Advice for nontraditional stats PhD prospect

[ihatemoney]

About me: White male US citizen, underrepresented minority affiliation

Undergrad: Top 3 uni, BS Math, ~3.9. One summer of statistical computing research and another in an industry research lab.

Linear algebra (Axler), calculus/ODEs, honors algebra+analysis, optimization, number theory, graph theory, full CS core + a good amount of CS theory. A-somethings everywhere but a B in analysis :'^(

Masters: Top 3 program, MS Stats, ~3.95

Regression theory, data mining, sampling, PhD probability, statistical learning, stochastics, RL, bit of biostats.

GRE, Recs: N/A (yet)

Interests: ML theory, graph mining, p >> n, manifolds

 

I've been working as a data scientist/applied scientist at a big famous Silicon Valley tech company since my MS (not being laid off lol), and after a couple years on the hamster wheel I'm wondering if I blew my chances at a great academic career. I wasn't aiming squarely at academia during undergrad/masters and didn't develop strong relationships with profs, do any great research or publish. I suppose I could ask stats PhD colleagues for letters of recommendation, but I'm not too confident I could get glowing letters from working academics considering both my unfocused past and that I've been out of the game for a while.

With that said, my questions:

• Should I try to hype up my industry experience or downplay it? Any general guidance on the value of an industry recommendation vs. an academic one?
• Am I likely to be perceived as a flight risk? (Is that even a concern in grad admissions?)
• Does it make any sense to spend some time working on my research portfolio before applying for a PhD? How best to go about this if so?
• Any pointers on which programs are realistic?

Thanks everyone.

[Stat Assistant Professor]

It shouldn't be a problem. There are a lot of Statistics PhD students who go back to get their PhD after years of working in industry. I can think of a few Stat PhD alumni from UC Berkeley who *began* their PhD 9-11 years AFTER they finished their Bachelor's degree (and so they finished their PhD 14+ years after graduating undergrad). If you have been out of school for awhile, you should mainly convey in your application why you are interested in obtaining a PhD and why you are motivated to do PhD research, so adcoms do not perceive you as a "flight risk."

If you have been working in industry for awhile, I don't think schools will expect you to have been doing much academic research. In general, the best letters of recommendation are those that can speak to your "research potential" and your mathematical aptitude, which is why letters of recommendation from academics are in general preferable.  I'm not sure how much weight a letter of recommendation from a boss/supervisor in a non-academic setting would carry unless it speaks specifically to your math abilities or your research 'potential' -- though I could see there being a few exceptions (e.g. if you've been working in a national lab or a research analyst at an economic organization or something like that).

I think your undergrad and grad performance are strong enough that you do not need to spend extra time working on your research portfolio. I would just apply for the PhD programs. Try to get some solid letters of recommendation, and I think you will see pretty good results at programs in the top 20... especially if you attended a top 3 program for both undergrad and grad!

Edited August 12, 2020 by Stat Assistant Professor

[bayessays]

All the important advice was covered above, but I was in your almost exact shoes and would like to ease your fears which are completely unfounded.

• Should I try to hype up my industry experience or downplay it? 

Working for a top company as a data scientist should be at least somewhat impressive and show good coding skills, and I know at my job we did real statistics stuff.  I think play up those things.  

• Any general guidance on the value of an industry recommendation vs. an academic one?

All else being equal, get some academic recommendations.  BUT, depending on the circumstance, I don't think the difference is huge.  My boss had a PhD in statistics from a top 10 university, published in JASA, and supervised work that was just as sophisticated as things in academia.  I don't view the letter as any different.  I'm sure there are snooty people who do, but not every letter needs to be from a famous professor.  If you have a boss (probably not just a co-worker?) with a PhD in statistics, that's a fine letter.  Since you got As in almost all your college classes, I'm sure you can find another 2 professors to write you decent letters - your profile is so good I don't think you need a lot of help from the letters department

• Am I likely to be perceived as a flight risk? (Is that even a concern in grad admissions?)

I think it's more likely that you'll be seen as less of a flight risk.  Most students coming into PhD statistics programs don't know a whole lot about the field.  You have an MS from a top school, got to experience what is widely regarded as the #1 job at one of the top companies in the world where you make tons of money, and you still like statistics enough to want to take a $100k/year paycut to go back to school.  What shows commitment more than that?  To put in perspective how little of an issue this is, I have almost the same path, but also dropped out of another graduate program after the MS, took years off of work where I didn't do anything I could put on a resume, and still got into programs I was very happy with 5 years after my MS.  I don't think anyone will bat an eye at your profile or even consider it non-traditional.

• Does it make any sense to spend some time working on my research portfolio before applying for a PhD? How best to go about this if so?

Nah.  Totally unnecessary. 

 

[ihatemoney]

Thanks a lot both, this is really helpful/reassuring. For some reason I'm simultaneously not surprised to read about the 35-year-old Cal PhD and can't imagine it being me. I think I'm guilty of comparing myself to my super focused peers who had a good idea of what they wanted to do a lot earlier than me (to the extent that I even do now). Better to wrestle with this now rather than several years deep into a doctorate, I suppose

[statsnow]

I agree with what Bayessays and Stat Asst Professor have to say.  I spent 4 summers at FAANG and the experience was well received.  It is  always a plus to have good coding skills.  Remember too that alot of the good people who get their phds go back into industry or some spend a bunch of time consulting with industry after their PHd.  The industry money temps alot of people. It also helps being a domestic URM .

At this point I dont think you need to spend more time working on your research portfolio.  I think alot of schools will take you