bayessays

I would personally disagree about FSU being below Columbia/Emory. They have a lot of strong people, especially if you're interested in image/shape analysis, and have some people doing some more theoretical work than is offered at McGill if that's your thing. McGill has a few people doing some interesting causal inference biostats stuff, but I don't think it could be considered on the same level as FSU *overall*. But if you find a good fit, the most important thing is that you have a good advisor you can publish with and that you're going to the program you feel is right for you. The FSU program is essentially their statistics PhD, so it will be *much* more theoretical (you'll have to take 2 semesters measure theory-based probability classes and pass an exam on some of that material - McGill seems to only cover master's-level material on their exams).

Yikes, this is why I almost never comment on international applicants. Bar just keeps getting insanely high. @cyberwulf, is this mostly because of citizenship-related funding issues in biostatistics departments with government grants, or some desire to keep a balance of international vs domestic students? Since OP completed his undergrad and master's in the US, I thought he would be judged similarly to a domestic applicant with this profile, who I think would have a good-but-not-guaranteed shot at schools in the 4-10 range.

The best thing you can do for your future PhD studies will be to spend all your time doing well in your real classes. Your MS GPA is on the low side, so work on getting that up. Online extension classes are not going to be useful for your application and are not necessary to learn programming. Use the swirl package or download The Art of R Programming to learn some R. SAS and Python will be a waste of time for most people. Machine learning/Data Mining classes will essentially be extensions of your applied statistics classes and will also not be useful -- you can watch youtube videos to learn any of these subjects.

I think you can apply to the top programs, but I think the top 3 will be hard to crack. I agree with @StatsG0d that you should cut some of the lower schools off the list -- although TAMU, Rice, NCSU and Rutgers are very good schools if you mean their statistics departments and I think are good schools to target. I think you should probably get into better programs than Pitt/Emory/BU/Columbia/UT-Anderson/Vanderbilt though, and you should cut these down to maybe only 2 of the above as safer options. I think the biggest zone to be targeting will be Michigan/UNC/Minnesota/UPenn, with a few below and a few above.

You are taking this artificial distinction your school is making between advanced calculus and real analysis too literally. The advanced calculus courses are what everyone here means you need to take. It's kind of weird that it is a two semester course though, as usually all those topics are covered in one semester. You do not have to take the Lebesgue class that your school calls real analysis, but with your profile you will not be competitive without taking the advanced calculus classes. If possible, I would highly recommend taking them next year because otherwise they will not be on your transcript before applications are due.

I don't think I would ever recommend somebody to retake the GRE if they got a 167+.

You already have plenty, but the most helpful thing beyond what you have would be an optimization class probably, and maybe some more statistics classes like linear models even.

I'd love to understand the distinction you're making here. Isn't everybody's research in some sense specialized? If Kosorok's recent papers have "causal" in the title and he's using, for instance, a potential outcomes framework to estimate causal effects in some new precision medicine setup, is that very different from what most people working on "causal inference" will be doing? I know there are some different schools of thought on the causal inference subject, but isn't Kosorok's work pretty much what you'll get at most places for a person researching causal inference? Or are you comparing him to someone like Judea Pearle or Tyler VanderWeele who are I suppose researching more "foundational" causal questions rather than new applications? I feel like there aren't a ton of people doing the latter, but would love to hear your thoughts. Thanks!

I had this situation come up and I think each school handles it differently, though my scores expired earlier, in August, so I didn't look into it that deeply and just retook the test. For instance UChicago's website says the scores need to be valid the day of the application deadline, whereas Penn State says as long as you have sent them a score before it expired, they will accept it after the expiration date. So I'd look at the application/FAQ pages of the departments you're looking at.

Hopkins has Stuart, UNC has Kosorok, Michigan has a few people working on that stuff in both biostat and stats. Washington just hired one of Judea Pearle's students. UPenn has quite a few people working in causal inference and might be the closest thing to a department that has a "focus" on it, although the number of people working on causal inference has grown *a lot* in the past few years. I think you'll find at least one person doing it in most top 10 biostat depts, but you'll have to look through faculty pages to see what's actually interesting to you.

Yes, that should be fine. Note that some schools have specific instructions about whether to send it to the department or just the school in general. But honestly, it should be able to get there no matter what. They're just sending an electronic thing that the school can easily check and I've never had an issue.

You definitely should not apply to masters programs. If your letters are good, you probably don't have to apply out of the top 30 and you have a shot at some top 10s (US News rankings). There are some really good programs throughout the top 40, so I'd look through them to find a good fit and have some safeties. As long as your letters are good, I don't think the analysis grades will hurt you that much. Especially outside the top 10 programs, schools don't get many domestic applicants who've taken grad analysis at top 10 schools, so I can't imagine them holding it against you much.

This was going to be my suggestion as well. Get rid of the long and extra-long reaches entirely. Schools around the area of South Carolina/Mizzou are good targets and will be more open to a non-traditional student. Since it sounds like you like teaching, some of these schools are more open about preparing people to be teachers specifically. As to @DanielWarlock's point, Miratrix got a PhD in statistics and just happens to teach in an education school. Most of these people are doing applied research in education, and this would be similar to getting a PhD in a subject like Quantitative Psychology where you are applying statistics to some field. There is, of course, the separate option of getting a PhD in math/stats education. A lot of people do this and become lecturers at a university. However, the options for people doing *statistics* education research, versus math, are significantly limited.

The bigger point is being missed here. Nobody is talking about grades here or worrying whether @catarctica will do poorly in the grad classes, as @DanielWarlock states. Most graduate classes are jokes and everyone gets As. So nobody will take an A in graduate probability or analysis seriously anyways -- it will raise more questions than it solves because the admissions committee won't believe you have the actual skills to succeed in this class. It is also very likely to not be productive for you personally. If you want to have a serious shot at a statistics PhD program, you need to take some more undergraduate math classes like analysis, and anyone telling you differently is doing you a disservice. OP, if I am wrong and you are a math genius, forgive me, but just be aware that this is not the route 99.9% of people would recommend for you.

@DanielWarlockYou're vastly overestimating the average student's math abilities. Yes, there are some exceptional cases. I'm sure at Harvard there were exceptional cases. But 95%+ of people who complete Waterloo's statistics major (I am looking at the requirements) would have literally zero idea what is going on in a graduate real analysis class.

If you have never taken any upper-level math or a first course in real analysis, you are going to have a really bad time in these courses. They're also overkill. You just need to take an undergraduate level real analysis class.

Absolutely, especially for a domestic student. For lower-ranked programs, it wouldn't necessarily even be the end of the world if you didn't have real analysis. But OP has essentially no theoretical math background, which is why these programs are not just unlikely but impossible given his current profile.

Washington is more than a reach - it would be a total waste of money. I also think your chances at UC Davis, UT Austin, and UCLA are close to zero. UCSD doesn't have a statistics program; if you mean the math PhD with specialization in statistics, that would also be a waste, but you may want to look into their new PhD in biostatistics. UCSB might be reasonable to apply to, but it will not be a safe school. Your math background is weaker than statistics programs require, and your grades are uneven as well. Unless you take some more math (including real analysis), I think your best bet is to apply to biostatistics programs outside the top 50 on US News.

You'll need to post a more detailed profile (look at other threads for a template) that includes your math background and grades in math classes. Even those backup schools are very competitive and you'll need calc 1-3, linear algebra, real analysis at a minimum to be competitive. You'll also need to take the GRE. My guess is that these will all be reaches for you, and my suggestion will probably be that you apply to lower-ranked biostatistics programs. But post the full profile and we can give you more help.

Look at the US News rankings from 60s-90s (approximately Medical College of Wisconsin down to VCU).

I don't think real analysis is necessary for MS programs, so I don't think that's a big factor, and one B should not have a big effect on admissions. Not an expert in MS admissions though. I think letters and SOP are the easiest thing to improve, as your GRE scores are great so I think you should be getting into good programs.

@JoyboyI agree that the programs you get into are fine, but yeah, if you have a good job lined up, go make some money, re-apply with some stronger letters and try to get a funded offer if you do end up going back to school. I am a little surprised given your stats that you didn't get into a few more programs on your list.

You'll probably spend most of your time cleaning data, figuring out how to do some really boring data pipeline stuff, or being confused as to what you are supposed to do. Every job has its flaws. See this for why almost everyone who works at Verily has left: https://www.statnews.com/2016/03/28/google-life-sciences-exodus/

No, it requires an MS in CS *or a related technical field*. You have one of these. You are too hung up on degree titles. It says you need to learn Python. Do that. Yes. You need to leave pharma, that is clear. Get a tech job and you won't be doing that. I guarantee you this job is not as cool as you think it is. You can do the same stuff at any tech company. The Verily job is not going to be much more interesting than the job you would get at an insurance company, a start-up, a financial company, etc. A million people with PhDs are going to apply for that job and most of them, including brilliant people, are going to be rejected and not get through the interviews. Banking on a specific type of job like this working on wearable sensors is setting yourself up for disappointment. Teach yourself Python and get any intro level data analysis job at a tech company where you use Python and SQL every day. In a year or two you can get a promotion to data scientist and then you'll have the work experience and be able to branch out more.

I think this is a key point -- if you really love statistics/data science so much that you want to do it as a hobby, do it all of that time that you're saving by *not* being in school. Take the extra money you've saved up from your job vs. being a PhD student and take a few months off completely and learn some new stuff full-time. There is literally no difference between this and what you're going to be doing in graduate school besides a mindset. Especially in this past year, the number of free online courses has skyrocketed. I think if presented this way, going back to school loses its charm for a lot of people. It is one thing to go back to school because you need the credential, or you have the resources where the money doesn't matter, but you can do research and learn anything you want about statistics and data science for free on the internet, more than enough to occupy you for a lifetime. I understand where @untzkatz is coming from in that it is a culture shock to be doing boring stuff at work all day when you enjoyed the subject during school. But I don't think going back to relive the glory days is the most productive path for most people. Although some people (me and trynagetby just in this thread) found the trade-off personally worth it.