Unusual Applicants - seeking advice on Applied Math and Stat PhD applications

[Spaceage]

Hi Guys - I was hoping the community would be able to offer some advice for unusual applicants. I've read a lot of threads that feature the typical undergrad Math/Stat/CS majors looking to jump right into a grad program, however, I haven't been able to find too many for people looking to switch industries or re-educate themselves with prerequisites with the intention of applying for Math/Stat/CS PhD programs. RE myself, please see below for some high-level stats :

Undergrad Institution: Honors program at small US public school
Majors: Pre-Med
Concentrations: Cognitive Neuroscience
GPA: 3.5/4.00

Type of Student: Domestic Male

GRE General Test:
Q: 166 (90%)
V: 161 (88%)
W: 5.0 (92%)
 

Research Experience: Endocrinology research experience, no papers. 
Awards/Honors/Recognitions: Full tuition scholarship to undergrad

For some more color, I am 3 years out of school. After school, I started/ran my own Math tutoring business and this past year, I've been interning at a large Hedge Fund in an operations role. I don't want to continue down the operations path, and I'd like to pivot into something significantly more quantitative. I've particularly interested in AI and Machine Learning after some exposure to their applications at work (I feel as though AI would be particularly interesting given my undergrad concentration). That said, I am not limited and am open to other interests as I dive deeper into math.

I've made a plan to go back to school and take 3 semesters of math/stat/prob/cs pre-requisites (essentially go back and get a second B.A. in Math) in order to prepare myself/apply to Ph.D. programs. That said, I'm worried that my lack of research will be a ding against my chances, as will my atypical background/path. I can openly admit that my brain wasn't fully switched on during undergrad and I essentially skated through with little regard for the future, smugly assuming that it will all work out in the end (shrink says I'm a "late bloomer"). Since then, I've developed a strong work ethic, identified my weaknesses, and am anticipating nothing but A's this time around as a Post-bac. Do I have chances at getting into top programs? What are some things that I can do to better my chances? I'd love to find some research opportunities but they're incredibly limited at the school I'm going to for post-bac (any advice on finding research opportunities in NYC?).

Open and thankful for any advice from this community. Thank you.

[orchidnora]

Hi there. First of all, is there a reason why you want to jump straight into a Ph.D.? Given your background, it might be more efficient to complete a Master's in CS/Stats/related fields. I have heard that some of these programs like having applicants from industry and other academic backgrounds. Then if you still want the Ph.D., you will be in a much better situation to apply for top programs. 

If you do prefer to get another Bachelor's, I can offer some advice on that. I don't think that being out of school for three years will significantly affect your chances in a negative way. But I think you will have to gain some research experience, like all applicants (traditional background or not), before being admitted to top Ph.D. programs. Moreover, it would be good to ensure that you like research, before committing to a five year Ph.D. program. It's not necessarily a problem that your school has limited research opportunities, since you can find research opportunities elsewhere (you can apply for summer NSF-REUs, government internships through NIH, U.S. Census Bureau, DOE labs, etc., although it might be difficult to get these positions until after more than 2 semesters of coursework...). But keep in mind that pursuing a Ph.D. typically means a willingness to move anywhere for the best research and career opportunities, so staying in NYC might restrict your options. Also if your school has limited ongoing research in general (meaning that there are few researching professors and graduate students) then you may consider attending a better school. Unfortunately, the rankings of an undergraduate institution do matter to some extent in graduate applications (Bs at MIT probably count more than As at an unknown school). 

[bayessays]

It's hard to imagine you could really know you want a PhD in a quantitative subject when you haven't really taken any math classes to know what that entails - but let's assume for a second that you do.  You might be able to get into a PhD program by just taking a few more classes.  If you took through Calc 3, linear algebra, and a probability class, you'd be a decent candidate for a lower tier but respectable biostatistics PhD.  There is almost zero chance of you getting into a math or CS PhD without literally completing almost an entire bachelor's degree.  My advice would be to take a couple of these classes and see what really interests you about them and evaluate how much time you're willing to put into that - but biostatistics is the clearest path open to you.

 

I would also second the advice that a master's (after the math pre reqs) might be a very good idea for you.

[Stat Assistant Professor]

Agreed with bayessays that you would have a decent shot for some Biostatistics programs if you complete the math prerequisites.

It's not impossible to switch to Math or Statistics for a PhD. But in your case, you would probably need to get a Masters degree in one of those areas first. I studied Econ/Political Science in undergrad and then got a Masters in Applied Math after completing several university math courses at local universities. Then I did a PhD in Statistics later, but I probably wouldn't have been able to get into a Statistics PhD program without having a Masters first. There are also several alumni from my PhD alma mater who did not study math/statistics as undergrad (they got BA's in Journalism, Economics, and History) but a few of them are now professors at R1's and one is working as a senior data scientist. But all of these guys needed to get a Masters first. For Mathematics, there are also a few accelerated post-bac programs for "non-traditional" applicants that are very good and that have good PhD placement.

I don't think your chances for a Computer Science PhD are very good without BOTH taking a number of advanced undergrad CS classes AND performing research in a CS professor's lab. Whereas research experience isn't strictly necessary for Statistics and Math, it is crucial for CS admissions, and you won't get in without any research experience. I have some friends in CS who took a gap year between finishing their BS in Computer Science and enrolling in a PhD program *just* to get more research experience by working as a CS Research Assistant and to prepare/size up their PhD applications. So if you want to go the CS route, keep that in mind that you'll need to do a bit more than just ace classes (research actually matters more than straight A's for CS PhD admissions).

 

Edited December 11, 2018 by Stat PhD Now Postdoc

[Stat Assistant Professor]

As for being out of school for three years, I don't think it matters too much. If you look at UC Berkeley's Statistics PhD Alumni and Current Students, you will see that there is one alumnus who obtained his B.A. in 1997 and completed his PhD in 2012, another who obtained his B.A. in 1985 but completed his PhD in 2013, and yet another current PhD student who completed his B.S. in 2002. Age and time spent out of school are not things that are considered relevant for PhD admissions for math or statistics.  However, having a solid math background is crucial for getting admitted.

Edited December 11, 2018 by Stat PhD Now Postdoc

[Spaceage]

Right off the bat, thank you. Your responses are already so helpful.

I know I'm overreaching in terms of timeline. I am by all means interested in Masters programs. 

I've mapped out the next 3-4 semesters to complete the math bachelor's degree. Again, responses from this community are incredibly helpful and I just have a few follow ups:

RE Python: I'd like to get a grip on python simultaneously. I've looked through both bootcamps and online classes. Are there any that you would personally recommend? Are there alternative ways you've learned (specific books that were better than others?) 

RE Research: After I get a few semesters under my belt and get a grip on python, what advice can you offer around finding research opportunities within Machine Learning specifically? 

RE Masters: Does a Masters in Stat typically better prepare people to work with Machine Learning? I'm looking through student profiles of M.A Stat programs and it seems like a number of them go on to PhD programs in machine learning. Which of the two would you say offers more flexibility/career prospects in industry?

 

 

[Stat Assistant Professor]

32 minutes ago, Spaceage said:

Right off the bat, thank you. Your responses are already so helpful.

I know I'm overreaching in terms of timeline. I am by all means interested in Masters programs. 

I've mapped out the next 3-4 semesters to complete the math bachelor's degree. Again, responses from this community are incredibly helpful and I just have a few follow ups:

RE Python: I'd like to get a grip on python simultaneously. I've looked through both bootcamps and online classes. Are there any that you would personally recommend? Are there alternative ways you've learned (specific books that were better than others?) 

RE Research: After I get a few semesters under my belt and get a grip on python, what advice can you offer around finding research opportunities within Machine Learning specifically? 

RE Masters: Does a Masters in Stat typically better prepare people to work with Machine Learning? I'm looking through student profiles of M.A Stat programs and it seems like a number of them go on to PhD programs in machine learning. Which of the two would you say offers more flexibility/career prospects in industry?

 

 

1)  You can learn the basics of Python just by doing Code Academy, edX, or one of the free online course providers. I have found personally, however, that the best way to learn programming, software packages, etc. is to use them regularly. I had limited experience with LaTeX and R before graduate studies, but I was able to pick up on them fairly quickly just by using them regularly.

2) For research opportunities, you will have to ask a professor if you can work as their Research Assistant.

3) A Masters in Statistics or a Masters in Computer Science gives you ample opportunities in industry. A PhD might be preferred for some of the bigger companies like Google and Microsoft, but you  can still get a decent job in Data Science/ML with just a Masters, sometimes with only a Bachelor's if you get the right experience. I have a friend who only has a Bachelor's in Biochemistry but now he's the Head of Data Science & Engineering for a health care startup. To the best of my knowledge, he got his first job as a data scientist out of college (no graduate degree), but he had to teach himself how to be a good "hacker" (Python, R, etc.) to get that job. After he had the relevant experience, his academic credentials didn't really matter.

I wouldn't say that getting a PhD is necessary to become a data scientist if you are a U.S. citizen (the bar is higher for non-citizens -- if they want to work in industry in the U.S., a PhD often makes it easier for them to get these jobs). You can get by with only a Masters and sometimes only a Bachelor's. Only do the PhD if you think you will like doing academic research.

[orchidnora]

2 hours ago, Spaceage said:

Right off the bat, thank you. Your responses are already so helpful.

I know I'm overreaching in terms of timeline. I am by all means interested in Masters programs. 

I've mapped out the next 3-4 semesters to complete the math bachelor's degree. Again, responses from this community are incredibly helpful and I just have a few follow ups:

RE Python: I'd like to get a grip on python simultaneously. I've looked through both bootcamps and online classes. Are there any that you would personally recommend? Are there alternative ways you've learned (specific books that were better than others?) 

RE Research: After I get a few semesters under my belt and get a grip on python, what advice can you offer around finding research opportunities within Machine Learning specifically? 

RE Masters: Does a Masters in Stat typically better prepare people to work with Machine Learning? I'm looking through student profiles of M.A Stat programs and it seems like a number of them go on to PhD programs in machine learning. Which of the two would you say offers more flexibility/career prospects in industry?

 

 

I agree with @Stat PhD Now Postdoc that there are great opportunities in industry for ML from a stats or a CS background. ML is essentially the intersection of math, statistics, and CS. Because of that, you should choose stats vs. CS based on what you like better. Typically, statisticians are more concerned than computer scientists with the mathematical theory underlying ML approaches, including checking assumptions, model validation, model interpretation, etc. In fact, many methods now under the umbrella of ML were originally developed by statisticians.

Keep in mind as well that there are Master's in Data Science programs out there. Some are cash cows (so be careful), but others seem reasonable. From what I can tell, they are typically not traditional research-based Master's degrees but rather sort of vocational, in that you learn exactly the skills to be directly applied in industry. A degree like that probably isn't enough to be the head of R&D at Google but can still offer lucrative opportunities in industry. Some of these Master's programs do not require an extensive quantitative background and even appreciate people from alternative backgrounds. 

Also when I asked about Master's degrees, I meant skipping a second Bachelor's and jumping straight into a Master's. Getting another Bachelor's involves repeating a lot of unnecessary coursework. I believe any Bachelor's should take more than 3-4 semesters. Are your previous credits counting toward it? Or do you mean that you plan to take 3-4 semesters of coursework but will not earn a Bachelor's from it? 

[Spaceage]

Again, thank you for taking the time to reply and offer your advice. 

@Stat PhD Now Postdoc - RE Masters: I meant to ask for your opinion of the preparation potential between a Masters in Statistics vs a Masters in Applied Math with regards to ML. 

@orchidnora - I've enrolled back at the same school that I received my first degree from, therefore, they start me on this degree with 90/120 credits, and I only have to complete the ~34 credits of Math/Stat to get the second BA in Math. I've built out a plan to take ~48 across 3 semesters with the option of maybe doing a 4th semester to throw some more graduate level courses in the mix. I'm scheduled to start back next month. Thoughts? My thinking is that I can take this Math BA (and Stat minor) with me wherever I go. Also, wouldn't that make me a more attractive candidate for top MA programs in Math/Stat? The concern I have is when will I learn Python and when will I do research.

[Stat Assistant Professor]

1 hour ago, Spaceage said:

Again, thank you for taking the time to reply and offer your advice. 

@Stat PhD Now Postdoc - RE Masters: I meant to ask for your opinion of the preparation potential between a Masters in Statistics vs a Masters in Applied Math with regards to ML. 

@orchidnora - I've enrolled back at the same school that I received my first degree from, therefore, they start me on this degree with 90/120 credits, and I only have to complete the ~34 credits of Math/Stat to get the second BA in Math. I've built out a plan to take ~48 across 3 semesters with the option of maybe doing a 4th semester to throw some more graduate level courses in the mix. I'm scheduled to start back next month. Thoughts? My thinking is that I can take this Math BA (and Stat minor) with me wherever I go. Also, wouldn't that make me a more attractive candidate for top MA programs in Math/Stat? The concern I have is when will I learn Python and when will I do research.

A Masters in Statistics would be more helpful than a Masters in Applied Math for machine learning. In either case, you could probably enroll in a Machine Learning class though (either taught by a Statistics dept or a Computer Science dept).

[orchidnora]

1 hour ago, Spaceage said:

Again, thank you for taking the time to reply and offer your advice. 

@Stat PhD Now Postdoc - RE Masters: I meant to ask for your opinion of the preparation potential between a Masters in Statistics vs a Masters in Applied Math with regards to ML. 

@orchidnora - I've enrolled back at the same school that I received my first degree from, therefore, they start me on this degree with 90/120 credits, and I only have to complete the ~34 credits of Math/Stat to get the second BA in Math. I've built out a plan to take ~48 across 3 semesters with the option of maybe doing a 4th semester to throw some more graduate level courses in the mix. I'm scheduled to start back next month. Thoughts? My thinking is that I can take this Math BA (and Stat minor) with me wherever I go. Also, wouldn't that make me a more attractive candidate for top MA programs in Math/Stat? The concern I have is when will I learn Python and when will I do research.

I see, that's an awesome deal! In that case, you may be competitive to apply straight for Ph.D. programs in Stats after these 34 credits, hinging on research experience of course. Do you know what language the stats courses at your school uses? For stats, R will be a big one to know. I wouldn't be too concerned about learning Python specifically, but by all means self-teach yourself in the meantime if you feel motivated. Most likely, you'll get strong in R, Python, or some other language of choice quite naturally through research projects. Now for research, since you say there aren't many research opportunities at your school, you should look elsewhere. I'm not sure if you have other commitments, but assuming not, then summers are prime time for doing research. As I mentioned before, typically it is difficult to get research without taking at least 4 semesters worth of math/stats coursework. However, if you emphasize your past research experience in science (to show that you know how to do research, even though it's a different field) + your independent projects, it may just work. If you can tie in your past Bachelor's degree and the research you've already done to your research interests in ML, then that could work even better for you. Google NSF REUs (there are some in machine learning; sometimes they're listed under stats, data science, math, etc...so just read the description) and national laboratories (Department of Energy, National Institute of Health, etc.). There are probably tons of other opportunities, but those are the ones I know. You could probably start applying for research in Fall 2019 and possibly get accepted to some research program for Summer 2020.  

[Spaceage]

Now what about which Bachelor's major you see as more useful/having more flexibility and longevity? Math, Statistics, or CS?

I'm going to take both CS and Math classes next semester and see what I'm more interested in/what comes more naturally, but just curious which you think is more useful. 

I figure that to be competitive for the top undergraduate research opportunities, I'd need to make sure I have some advanced coursework under my belt prior to applying. It seems like CS moves a bit slower in my school (prerequisites) and takes longer to get to the more advanced coursework. Though in the long run, it seems like there are more research opportunities for CS undergrads (based on my initial due diligence)?

[Stat Assistant Professor]

On 12/14/2018 at 11:05 AM, Spaceage said:

Now what about which Bachelor's major you see as more useful/having more flexibility and longevity? Math, Statistics, or CS?

I'm going to take both CS and Math classes next semester and see what I'm more interested in/what comes more naturally, but just curious which you think is more useful. 

I figure that to be competitive for the top undergraduate research opportunities, I'd need to make sure I have some advanced coursework under my belt prior to applying. It seems like CS moves a bit slower in my school (prerequisites) and takes longer to get to the more advanced coursework. Though in the long run, it seems like there are more research opportunities for CS undergrads (based on my initial due diligence)?

Math, statistics, and CS should all be okay as far as industry career prospects... even a lot of those who studied pure math (which doesn't have as many industry applications) can move into successful careers. A large number of the PhD graduates in Mathematics from the department where I got my Masters went on to become software engineers/developers at big companies like Google, Bloomberg, etc. For academia, the math job market is very competitive, with many PhD graduates needing to do multiple postdocs to get a job (it's not unheard of for math PhD graduates of schools like MIT and Harvard to be postdocs for 4-5 years before landing their first TT job). The academic job market is better in Statistics and CS, where one postdoc is more the norm.

I think it is a good idea for you to figure out what you like first and then approach professors for research opportunities. For math and statistics, the undergraduate research opportunities seem to be mainly summer REUs, whereas for CS, professors will let undergrads work as lab assistants. In either case, it is probably a good idea to get a flavor for research to see if you enjoy it enough to pursue a PhD. The PhD is a long grind and if your ultimate goal is to work in machine learning in a non-research setting, it may be better to get a Masters and relevant work experience.