wine in coffee cups

Doesn't Berkeley have a steady pipeline of people working on machine learning under Jordan/Wainwright moving on to the big tech companies? I imagine their students are getting the jobs no matter if they're coming from the stat department or the EECS department. I think it's unlikely that industry positions you would reasonably be considering coming out of Berkeley would have a hard CS degree requirement. (If they do...you are looking at the wrong jobs.) As a quick check, I don't see many master's in CS degrees among this crowd. I don't think you will need that extra CS credential to be marketable as a statistician/data scientist as long as you pick up relevant experience during your research.

The five-year program could be a good idea for a couple of reasons. I think one big benefit is you will be delaying applying for PhDs for another year, and so you'll have an extra year's worth of (hopefully very good) grades in advanced classes at the time of applying. The other upside is that you'll have the extra year to impress potential recommenders. You'll absolutely need to have strong recommendations to be competitive. I would direct you to the chain of posts from cyberwulf and biostat_prof starting to get some ideas about how much where you go matters vs. your grades vs. your letters. I think an REU could help strengthen your references so that it's not just three letters saying "he was good at classes". You might investigate where previous math/stat majors (and particularly stat master's students) from your school have ended up for stats PhDs and how you stack up to them to get a rough sense of where you might be competitive, though it's premature for serious comparisons.

This will come off as obvious and patronizing, but seriously, get way more A's. The three programs you specifically identified have very, very high average overall GPAs for incoming students.

SAD light boxes were all the rage when I was an undergrad in rural MN. Get one of those suckers on Amazon and you're set even in the dark frigid north.

It sounds like you have a competitive and interesting background, but I would definitely broaden your list to include less selective departments. Is your MS program considered solid, does it have a track record of good PhD placements? From what I know of the research in the UCLA bioinformatics department, that actually sounds like it might be a good fit for you. Duke and CMU stat, not sure about those for more computational genomics-y things? Maybe, but it's not what comes to mind, perhaps too theoretical for your tastes. Also, I think you'd have a much easier time getting into NCSU than CMU because of its size, even though it's a strong department. Your light math background could be an issue both in getting admitted and surviving PhD-level coursework. My guess is you'd get more leeway from non-statistics departments. The real analysis summer class sounds skimpy, but fortunately for you grad schools probably won't know that. You didn't mention it, but I assume you have linear algebra on your transcript somewhere? That's a pretty firm requirement at most places and I don't know that obviously needing to use it in other classes would count. The Coursera classes aren't going to matter for anything besides personal edification. Utterly baffled by your climate/quality-of-life rule-outs. Seattle is awesome, love it here! Baltimore and Boston but not NYC, Minneapolis, or Philly, what? And I think any proper city is going to offer more for someone in his late 20s/early 30s than a college town like Gainesville, College Station, or Chapel Hill so I'm surprised you seem to be leaning that way. But of course if you already have firm (and reasonably informed) opinions, no point wasting your time and money.

You really don't think there is a big gap in the difficulty of that qual vs. the first years master's exam taken at, say, UW bio/stat? Because I sure as hell wouldn't mind taking the Stony Brook qual instead of the UW's!

Wow: only ONE person in the history of TheGradCafe results has been rejected from Stony Brook's AMS program, not just the statistics track. Many (most?) students admitted without funding, based on the comments. Here is this year's common qualifying exam students in AMS take after their first semester: part A (2 hours) part B (1 hour). Here is this year's statistics track qualifying exam taken in the middle of the second year: math stat (2 hours) applied (24 hours). Suffice it to say, some red flags about the quality of the students and the training. I had no idea!

Yeah, this. Forget whether SB stats is "fraudulent" or not. Why on earth would he spend 5+ years of his life being trained as a biostatistician when he is specifically interested in banking, marketing, and finance jobs? I really do not think he should enroll in grad school this fall if these are his options, unless he can switch to quantitative finance at SB. (Why didn't he apply for that in the first place?) Certainly an MBA, finance, or operations research would be much more relevant and more likely to lead to the outcomes he wants. WaddleDoos, WTF was your boyfriend thinking when he sent off these applications? Choose your programs to match your goals, people!

I summarized the demos for stat and biostat from three of the above sources a few days ago, plus UMN. Reposting for the benefit of this discussion and adding in the Wisconsin data for 2012-13 (hadn't seen that before, thanks): UMN statistics: acceptance rates of 30/121 (25%) for females, 40/173 (23%) for males, 4/14 (29%) minority (includes Asian-American), 53/257 (21%) international, 17/42 (40%) domestic UMN biostatistics: 35/68 (51%) female, 28/76 (37%) male, 6/12 (50%) minority, 31/95 (33%) international, 32/49 (65%) domestic Duke PhD statistics: 8/75 (11%) female, 8/121 (7%) male, 0/7 (0%) under-represented minority, 8/140 (6%) international, 8/56 (14%) domestic UWashington PhD statistics: 12/156 (8%) female, 23/210 (11%) male, 8/46 (17%) minority, 10/233 (4%) international, 25/133 (19%) domestic UWashington biostatistics: 26/148 (18%) female, 22/113 (19%) male, 7/40 (18%) minority, 22/151 (15%) international, 26/110 (24%) domestic UNC statistics/OR: 0/7 (0%) under-represented minority, 45/386 (17%) international, 25/125 (20%) domestic UNC biostatistics: 11/17 (65%) under-represented minority, 52/166 (31%) international, 59/96 (61%) domestic UWisc PhD statistics: 1/4 (25%) under-represented minority, 49/182 (27%) international, 26/42 (62%) domestic UWisc MS statistics: 0/2 (0%) under-represented minority, 13/196 (7%) international, 10/21 (48%) domestic UWisc statistics, biostatistics option: 0/0 under-represented minority, 4/57 (7%) international, 9/19 (47%) domestic (Unrelated: what is with Wisconsin's MS program being way more selective than the PhD?)

Here are the most recent acceptance rates publicly available for each group where I can calculate it (although there is uncertainty about the intersection of citizenship/sex/race, and most of these combine across master's/PhD): UMN statistics: acceptance rates of 30/121 (25%) for females, 40/173 (23%) for males, 4/14 (29%) minority (includes Asian-American), 53/257 (21%) international, 17/42 (40%) domestic UMN biostatistics: 35/68 (51%) female, 28/76 (37%) male, 6/12 (50%) minority, 31/95 (33%) international, 32/49 (65%) domestic Duke PhD statistics: 8/75 (11%) female, 8/121 (7%) male, 0/7 (0%) under-represented minority, 8/140 (6%) international, 8/56 (14%) domestic UW PhD statistics: 12/156 (8%) female, 23/210 (11%) male, 8/46 (17%) minority, 10/233 (4%) international, 25/133 (19%) domestic UW biostatistics: 26/148 (18%) female, 22/113 (19%) male, 7/40 (18%) minority, 22/151 (15%) international, 26/110 (24%) domestic UNC statistics/OR: 0/7 (0%) under-represented minority, 45/386 (17%) international, 25/125 (20%) domestic UNC biostatistics: 11/17 (65%) under-represented minority, 52/166 (31%) international, 59/96 (61%) domestic Being a US citizen clearly gives you an advantage: acceptance rates are about twice as high for most of these programs. However, it's not at all obvious that being female and/or an under-represented minority gives you a leg up.

Glad you had a good experience and an easier decision

Some armchair theorizing: The 3.8s fundamentally don't understand the 2.8s. They just can't imagine not caring and not trying to do well in all classes. They mostly came from upper middle class backgrounds and took academics seriously their whole lives and went to top schools. The privileges they have don't register. To them, it looks like you had a good opportunity and squandered it, even if you have since made up for it. Unfortunately for you professors sitting on admissions committees are 3.8s. Statistics is a liberal arts discipline. Lots of collaboration with non-statisticians, lots of consulting, lots of reading, lots of writing, lots of presentations, you get the idea. The data doesn't collect itself. You need to be curious about things that aren't statistics, or at least show you are nonetheless willing to learn and can take them seriously. As with all academic fields we work for ourselves, but we spend much of our time serving others to help them perform good science. Poor grades in superficially unrelated classes signal a potential lack of caring about the context in which you will conduct your analyses. (This may be less of a concern in more finance-oriented departments.) My sympathies. Good luck.

Oh, I have truly no idea about that, sorry man. I know the UCSD math department is overall solid, so I would guess the prob/stat group alumni fare pretty well, but you'll have to investigate that yourself if you don't know already. I think the issue is more if you feel that your interests can still be served by the department even if they shift around a lot, which they basically do for everyone. Also, you are sooo young and have blitzed through undergrad in two years!! I would be wary of a program that limits your intellectual options and perspective because you haven't really had the opportunity yet to explore or learn what it feels like to hit a wall/ceiling. My interests are still solidifying, but the general areas I've more or less settled on were picked up at various points over the six years from my senior year of college to the present day. My education before that certainly helped develop my mathematical skills but was not informative as to what I wanted to study (even though at age 20 I would have disagreed and told you that I'd do a pure math PhD right out of college, lololol, because that's all I really knew at the time). I personally needed those extra years to encounter some limitations, get sick of where I was heading, try out new things, read a lot, start caring about areas of application I had no interest in before, etc. And now I'm glad to be in a place that serves my current set of interests well but will continue to work even with some evolution. Statistics is a broad field, so I think if you feel awesome about it now you're not likely to have a complete about-face, but I wouldn't count on more intellectual stability than that. Maybe UCSD can accommodate a wider range of interests than I am uninformedly assuming, maybe not, but I think you should mull on that. I also think you'll learn a little more about the possibilities open to you and which doors you feel okay closing at this early stage in your career by way of comparison with UNC this week.

You have a general interest in statistics and probability now, no particular areas of focus? I would want to make sure you can get exposure to some diverse perspectives to help you figure out what kind of problems you like working on and what your career path could look like. I think that's most easily accomplished at a department that gives you a lot of options: a large faculty across all career stages, lots of researchers currently publishing, frequent seminars and visitors, perhaps interdisciplinary collaborations/funded projects to get inspiration for methods work from applications, things of that nature. I truthfully don't know much about the programs you are considering, but I suspect UNC is the frontrunner in this regard (especially with a good biostatistics department and proximity to NCSU/Duke). I would also want to be around peers who had similar career ambitions as me. If you hope to go on the academic job market, you will really benefit from observing advanced PhD students go through the process themselves, not to mention getting advice from early career faculty. You don't want to be stumbling around without good mentoring, so again, a large department helps out a lot here, particularly one that has good academic placement. Strongly encourage you to wait to decide until after you visit UNC. Nobody can hold that against you. You might save some people on waitlists a little grief and at least turn down UConn now, though.

Okay, so I'm very much not faculty or on the admissions committee, but based on my personal experience, I disagree that improving SoPs is a waste of time. Perhaps cyberwulf's department usually doesn't give these essays a lot of credence for what sound like very sensible reasons. But I will say that at least after getting in, I had doors open for me at several departments based on how I had described my professional background and my research interests in my application materials: RAship offers, extra funding, TAships better aligned with my interests. I know from conversations with some of the faculty in my department around visit days last month that they took notice of students who had memorable descriptions of their motivations and potential areas of interest (particularly for applicants who were no longer in school). Sure, everyone has to have the high GPA and the math classes and the strong letters and such, but once you're on the margin -- and isn't being on a lot of waitlists as marginal as it gets? -- submitting a good and well-written story seems like such an obvious way to tip that next time. Also, as a practical matter, I think writing a strong SoP takes up way less time (and $$$) than trying to ace measure theory and learn a broad enough swath of an undergrad math major curriculum to beat a bunch of test-taking machines in the math GRE percentile game. And all while working full-time! It's one thing to do well in upper level abstract math when you're a student and that's all you do and you're in the study groove, but entirely another to do so when you've been away from classes for a while and have inflexible real life to deal with, where another B+ in upper level math or a 50th percentile subject score might be a real accomplishment but won't actually help matters. Not like all these areas of application improvement are mutually exclusive, but I argue to at least pick that low-hanging SoP fruit. A lot of good opportunities were presented to me that would not have been offered if I had written something more generic. Don't forget that there is synergistic potential with writing a great SoP and sharing that with your recommenders before they write their letters next time, too. They might be able to say more emphatic and specific things about your accomplishments and potential once you give them a clear picture of where you've been and what you hope to do.

Duke sounds like the clear best choice to me, but good luck in making your decision in any case!

Agreed. I would interpret getting waitlisted at half the programs your applied to as evidence that you probably already have the academic credentials to get in to a good department. As such, I think taking more math classes or the math subject GRE is not likely to help much (and could very well hurt if you don't perform as well as you hope). So I don't think you were being unrealistic, but when you reapply, you should obviously cast a wider net because admission rates for non-US-citizens are already incredibly low and only going dropping more next year. I note that you did not mention your letters of recommendation or your statements of purpose! Those seem to me to be the obvious areas of improvement for someone who is academically qualified but not ranked high enough in the pile. You should definitely reflect on both of these areas in the next year and talk to your recommenders. You came from a top 10 stat department for undergrad, so if your letters came from statistics faculty there, admissions readers actually know who those people are and take their opinions very seriously. My guess is admissions committees place more weight on recommendation letters in your situation than would be the case for applicants with less familiar reference names from pure math departments or unknown schools. Out of everything in your application, the main place you have complete control over now is your statement of purpose. No excuse not to get that in tip-top shape next time you apply. Ask a variety of people for feedback on your current materials when you can: the people who wrote your reference letters, others you know in the stat/ML world, and friends who are good writers/editors from any background. Surely there are aspects that could be improved. Combined with strong letters from credible references, a clear and compelling SOP could make all the difference next time.

Really interesting discussion. One thing I wonder is how generalizable those high academic starting salaries in biostatistics actually are to biostatistics PhDs, though? My university's biostatistics recent hires and job candidates seem to mostly have been trained at the top theoretical statistics programs (whether direct from grad school or after a postdoc). My suspicion is that if I were to hunt down the 15 respondents who started in biostatistics departments last fall and 19 from the year before is that maybe half of those actually have biostatistics degrees and that they would come from just a few universities.

The OP didn't mention machine learning specifically so this may be entirely irrelevant to his/her interests, but I would dispute ML/data mining as a reason to choose Yale stat over UW biostat. The UW is spending gobs of money in a well-publicized effort to become one of the top universities for ML. Lots of new faculty have been hired in this area just in the past year (see http://news.cs.washington.edu/2012/09/06/uw-cse-makes-game-changing-hires-in-machine-learning-big-data-computer-vision-and-computer-systems/ for just the ones as of last fall -- definitely more hires coming for the next academic year) and a new concentration is being developed for ML/big data to launch this year. I don't know how the ML/big data initiative has been to advertised to current/prospective biostat students, but is certainly a big new thing for UW stat/CS students and something I think the biostat students would have ready access to. It's a very bold claim to say one would get a better experience in doing anything related to ML at Yale given how serious the UW is about beefing up its ML research and the combined size/reputation/breadth of the stat/biostat/CS departments relative to Yale's equivalents.

The Davis offer is similar to my funding, which was the worst of my offers materially but at a good department with the best personal and research fit so I sucked it up. It looks like you might only be TAing for one semester and funded by fellowship for the other one? Or maybe you only have half-time TA duties? You might want to clarify, that would be pretty nice to be relieved of TA duties for your first term as you adjust to the US. And if you do get the university fellowship (which I think is very unlikely to be the same thing as the out-of-state tuition fellowship you already got -- that is a just a standard part of an offer for a non-state-resident), that will likely waive TA duties for the whole year and/or increase your total stipend above $18K. You also should ask about total amount of fees of any kind that are not covered, particularly health insurance. Health insurance is provided at no cost to graduate students in my program, but you really ought to make sure you don't have to pay a fee for that or that it is small and included in the $800/yr total they mentioned. Also, is this level of funding guaranteed for the duration of your program? I was not exactly a freewheeling Indian bachelor before my program, but I made more in my last year of employment than I will over the entirety of my 5-6 year PhD, so quite the adjustment. You'll do okay, but you won't be saving anything. Living on this budget is not impossible, but I easily spend the entirety of what I make now without living a particularly extravagant lifestyle (would help if I could give up the $80/mth smartphone...). I am really glad to have savings from my former life to smooth things out so that I am not sensitive to the timing of my paychecks or carrying credit card debt from buying one-time expensive purchases like plane tickets, etc. that can more than eat up your entire biweekly pay.

To add to Statistique's anecdote: on a visit last year, one of my fellow prospectives was a waitlisted student. During the visit, he met with the director to plead his case and was eventually told that he would be admitted conditional on someone else declining their offer, so basically he jumped right to the top of the waitlist. Pleased to say there is a happy ending: I see his name on the department website. I would guess your chances of moving off the waitlist are much higher when you do what he or Statistique did, especially when there isn't a defined ranking a priori. Some departments are probably more amenable to letting waitlisted students visit (or set up Skype calls) than others, but it seems like you should do whatever you can to help the admissions chair warm to you and convince them that you are seriously interested. (And secondarily, guilt admitted students who won't attend into declining their spots ASAP!)

I wouldn't wear a pit-stained baggy math t-shirt and white athletic socks, but I definitely wouldn't wear a tie. Button down shirt or nice sweater and non-denim pants for men, somewhere around business casual. IIRC last year they said they would admit about half of the people they interviewed between flyouts and Skype calls to internationals. So I'd say you have a good chance, of course, but far from a sure thing. Last year they had candidates meet just about every professor in the department in 30 minute meetings, typically the whole group meeting with one or two faculty at a time in their offices. The profs do most of the talking and describe the program, their research, life in the Research Triangle, etc., but you will have a lot of time for questions so make a nice long list. You'll also spend some time socializing with current students and will most likely go to dinner with some of them. They're great, pick their brains as much as you can. You're not getting interviewed in a traditional sense, but you will definitely get asked by the director and probably quite a few of the faculty about potential topics/areas you are interested in, why you want a PhD, that sort of thing. (Not that you still don't get asked about these things on other visits even after getting in! It was very similar to the accepted student days I went to elsewhere but for the fact that I wasn't sure if I was admitted yet.) I had the impression they were screening for personality fit within the department and some degree of maturity/seriousness in purpose, so be yourself but be your best self. Good luck, it's more fun than you think!

Are you not aspiring statisticians who need to redirect anxious energy? Scrape database results for stats/biostats applicants from past years. Parse out GPA/GRE/subject/citizenship when provided. Match results on those keys to track outcomes for individuals over the entire application season (maybe with manual cleanup to account for typos/transpositions). Use more complete posted profiles here and on mathematicsgre.com to augment and correct results. Then you have some data to actually estimate the conditional probabilities. Non-random sample and has errors, but better than nothing. You'll learn valuable skills on the way there.

Second learning R, being more familiar with it would have saved me a lot of time. I also really wish I had brushed up on calculus since it has been 9 years since I last did integration by parts and polar transforms and such. LaTeX is good to pick up in advance if you are not already comfortable. All the fancier math that is somewhat more recent to me has not been of much use thus far.

No salutations, that's just a waste of words. Launch right into it. This is not a letter and you don't need niceties.