Applying to Statistics PhD Programs

Lee Richardson just graduated from the University of Washington with a BS in mathematics and statistics. He worked a year for the Institute of Health Metrics and Evaluation and is now a PhD student in the department of statistics at Carnegie Mellon.

Editor’s Note: Reposted with permission from Lee Richardson’s blog, Statistical Signal.

During the last several months, I’ve spent the large majority of my free time applying to statistics PhD programs. After accumulating so much information on the topic, I felt it would be a shame not to pass it on to applicants next year and beyond. This post is an attempt to pass on everything I’ve learned about the application process, with hope it will be of value to future applicants.

Statistics has risen in popularity over the past several years. Clearly, this is due to the meteoric rise of Nate Silver (just kidding). I would actually say it’s a multitude of factors: financial prospects, media coverage, and, of course, sexiness. Whatever the reasons, the implication for aspiring PhD students is that the competition is growing. I remember first asking graduate students last year about my chances, and the consensus was, “Of course you’ll get in everywhere!” Most were genuinely surprised by how many departments accepted them.

While this may have been true in the past, a deep look into the epic Grad Cafe admissions thread tells a different story. There are qualified students being rejected from the majority of schools they apply to, and this trend does not appear to be reversing anytime soon.

The first question to ask yourself is: Do you want to pursue a master’s, one of the newly developed one-year analytic programs, or a PhD? This question could spawn an entirely new discussion of the pros and cons of all three, but it really depends on what you want.

Attributes of a PhD

Generally fully funded, 4–5 years

Attributes of a Master’s

Generally not funded, 1–2 years

I’m dramatically simplifying, but these are two pretty significant things to consider. Since I only applied to PhD programs, that’s the realm in which my advice will be the most useful.

Applying to Statistics PhD Programs

There are four main components graduate admissions committees will use to evaluate your application:

• GRE scores
• Letters of recommendation
• Grades/transcripts
• Statement of purpose/CV

Below is an attempt to go through these with as much helpful information as possible. This will include tangents of other things I thought were important.

Statement of Purpose

Something I found controversial was the debate over the statement of purpose (SOP). See the snippet of a Twitter conversation I had with Julian Wolfson, a professor at the University of Minnesota Department of Biostatistics.

This isn’t just shameless self-promotion of my Twitter handle, there’s serious debate as to how useful the statement of purpose is. I’m a little biased here, because I wasn’t aware of this until I spent a lot of time on the SOP, but I do think it’s important! This is the only portion of your application where you can explain things that didn’t fit anywhere else. The SOP is also an opportunity for committees to assess the applicant’s English, so non-native English-speaking applicants should seek out a native speaker to read their draft before submitting.

The key argument against the SOP is that graduate committees are primarily concerned with mathematical ability and research potential, and the SOP isn’t a great indicator for this. I think there’s merit to both views, but since the SOP is the most laborious process that forces you to think about your strengths, why you’re applying, etc., there is value beyond just increasing the likelihood of acceptance.

My intuition tells me that some schools look closer at the SOP than others do. I have my department-specific hunches based on how quickly some schools responded, but this could vary from year to year based on things like committee members, number of students leaving, etc. There’s also the argument that since the number of applications has risen, there’s less time for committees to spend with individual applications. A lot to consider for sure.

NSF-GRFP Fellowship

Spoiler: Here’s what I think is one of my best pieces of advice. Although it was time consuming, I think the best decision I made was applying for the National Science Foundation Graduate Research Fellowship. The biggest upside is the deadline is much earlier than departmental application deadlines, so it forces you to think hard about why you’re applying to graduate school and what type of research you would like to be doing. Then, you can write what can easily be adapted into a SOP. Even if you don’t get the fellowship (I didn’t), you will have a huge jump-start on finishing your applications. I came away with a polished, general template that worked as an SOP for nearly all my applications (only Michigan and Berkeley required personal/diversity statements beyond the SOP).

Another significant benefit of the NSF GRFP is that it requires a research statement, which means you have to think deeply about what type of statistical research you will be doing in graduate school. You might be thinking, “What the hell is statistics research?!” Don’t worry, you’re not alone, and it’s perfectly all right if your research statement is bad. In fact, one of the most retrospectively funny moments of my application process was meeting with one of my professors to discuss my application a couple days after it was due. He remarked, “Your personal statement is good, but your research statement needs a lot of work!” Oh well, another plus is that you’ll have two more years to apply. (Note: There are more fellowships than the NSFGRFP fellowship, Such as the National Defense Science & Engineering Graduate Fellowship and the Hertz Foundation.

The research statement ties into a related debate regarding the importance of contacting professors beforehand. Visit this for a good summary of best practices if you decide to do it from a related field (yes, that’s the same person who teaches Udacity’s CS101). I invested an insane amount of time researching professors, reading their papers, and looking closely at the structures of many departments. I found this to be very useful, but it’s my no means necessary; there are indeed people who got into great programs without emailing anybody. The reason this is related to the NSF GRFP fellowship is that when you’re putting together your research statement, you’ll be reading papers of professors, and, naturally, those are the one’s you’re most likely to contact.

CV/Research Experience

Another requirement is to submit a CV, which has no real defined structure, but a vague Google search can lead you to many templates/examples. The key part for me was being able to talk about research experience and ASA membership and deliberately highlighting the best parts of my application. There’s not a lot of unique information in the CV that isn’t covered elsewhere, but it gives an opportunity to frame you as attractively as possible.

This leads to another conversation regarding research experience. It can be comforting to know that some say having real, statistical/mathematical research experience before attending graduate school is rare and not strictly necessary. While it’s perfectly acceptable to enter without research experience, it’s obviously a plus if you do. If you’re in your early years of undergraduate study, I highly encourage seeking out research experiences at your institution or elsewhere through an REU or job. The real benefit of doing research before applying to graduate school is trying it out and seeing if it’s something you enjoy. You don’t want to plunge into a PhD program not knowing whether you enjoy research, because if you don’t, it’s going to be a long ride.

GRE

If you look at the number of applications Stanford receives, you might think “Hey, something isn’t right here.” Well, the reason is that they require the mathematics GRE test, which weeds out lots of potential applicants. If you’re hell-bent on going to Stanford, then you have to take the mathematics GRE. (Some other schools—like UW, Chicago, and Columbia—strongly recommend it.) I didn’t take the exam, but I’ve heard it is pretty hard and requires a lot of studying and deep coursework in pure mathematics. There’s uncertainty around what particular score would hurt/help you if you include it in your application. Stanford posts an average of the 82nd percentile (thanks, wine in coffee cups), but it’s unclear what the distribution of scores are, and whether the scores at Stanford are similar to scores at other schools that strongly require it.

The most annoying part of applying was the GRE test. You’ll be quizzed on vocabulary words you can effortlessly look up, write timed essays on strange topics, and answer short quantitative questions. For our purposes, the most important part is by far the quantitative section, a significant piece of information graduate committees use to assess mathematical ability. I can understand this; as a statistics department you probably don’t want to invest in someone who has evidence of struggling quantitatively. However, if you’re coming from a relatively quantitative background and take some practice tests, you should be able to score in the mid-high 160s. The GRE won’t get you in anywhere, but it can disqualify you. So the takeaway message is to not bomb it.

After you take the GRE test, you can send your scores to four departments for free. If you want to send your scores to anyone after this, it costs you $25 per department. That’s right; the ETS is charging $25 to send about 8 bits of information. The Singularity Is Near.

Grades/Transcripts

This is pretty straightforward. Basically, you want to take as many math/stat/computer science classes as possible and get the highest grades you can. It’s not strictly necessary to be a math/stat/CS major, as Kristin Linn famously majored in music before getting into the PhD statistics program at NC State, but it will boost your chances.

The most common classes you see as strictly required are multivariate calculus and linear algebra, and some departments strongly recommend taking a real analysis sequence. If you’re still an undergraduate, I recommend taking as many mathematical proofs courses as possible. In fact, Peter Guttorp recommended that I should take the real analysis sequence in my gap year.

Letters of Recommendation

Letters of recommendation (LOR) are a tricky topic. It seems you have to be strategically fortunate to stumble upon good LOR writers. Obviously, the younger you are, the more time you have to develop a relationship with a professor or boss through a research project and lock down a good letter testifying you will be successful carrying out research in graduate school. That’s the golden standard for an LOC writer. Another thing to do while you’re still in the early years of undergrad is participate in an REU. I’ve heard many people get their LORs from REU supervisors.

You will hear this ad nauseam, but I’ll reiterate that it’s better to get an LOR from someone less well known who knows you well than someone well known who barely knows you. If you’re a senior and you don’t have anyone who can write a good letter, my suggestion is to go all out next quarter, be a totally overachieving student (going to all the office hours), and see if you can connect with a current professor. Better yet, it would be good to seek out advice from people in the department about working on an undergraduate research project. Chances are there are professors in your university looking to have some data collected/cleaned/analyzed. It’s never too late to establish a good relationship with a professor; however, it does take time and effort.

The LORs are tricky, but they’re also a critical piece of your application. It can certainly be awkward asking a professor to write one for you, as professors are very busy people. However, you should keep in mind that they are asked for LORs all the time as part of their job. Try to establish as many good relationships as possible, but if it’s not feasible, it’s not the end of the world to default to professors you’ve simply taken a course with. One piece of advice is that, instead of asking whether a professor can write a letter, ask if they can write a good letter. By asking directly, you can avoid sending unflattering letters to admissions committees.

Where to Apply

At a certain point, you’ll have to live with your performance in these categories and start filling out applications. The natural question to ask is, “Where should I apply?” There are some basic places to start looking—the most recent graduate department rankings and this insanely comprehensive list of all statistics programs—but I think the most important thing is to meet with someone in your department (if you’re still an undergraduate) and solicit suggestions on where to apply. If you’re not an undergraduate, an alternative to this is posting your profile on TheGradCafe, and you’ll probably garner some candid responses.

One of the biggest mistakes I made while applying was never getting an honest assessment from someone inside the process about how competitive my application would be. Per sources at TheGradCafe (biostat_prof, cyberwulf), the tier system is a realistic way to view schools. This basically means that if you’re accepted into one highly ranked school, chances are you will be accepted into others—the converse is also true (IFF!). I’m somewhat skeptical about this. I do think some schools look more in depth at different factors than others do, and that they’re looking for students who fit their culture. There are applicants who get into one of their reach schools, despite being rejected by others reaches (yours truly). The bottom line: You’re not going to get into anywhere you don’t apply, so I wouldn’t let rankings intimidate you if it’s a school you really want to attend.

My personal criterion for applying was: “If you get rejected everywhere except this school, would you actually go?” This was retrospectively pretty risky, and it’s certainly possible I could have been rejected everywhere (everyone’s worst-case scenario). If your goal is to be in a PhD program no matter what, I would suggest having someone give you an objective, candid assessment of which schools you could potentially be accepted into (described above).

Apply to some reaches, but also apply to schools you’re fairly confident will accept you. Other than about $100 and some opportunity cost, being rejected from schools doesn’t really carry any downside.

A reason PhD programs are so selective is that they’re investing money into your potential as a student, and it’s a waste for them to invest resources in someone who won’t be successful. Odds are certainly higher applying to master’s programs; however, you will most likely be footing the bill (although I’ve heard of some master’s programs funding students).

TheGradCafe

Chances are the majority of your social circle will not be simultaneously applying to statistics PhD programs. Due to this, you may be seeking people eager to talk about the application process. TheGradCafe is a terrific community where fellow applicants can discuss the application process. I found it a useful place to discuss various topics relating to application season (as well as express my fears of being rejected everywhere). Also, it’s extraordinarily entertaining to meet people on visits whose usernames you recognize.

Costs/Suggested Scholarships

Applying to graduate school is quite costly. You can get a rough estimate from the following:

200*(# of GRE Tests) + (25[GRE]+75[App Fee]) *(Number of Schools)

Plus countless hours and -5 years of life expectancy due to stress (sort of kidding). So, if you take the general GRE once and apply to 10 schools, you’re looking at roughly $1,200. One thing I’ve always felt would be a good idea is for undergraduate institutions to foot the bill for their students’ application fees with something like an “applying to a STEM PhD” scholarship. This could easily already be happening, so my apologies to anyone who’s already doing it.

Post-Application Stress

I didn’t want to make this post about anything except the application process, but in case I don’t revisit this topic: It’s pretty common to feel as if you’re not going to be accepted everywhere directly after you’ve applied everywhere. Relax, you probably will, especially if you’ve gotten honest assessments of where you stand as an applicant. Applying to PhD programs is just extremely stressful.

If you have specific questions for Richardson about applying to a graduate statistics program, visit his blog, Statistical Signal.