alicealice

Be sure that your letter writers are able to assess your suitability for graduate study. Some employers are able to do so (e.g., industry researchers in the sciences, some government advisors, etc.); most simply aren't familiar enough with graduate studies or the particular subject area. Even if you've been out of undergrad for several years, it wouldn't take long for a professor to reacquaint him/her self with your profile and provide a meaningful evaluation. Best of luck!

The one-page rule that is common for professional resumes does not apply to CVs. In your CV you may wish to add short descriptions of each of your research and industry experiences, focussing on the aspects most directly related to your proposed area of research. Don't include non-standard material when it isn't solicited by the program: provide a Quantitative Background document for the program that requires it, but for the rest of the applications use your `statement of purpose' to describe briefly the courses you have taken.

Kent is absolutely right about this. Send the professors a friendly email letting them know that after 5 years away from school you've set your mind to returning to your studies. Meeting with them to discuss your plans and provide them with material for letters of recommendation might even leave you feeling ever more motivated and excited about graduate studies and your subject area.

Gustavo, you appear to have quite a strong profile. I would encourage you to apply to at least a few of the top `theoretically inclined' PhD programs in statistics if you can afford the application fee. Worst case scenario, you waste some time, effort and money in the application process. But who knows---you might also wind up very pleasantly surprised at the results. If you are certain that your interests lie in extreme value theory, then you should consider some top programs with strengths in time series analysis. Chicago and Columbia come to mind. Chicago is very strong in theory and time series, and Columbia recently hired Richard Davis away from Colorado State (he has done extensive work in the field of extreme value theory).

Students who are admitted to MIT CS typically apply to other CS programs also, many of which either recommend or require that students take the GRE subject test in CS or a related field. So it may indeed be true that MIT CS admits have high subject test scores even if MIT itself does not use the scores in their admission process. That said, I have never come across a statement on a PhD admissions page explicitly stating how much weight is given to the GRE subject test. We may speculate all we want about test scores, but don't lose sight of what admissions committees at top programs are looking for: outstanding research potential.

I know a little about U Toronto's program, less about UBC's. From what I know, Radford Neal and Michael Evans (UToronto) are well regarded in the Bayesian community.

Since you come from a machine learning background and are already familiar with some Bayesian methodology, I would encourage you to apply to a few of the "top" Bayesian programs (i.e., the ones in my previous post). Harvard also has strengths in this area. A strong statement of purpose outlining your work in ML and your interests in Bayesian methodology might just get you admitted to a good program. As for locating Bayesian research groups, my advice would be to navigate over to the ISBA website and take a look the list of award recipients, members and listed journals. See who's publishing in the major Bayesian journals and what universities they're affiliated with.

Stanford and Berkeley are not schools that one would typically associate with Bayesian inference. Departments that do have strong Bayesian research groups include CMU, UW and those part of the NC Research Triangle (look up SAMSI).

Applicants coming from a machine learning/data mining background are generally viewed favourably by statistics admissions committees. You say that you wish to study mathematical statistics but that you are looking to get into math programs. Math and statistics are two markedly different fields. I would advise you to browse through the course offerings and program descriptions of some master's programs in both math and stats so as to get a better sense of which subject better fits your interests. Do keep in mind that stats programs will generally allow you to take a couple of grad level math courses as part of your degree. Edit: mathematical statistics is statistics, not math.

skibum seems to have in mind the sort of preparation that admissions committees would like to see from their PhD applicants. Since you wish to do applied policy/social science statistics and apply for Master's programs, I would suggest that you take: (As suggested by skibum) -- Linear algebra or an optimization course, whichever one with introduce you to various matrix decompositions -- Probability theory (one where you will be introduced to forms of convergence and basic results like the Dominated Convergence Theorem) -- Real analysis (at the 3rd or 4th year level, whichever your university offers) Then also -- Regression analysis or another course in applied stats. Econometrics probably introduced you to some basic regression techniques, but you'll need a more indepth understanding of regression to do non-econ statistics. Definitely take a course in regression analysis/linear models and/or applied multivariate statistics. These courses will put you into the right mindframe. If your university offers an undergrad level course that covers maximum likelihood estimation, take it. But I wouldn't worry too much about theoretical statistics beyond basic likelihood theory. You'll have to take a stat theory course as part of your master's degree, and I don't see much of an advantage to studying theory in any depth before grad school. Most universities don't even teach decision theory at the undergrad level, so you won't feel like you're behind your cohort for not having studied it before.

Unless you are enrolled part-time, I believe that students generally complete Toronto's Master's program in one year. Feel free to PM me if you'd like to chat a bit about your options.

If you get a tenure track position straight out of a PhD program then you'll generally have no choice but to continue the research you started in grad school. Postdoc positions give you the opportunity to change your research focus and to further distance yourself from your PhD supervisor's research agenda.

In December you should've received an email from your university informing you that your application has advanced to the final stage of the process and has been sent off to NSERC. Your university should've also mentioned in that email that decisions come out around mid-April.

It's quite possible that studying your favourite subject away from Stanford will produce better work (and a generally more happy you) than studying a somewhat related subject at Stanford. Statistics isn't the sort of field where you have to worry about securing a job post-graduation, especially not if you graduate from a top 10 program. My advice would be to visit both the top 10 school and Stanford if you have the chance, and see which one fits better. Regardless of whether you can visit, keep in mind that your supervisor and the work that you produce are more important than the brand name of the school you graduate from--at least that's the case with stats.