Math GRE Subject test for (bio)statistics PhD programs

[migos_season]

Hi all,

Scores were just released today for the April math GRE subject test. I was wondering whether anyone had any knowledge about how useful this was for admissions. I've seen a lot about the regular GRE but not much about the math GRE, at least in the context of stat/biostat PhD programs. I know that most stat/biostat PhD programs don't require or even recommend this test, but I had a few questions:

1) What is a "good" score (for, say, a top 10 program)?

2) How much will doing well help my chances at top PhD programs?

3) Is there a threshold used by any schools?

Just generally would like to know a bit more about how the subject test can help/hurt candidates, etc. Thanks!

Edited May 11, 2018 by migos_season

[Bayesian1701]

It's hard to generalize because not every top program (Duke doesn't take scores) requires it, and a good score is best understood in context of your entire application.  If you already have evidence of mathematical maturity in your transcripts (high grades in upper-level and graduate courses) and an overall great application your score is not as crucial as someone whose application is below average (at specific programs).  But a near-perfect math GRE will not by itself guarantee admission to any program. Over the years there has been debate over this in the forum, and you can find the old threads by searching.

  I never took the Math GRE, since I was unwilling to take it just for Columbia (the only program on my list that recommended it).   Stanford says that the average admitted score is in the 82nd percentile, so something close to that is probably a good score (or at least a score good enough to make you stand out).    I think this is the only average that has been released by a program. Hypothetically If I had a score of over 82nd percentile, I definitely would send it because that would be higher than Stanford's average.  But how low is to low to send is a tricky question.  For Columbia, I probably would have set a cutoff for myself (had I taken it) of at least 60th percentile to send.  

[Radon-Nikodym]

3 hours ago, Bayesian1701 said:

Stanford says that the average admitted score is in the 82nd percentile

This is no longer the case. Stanford updated its FAQ page and now the average subject test score for admitted Statistics PhD students is 89th percentile, which roughly corresponds to a score of 870. However, I received a 900 (95th percentile) as a domestic student, and I was not accepted at Stanford, so you are right that high GRE scores do not guarantee acceptance. Their standards are incredibly high.

Edited May 12, 2018 by Radon-Nikodym

[abstract_art]

Just want to throw out that I've heard anecdotes of students being admitted to all of the programs that recommend the Math GRE w/o taking it (including Stanford!)

[migos_season]

Thank you all for the help. I don't know if anyone can really give advice on this, but I got an 82nd percentile. Would submitting the 82 hurt me? Do you think I should try and study to retake it and hopefully get a better score?

Edited May 12, 2018 by migos_season

[Bayesian1701]

On 4/3/2017 at 2:20 PM, cyberwulf said:

The math GRE is most likely to benefit a student with an otherwise excellent record but who has taken relatively little math (for a top program, say "only" up to undergraduate real analysis and abstract algebra). Then, a good score in the math GRE (which I would qualify as about 75-80th percentile) would be helpful. Another student profile that could benefit is someone who has a very good but not outstanding record in math classes, but absolutely crushes the math GRE (say 90-95th percentile).

In both cases, the math GRE score adds useful additional information to the profile which could elevate them enough get into a place they might otherwise not.

cyberwulf is a biostat professor and this is what they posted about the math GRE.  I would say submit.  I would guess that the 82nd percentile on the test would translate to a least 82nd percentile (probably higher) if you focused on only statistics applicants.  Looking back at past profiles for 2018, 2017, 2016, and 2015 that seems higher than most people.  Of course, the profiles are an incomplete sample and very few people who took it posted their results so there isn't much data.

@Radon-Nikodym  Thanks, their website was down when I was writing the post so I couldn't verify if it had changed when I wrote it.

[Gauss2017]

The only place it is necessary for is Stanford.  So generally it makes absolutely no difference whatsoever