is theoretical statistics according to the statistics community at large.  But when most people not already in the field say theoretical, I suspect they mean "not just applying an R package to a set of data", so the stuff done at any top biostat or stat department fits.  For instance, Tyler VanderWeele at Harvard is one of the biggest names in causal inference, a professor of epidemiology as well as biostats, and his research is described on wikipedia as "applications of causal inference", but I can't imagine reading this paper and thinking it's not a theoretical paper -- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3166439/ it's about creating a framework to solve a class of statistical problems and how to think through them - if that's not theoretical, I don't want to be a theoretician.

Research is usually divided into applied, methodological, and theoretical, which I think are pretty arbitrary, but I think prospective students should think a lot about what they mean when they want to do "theoretical" research.  Do they want to do something that focuses on conceptual ideas of how to solve statistics problems, rather than just analyzing a data set with a known method?  To me this is a more important distinction than "uses complicated enough math to go in Annals of Statistics."

The line between theoretical/methodological/applied is blurry to be sure. I would consider pure "theoreticians" to be those who mainly publish in places like Annals of Statistics, Bernoulli, Annals of Probability, etc., and I would consider "theoretical" research to be research which is mainly concerned with formalizing statistical analyses/procedures through mathematical theorems and proofs.

If pressed, I think most statisticians in the U.S. would consider themselves methodologists, but they may tilt a bit more towards the theory side or towards the motivating applications.  

Link to post
Share on other sites
16 hours ago, Stat Assistant Professor said:

There are some outstanding theoreticians who graduated from Johns Hopkins Biostat and Harvard Biostat. Some I listed above, but there are others as well. I think it probably depends on who you work with as your PhD advisor at these schools.

Right. I was not claiming that one cannot do theoretical work at other biostats programs around the top 5 nor that these programs do not graduate really good theoretical students, I was simply trying to state that the training is not as mathematically rigorous as UNC or Washington (hence why I used italics above for emphasis).

For example, Harvard requires only Casela-Berger mathstats, while Washington / UNC require at least a little measure theory. 

Edited by StatsG0d
Link to post
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.