HOWTO: Get in to grad school for science, engineering, math and computer science

Juniors and seniors often ask me how to get into a Ph.D. program. Having served on the graduate admissions committee, I finally feel like I can offer some good advice. [This advice applies for masters students too, but the bar for acceptance is lower.] The one-word version of that advice is: PUBLISH.

Feel free to if you have questions.

What graduate schools want

When graduate schools are admitting students, they are trying to answer one question from your application materials: "Does this person have the potential to conduct scientific research?"

When I review applications, I look at three things:

prior research (published or unpublished);
letters of recommendation; and
personal statement.

If you've already done research and published it, then you've answered the question for the admissions committee. If you have unpublished work, submit it with your application, and someone from the field will judge the quality of the work. If you don't have any prior research, don't lose hope. You can still get in, because schools are trying to judge potential to do research. [I didn't have any research papers when I applied to Ph.D. school.]

Masters-only tip: If the school lets you select thesis or coursework as a preference, select thesis. Selecting coursework signals the admissions committee that you're not interested in research!

Letters of recommendation

When letters of recommendation come from active, well-known researchers in your field of interest, a sentence in your recommendation like, "I've supervised her on a research project, and I have witnessed and believe in her potential to do research," counts for a lot. Recommendations like, "This student took my class and got an A," can't really help the admissions committee discern your research potential. Doing a supervised independent study on some research topic is a great idea, because you'll get a fantastic letter of recommendation out of it.

When you ask for a letter of recommendation from a professor, don't ask them if they can write a letter of recommendation. Of course they'll say, "yes," to that. Ask a professor if they can write a strong letter of recommendation. This provides them a way to say "no," and save you the embarassment of a crappy recommendation letter.

Provide your recommenders with all the necessary materials: pre-addressed stamped envelope, due dates and your application materials. Send email reminders and check with schools that they've received recommendations.

Also, talk to a professor at your school in the field in which you plan to do research. Ask them to which schools you should apply for that field. At this point, they should contact professors at those schools in these areas, let them know you're applying and give a candid assessment of your abilities. These behind-the-scenes recommendations are priceless.

Personal statements

Personal statements should be short (one page), and anything important like the name of Professor X should be in bold. "Personal statement" is a terrible name for this document, because it confuses applicants. Use this statement to answer the following question in essay form: "Why should we, the admissions committee, believe that you, the applicant, have the potential do research in field X?" and "What kind of research could you see yourself doing and why?"

A personal statement should not cover your childhood experience with science, computers or math.

Do the math

There is way too much randomness in the grad-school admissions process. Many excellent applicants will be rejected for reasons totally apart from their research potential: they selected the wrong potential advisor, or their personal statement was too long, or their application was read later in the process when reviewers are exhausted, or no one even bothered to read it at all.

Recognize that acceptance rates, even at "lowly-ranked" schools hover around 10%. At "top" schools, acceptance rates will be in the low single-digit percentages. If you only want to go to a top grad school, then you're going to grad school for the wrong reason, and the odds of you getting in are low. You should go to grad school because you want to do research, and you don't need to go to a "top" school for that.

A top schools, virtually all applicants are qualified, which means that your probability of getting in is roughly the same as the acceptance rate. Assuming a 5% acceptance rate, if you apply to 10 "top" schools, your probability of going to grad school is (1-0.95¹⁰) = 40.2%. If you really want to go to grad school, then the odds are that you'll end up disappointed if you take this strategy.

If you apply to ten "top" schools (5% acceptance rate) and ten "regular" schools (10% acceptance rate), then your probability of going to grad school is (1-0.95¹⁰0.90¹⁰) = 79.2%. This is a lot better, but it still feels a little low to me.

Decide ahead of time on the probability you'd like to get into grad school, and compute the appropriate mixture of "top" and "regular" schools to which you should apply.

What doesn't matter

GPA? I don't care if it's 2.0 or 4.0. I won't even look at it. The school you went to? I'll judge you the same whether you went to Nowhere State U or MIT. Transcripts? Never seen one. GREs? Irrelevant. Where you work/worked? Unless it's a research lab, it's not important. I don't think these items have much predictive capacity as to whether or not someone can complete a Ph.D.

Ten application tips

Contact a faculty member you'd like to work with. Email them a month or so before you apply. Tell them you were considering applying, and you're curious about the research opportunities available in the field. Comment intelligently on some research that faculty member has done. Attach any research you've done, and briefly summarize your research interests. That faculty member can then make sure your application receives a thorough review. Bear in mind that professers receive lots of form-letter spam from prospective students. It's painfully obvious when the email is form-letter spam, and most professors will summarily discard it.
Be brief. Even "lowly ranked" schools will receive hundreds or even thousands of applications for a few dozen slots. Most applications are skimmed first, and read only if something catches the reviewer's eye. Bold-facing items can help catch a reviewer's attention. There simply isn't time to read long-winded applications.
Choose your area of interest/preferred faculty carefully. Applications are reviewed by the faculty in the area for which the prospective student states an interest. If you choose this poorly, the right person will not see your application. Reviewers also get annoyed when there is a mismatch between area preference and faculty preference. At least skim the home pages of every faculty member. It's also a good idea to look for faculty with an active research program and current Ph.D. students. Faculty without funding can't easily admit students. New/pre-tenure professors are especially eager to find good graduate students, and sometimes they have start-up funding to use until they get a grant.
Be different. Don't talk about how you've been interested in the field ever since you were a child and that you wrote your first program/proved your first theorem at age eight. The admissions committee already knows that it's been your lifelong dream to become a scientist. (That's why you're applying.) Many personal statements start off this with this "standard back-story," and it's a waste of space.
Use quotes carefully. A lot of personal statements start off with a quote. If you use a quote, make sure it's witty, relevant and one that the reviewer has never seen.
Put up a personal home page. Make it professional. Highlight any interesting projects you've worked on there. Remove all references to politics and religion. (Clean up your facebook profile, too.)
Proof-read your documents. This should go without saying, but having typos in your statements looks sloppy. If you're not a native English speaker, have a native English speaker proof-read your materials.
Make your application look good. I know this shouldn't matter, but I find myself putting more effort into well-typeset applications. I notice when applications use LaTeX (and use it well), too.
Choose your recommenders carefully. Cultivate working relationships with your recommenders. This is the only way to get convincing recommendations out of them. (If you tell a professor you're interested in research after class one day, they'll have you helping out on a research project by that evening. Professors are always short on research manpower. Trust me.)
Don't get a job. Once you get accustomed to a real salary and you start putting down roots, it's going to be difficult to go back to being a student. You'll be living in a small apartment, working on demanding problems all day long and getting paid a subsistence wage to do so. This will be your life for four to seven years. The least inconvenient time to do this sort of thing is right after undergraduate school. I often tell undergrads pondering a Ph.D., "You're only dumb enough to get a Ph.D. once, so you'd better not let the moment pass you by."

External resources

Mor Harchol-Balter on Appyling to Ph.D. programs in Computer Science.