GRE Math Subject Test

[maxent]

Are there any guidelines for what constitutes a "good score" for Stat PhD programs that recommend the GRE Math subject test? Top Math PhD programs set 80th percentile as a bar, so presumably Stat PhD programs consider lower scores to be good.  Stanford notoriously requires the test (and they report an average of 82 percentile), but presumably programs that only recommend the test have lower ranges? Previous threads that have discussed this topic haven't really come to any conclusions, and I wonder if anything has changed since then. Is 70-75 percentile considered "good"? 65-70? At what score should an applicant forgo sending to a program that recommends it?

[MathStat]

I would also be really interested in finding out this, especially for a score in the low 70s percentile. 

[miserablefunction]

My impression is that, for international student, for statistics PhD, I think the lower bar is 80%. Programs which require or encourage GRE math subjects are very competitive. For domestic, I think it is lower, somewhere 70%. If you check the stanford website again, I believe they updated that their average subject test score for incoming students are like 89%. So for internationals, to have an edge through gre subject test, I think 90% or above is a safe guess. Indeed, many students at the top programs have near perfect scores. At least my friend who is in top 10 statistics program, his raw score was above 900. And similar for other asian international students that I am aware of. But this is just my impression.

Edited October 13, 2018 by miserablefunction

[statfan]

According to my supervisor, the GRE subject test most often acts as a screening test and is less important than you think, even for schools that require it. With that said, if you score too low, it could raise concerns, however, if you pass the cutoff it won't matter much. For top statistics programs, I guess 70% would be ok and anything over 80% would be considered a very good score, and I doubt the bars are different for international and domestic since afaik some international students still got into these schools without taking the test. If you get rejected with an over 80% GRE math score, it is not because of your GRE. 

[maxent]

Does Stanford really have an average of 89%? For reference, I looked at https://statistics.stanford.edu/academics/phd-admissions-faq

@miserablefunction It seems odd to me that the score bar should be that high for international students at top programs. It strikes me as very weird that a top program like Stanford requires the test, but another top program, Berkeley, does not require the score. If a really high score was an important signal for success in statistics graduate studies, shouldn't all of the top programs require the score (like they do in Math PhD programs) ? The fact that many programs (such as Berkeley, CMU, Duke, Michigan, Wisconsin, UNC etc) do not require let alone recommend the test leads me to believe that the bar is lower. But again, this is just my (very) naive impression. 

@statfan Your description makes a lot of sense to me. It is kind of unfortunate that programs aren't very transparent about expectations, unlike the math world.

 

Edited October 13, 2018 by maxent

[statfan]

5 minutes ago, maxent said:

Does Stanford really have an average of 89%? For reference, I looked at https://statistics.stanford.edu/academics/phd-admissions-faq

@miserablefunction It seems odd to me that the score bar should be that high for international students at top programs. It strikes me as very weird that a top program like Stanford requires the test, but another top program, Berkeley, does not require the score. If a really high score was an important signal for success in statistics graduate studies, shouldn't all of the top programs require the score (like they do in Math PhD programs) ? The fact that many programs (such as Berkeley, CMU, Duke, Michigan, Wisconsin, UNC etc) do not require let alone recommend the test leads me to believe that the bar is lower. But again, this is just my (very) naive impression. 

@statfan Your description makes a lot of sense to me. It is kind of unfortunate that programs aren't very transparent about expectations, unlike the math world.

 

The average varies from year to year and the 89% average should be the average for the incoming class last year. Unfortunately, the GRE math test is neither an accurate measure of math skills nor a good indicator of future success in graduate programs so most schools don't take it very seriously. If your math background is relatively light or you come from an unknown school, a decent score, say over 80% may help. Otherwise, your math skills are reflected on your transcript much more than the GRE.

[Gauss2017]

There have been lots of studies to indicate that the use of the GRE subject test causes discrimination against women and under represented minorities.

This years Stanford cohort does not appear to have any women.  I don’t think Stanford statistics in the last 5 or 10 years has had any URMs.   There appears to be a causal connection.

Edited October 13, 2018 by Gauss2017

[miserablefunction]

I never said it is the only factor which would decide your admission. I am addressing the question what range of scores will be considered as high. I will only talk about international student case here. It is just a fact that domestic admission bar is much lower since they have many funding resources. I guess they won’t have explicit bar to screen out candidates, but adcomm will naturally expect higher scores for international students. In the view of statistics department, international students require more funding, which makes them carefully screen over and over. You can get into programs which do not require gre subject without a good score, but you should have something strong, like publications to be competitive. My point is schools like Stanford, Chicago, Columbia, Upenn and UW, which encourage(or require, in case of Stanford) one to submit GRE subject score do have pretty high average of scores for incoming students. 820 was stanford’s data 7-8 years ago. The admission got much more difficult recently. If you check admitted international students profile of top programs, you can easily see that they are just perfect. Many of them have publishments and have taken advanced graduate level math/stat courses. So if you are an international, and want to impress adcomm, you defintely need to have a higher score than domestic. And you are right that some people get in with low scores, but I think submitting 700-800 scores for aforementioned schools are quite meaningless for international students. Adcomm will simply ignore it. International openings for the schools I mentioned will be filled by top people from Peking, Tsinghua, IIT, ISI, SNU, NUS and etc. To these kids, getting 900 range score is no problem at all. I agree that gre subject math test is not an accurate measure of one’s math proficiency, but in the view of adcomm, this can be another reliable source to assess one’s mathematical maturity. And there are sufficient number of strong candidates who would do well in their programs, so my guess is that they have no choice but to rely on every little thing they can distinguish applicants. Also math departments are known to have 800 as their bar but if you actually look at profiles who got in, most of them have 85 above percentiles. So I think to give an impression that one did well(fine) on subject test, I think the bare minimum would be 80 percentile. If you add all international openings for the five programs I mentioned, it would be around 25-30 slots. There are many strong applicants from China, India, Hongkong, Singapore, Korea and Europe. And also there are international students in ivy leagues, top public and etc. These number simply outnumbers the openings easily. The link below is updated info of stanford stat phd subject gre test. They did not say it is this year’s data, but recent. Indeed the admission got much fiercer and if you just rely past data on this forum, you are missing something. https://statistics.stanford.edu/admissions/phd/phd-admissions-frequently-asked-questions

Edited October 13, 2018 by miserablefunction

[statfan]

43 minutes ago, miserablefunction said:

I never said it is the only factor which would decide your admission. I am addressing the question what range of scores will be considered as high. I will only talk about international student case here. It is just a fact that domestic admission bar is much lower since they have many funding resources. I guess they won’t have explicit bar to screen out candidates, but adcomm will naturally expect higher scores for international students. In the view of statistics department, international students require more funding, which makes them carefully screen over and over. You can get into programs which do not require gre subject without a good score, but you should have something strong, like publications to be competitive. My point is schools like Stanford, Chicago, Columbia, Upenn and UW, which encourage(or require, in case of Stanford) one to submit GRE subject score do have pretty high average of scores for incoming students. 820 was stanford’s data 7-8 years ago. The admission got much more difficult recently. If you check admitted international students profile of top programs, you can easily see that they are just perfect. Many of them have publishments and have taken advanced graduate level math/stat courses. So if you are an international, and want to impress adcomm, you defintely need to have a higher score than domestic. And you are right that some people get in with low scores, but I think submitting 700-800 scores for aforementioned schools are quite meaningless for international students. Adcomm will simply ignore it. International openings for the schools I mentioned will be filled by top people from Peking, Tsinghua, IIT, ISI, SNU, NUS and etc. To these kids, getting 900 range score is no problem at all. I agree that gre subject math test is not an accurate measure of one’s math proficiency, but in the view of adcomm, this can be another reliable source to assess one’s mathematical maturity. And there are sufficient number of strong candidates who would do well in their programs, so my guess is that they have no choice but to rely on every little thing they can distinguish applicants. Also math departments are known to have 800 as their bar but if you actually look at profiles who got in, most of them have 85 above percentiles. So I think to give an impression that one did well(fine) on subject test, I think the bare minimum would be 80 percentile. If you add all international openings for the five programs I mentioned, it would be around 25-30 slots. There are many strong applicants from China, India, Hongkong, Singapore, Korea and Europe. And also there are international students in ivy leagues, top public and etc. These number simply outnumbers the openings easily. The link below is updated info of stanford stat phd subject gre test. They did not say it is this year’s data, but recent. Indeed the admission got much fiercer and if you just rely past data on this forum, you are missing something. https://statistics.stanford.edu/admissions/phd/phd-admissions-frequently-asked-questions

I have to disagree with you. My supervisor obtained her phd from a top 5 statistics/biostatistics program and she knows much more about the admissions process than we do. According to her, although more and more people apply each year, the competition does not get fiercer and fiercer. Many students apply to statistics phd simply because statistics gained popularity in recent years. However, some of them did not even meet the minimum requirements and they won't get in. The reason why it is harder for international students to get into top phd programs are: (1) Their referees may not be well-known in statistics. (2) The admissions committees are concerned about their English skills. (3) They come from an unknown institution. Interestingly, applicants from a Canadian top school seems to do much better than other international students because they subject to none of the three. Personally, I know quite a few Canadian students in my school who got into UW/Chicago/Columbia without having taken the math GRE test. But they have great research experience and excellent math preparation.

[miserablefunction]

Canadians are different. Here I am talking about international students whose native language are not english. As many others in the forum indicated, US programs treat Canadians quite differently from other international students who are from Asia. Believe it or not, I am from top 10 stat department and this is what I heard from the admission chair in a private conversation. Recently the bars for admission got higher and candidates at the level that they used to accept were not accepted in the past two years(they made a long waitlist but still they ended up giving rejections to most) He carefully guessed that due to increasing popularity in statistics many more asian students from top universities like peking, tsinghua start to apply for phd in statistics. Those who are top math not statistics major nowadays also apply to statistics/machine learning programs and indeed stat phd programs love them. If you are from Peking, Tsinghua, ISI and are top student, no one will doubt your capabilities. Just count the number of prominent statisticians who have BS in Peking. It is just massive. In addition due to huge chinese community in statistics, the university that is well respected in admission does not just stop at Peking, Tsinghua. There are at least 5-6 universities in china which would be regarded on par as competitive universities in north america when it comes to PhD admission. I haven’t seen any single top program who did not have a chinese admitted student from tsinghua or peking. They have a solid record of students. Canadians are better in terms of communications but not international students in north america are in general better than those who studied in top universities in China or India. Come on, India’s native language is even english. It is true that the chance is higher for those who studied in north america will get better letters of recommendation, but nowdays there are just so many good researchers in statistics in peking or HKU as well. And given the population in china and india, the stat department openings are quite smaller than you expect when you experience this year’s application, especially for international students.

Edited October 13, 2018 by miserablefunction

[miserablefunction]

And I do not know why your advisor thinks like that but there are actually quite many qualified applicants among the pool. Statistics program is just not as big as other engineering, and surprisingly their size has remained similarly for the past 10 years. I am one of those who wish them to expand their size to provide more education opportunities who wish and I believe they should, given the interest of the discipline nowadays. 

[Schweinchen]

What I heard from somebody who's on the admissions committee at a top school is that if you're applying from China and you get anything below the 90th percentile, it's going to hurt you; if you're applying from India, then the cut-off is somewhere between the 80-85th percentile; if you're applying from within the US, you can go as low as 75th, particularly if you belong to an under-represented group. The bottom line is that average percentile is a poor measure; say you're 80th percentile, that would be fine if you are at an American college but you're screwed if you're applying from China.