Importance of GRE Math subject test in top stat PhD programs

[Robbentheking]

Hey guys,

I intend to apply for stat phd programs in the winter and am planning on taking the GRE math subject test in October. Many of the top programs desingate this test as 'strongly reccommended.' I studied math in undergrad, but I took computation-heavy calculus like four years ago, something that makes up a large portion of the test. I've started my review, but even though I would say it comes along well when I get down to it, I'm struggling to find time as I'm currently working full time, and there is quite a bit to go over.

I've seen other posts on this topic (from a few years back) with a lot of mixed feedback on the question. I also know that these programs have only gotten more competitive in the past couple years. So, for people who have some inside knowledge on this, I'm posing the question again. How important would it be for me to submit scores on this test? What would be the cut off for submission (e.g. 70th percentile)? Probably way too specific, but any guesses on what percentage of accepted students at a program like Chicago or Wharton submit the test?

Thanks for any responses.

Edited September 12, 2016 by Robbentheking

[Residuals]

This is one of the great mysteries when it comes to top stats departments. Math departments tend to talk about the GRE math subject test in terms of lower bounds (e.g., UCLA's math states that most applicants are at least 80%). Statistics departments are much more ambiguous. Stanford states that the average for admitted students is 82% but what is the distribution around that average?

Chicago and Berkeley say the test is "optional" but encourage its submission. Students who do really well (say 80% or greater) will clearly submit. But, how many students score poorly and don't submit but get admitted in the end? Thus, we are dealing with censored data. But then I know of one Berkeley PhD graduate who submitted a math subject test score of less than 40% and was admitted. Berkeley clearly saw much more to this person's package than the score (coursework, letters, etc.). Berkeley was right in that this person did cutting edge research and has a successful career.

[Robbentheking]

@arima

First off, thanks for the reply!

15 hours ago, arima said:

This is one of the great mysteries when it comes to top stats departments. Math departments tend to talk about the GRE math subject test in terms of lower bounds (e.g., UCLA's math states that most applicants are at least 80%). Statistics departments are much more ambiguous. Stanford states that the average for admitted students is 82% but what is the distribution around that average?

It's honeslty pretty funny in my opinion. Though it's probably a poor indicator for research potential in both stats and pure math, I feel like this test might be a better indicator for graduate statistics preparation than for graduate math preparation. Timed computations are just the total opposite of everything my undergrad experience indicates about math grad school. Testing was already trending towards take home exams where you had like a week to do some proofs by the end of junior year, and I only was taking undergrad courses at that point. Sure, being able to do basic stuff fast helps a lot in pure math and this isn't to say that stats is just a bunch of computations, but I just see a real discord between what I perceive higher math to be and this test.

15 hours ago, arima said:

Chicago and Berkeley say the test is "optional" but encourage its submission. Students who do really well (say 80% or greater) will clearly submit. But, how many students score poorly and don't submit but get admitted in the end? Thus, we are dealing with censored data. But then I know of one Berkeley PhD graduate who submitted a math subject test score of less than 40% and was admitted. Berkeley clearly saw much more to this person's package than the score (coursework, letters, etc.). Berkeley was right in that this person did cutting edge research and has a successful career.

From Berkeley's website: 'We do not require the Math subject test but if the scores are available you may submit it as part of your application.'

Obviously different departments are going to weigh all this stuff differently. Some seem to be more insistant. Or maybe I'm just reading into this too much. The thing is, I feel like sucess on this test is strongly correlated with time spent studying, especially if you narrow the pool to people who would otherwise be competitive at a top 5 stat program. Not saying I am one of those students, just making an observation that it probably is a bit of an artificial differentiator for otherwise qualified applicants.

 

[Residuals]

I agree with your sentiments on a timed standardized test not being a true measure of a person's academic abilities. I can tell you of numerous stories of UG pure math students who get incredible standardized test scores on the math subject GRE but, in class, they are only "so-so" students. And there is the opposite, stellar and meticulous in-class students who just can't score high on the standardized test.

To your original question, what is the threshold for submitting? Does one submit a 50-70% score? I told you of a case where a less than 40% score was submitted and he was admitted. Was he lucky in that the committee brushed it aside and looked at the rest of his record? We can't infer from a sample size of 1. It is a tough decision. What are you thinking of doing given whatever score you get?

[abstract_art]

Just wanted to add that from posts from previous years in this forum it looks like the math gre is more important for internationals than domestic students (as are most parts of the application)

[Robbentheking]

1 hour ago, arima said:

To your original question, what is the threshold for submitting? Does one submit a 50-70% score? I told you of a case where a less than 40% score was submitted and he was admitted. Was he lucky in that the committee brushed it aside and looked at the rest of his record? We can't infer from a sample size of 1. It is a tough decision. What are you thinking of doing given whatever score you get?

Well, luckily I do have some time to mull it over, but I'm thinking 60-70th percentile as some sort of cutoff if it's recommended, maybe a bit lower if it's "strongly recommended." But to be honest, I really don't know. Couple of things I'm considering:

1) Because it seems like it holds more importance in the math world, I think students probably place more emphasis on it and do better on it there on average. So, I'm guessing the submitted scores are a bit lower for stat programs in general. There seems to be slightly more complete data on cutoffs at math PhD programs, so I might just find as much info on that as I can and subtract an amount that sounds reasonable. Considering how much of a pain it is to go back and do this review and the fact that it's technically option everywhere I'd apply (Stanford is realistically out of the question lol), I think a sort of lower but respectable score (like high 50s) might not even look that bad, depending on where you're applying.

2) I'm actually fortunate enough to have gone to undergrad at a place with this ambiguous recommendation. I may try to send a few emails around and see if I can't learn a bit more. Then again, I might end up applying there, so I don't really want to come off as lazy or something to someone in the department, especially considering if there's any place with a recommendation where I could probably get away without it, it's my old school, because after all, there's probably someone on the admissions committee who's friends with a math professor who can tell them exactly what it means for a student to get an A- in Topology haha

[MLHopeful]

For what it's worth, I was admitted to three schools that "highly recommend" the math GRE without submitting it (Chicago, Washington and Columbia), as well as Berkeley. I got ~70% and only submitted it to Stanford (where I did not get in), based on the recommendation of my undergrad research adviser. I also had a very strong math background which probably earned me the benefit of the doubt - multiple graduate level measure theory courses with all A+s, and a letter from a prof I spent a summer with doing research in abstract analysis. You can probably find my profile somewhere online if you want more context.

[Residuals]

MLHopeful, you clearly have a strong mathematical background. May I ask why you didn't submit a 70% Math Subject scores to schools like Berkeley and Chicago? I would think that a 70% is very respectable score for a stat applicant.

I do have a question that I have always wondered. Do you have a sense of how much weight is given to the verbal score on the general test? I am guessing not much.

Edited September 18, 2016 by arima

[yosemitesam]

Hi all, reviving this thread; let me know if it would be better if I start a new one.

I took the GRE math subject test last weekend, and unfortunately I kind of blew it. Didn't get great sleep the night before, plus I lost track of time and wasn't able to bubble in guesses for any of the questions I didn't answer. Since there's no penalty for guessing anymore, that cost me at least a few points.

Anyway, I have the same question that didn't seem to quite get a definitive answer here--at what percentage level do GRE math subject scores become a liability rather than an asset? (I'm sure there's no definitive answer, but I'd like to get some insight if possible.) Here's my profile: 

I was hoping to score well on this test to bolster my math credentials since my math background is a little thin for top stats programs. If I managed to squeeze into, say, the 60th percentile, should I include those scores? What about 50th? I don't think any of the programs I'm applying to REQUIRE scores. Any advice and input is appreciated!