Spatial Temporal Data: PhD vs. Industry

[Space+TimeStats]

I am interested in pursuing a PhD in Statistics to specifically research Spatio-Temporal Data or working in industry working with spatial data (finishing my first year at a Masters in Statistics at Top 5 Stat program & interning at top GIS company this Summer).

I am trying to see if anyone can share their experiences in this field (Spatial Temporal Modeling) about going the PhD route or straight to industry (government, tech, or other)? 

Also,  after searching the web, I came up with these prof's/ uni's as a starting point for reading literature: Chris Wikle (U. of Missouri), Debashis Mondal (Oregon State), and Mevin Hooten (Colorado State). Any others to mention as I begin reading papers this Summer to better compare/ contrast the two routes?

Thanks! Any and all advice is appreciated!

[Stat Assistant Professor]

I don't have firsthand experience with this. But I co-authored a paper on spatio-temporal statistics with a friend who got a Masters in Statistics and then earned his PhD in Agricultural & Biological Engineering with a concentration on spatio-temporal statistics. He is doing very well now, working as a Senior AI Quantitative Researcher at the Climate Corporation (he also interned there as a geospatial statistician the summer before he went to go work there full-time). Since he had a Masters in Stat already, he was able to complete his PhD in four years. I'm sure that you would be well-positioned to do so as well if you can bypass having to repeat Masters-level courses.

I would look at the job requirements for jobs that interest you. If most of the job descriptions say "PhD preferred," then it might be worthwhile to go straight to the PhD. If you decide to go into industry in your final years, you can always do another internship the summer before you graduate. I would also assess how important it is for you personally to earn a PhD. If this is very important to you, you should go straight into it from your Masters. If you're on the fence, it might be better to go make some money.

[Space+TimeStats]

13 minutes ago, Stat Postdoc Soon Faculty said:

I don't have firsthand experience with this. But I co-authored a paper on spatio-temporal statistics with a friend who got a Masters in Statistics and then earned his PhD in Agricultural & Biological Engineering with a concentration on spatio-temporal statistics. He is doing very well now, working as a Senior AI Quantitative Researcher at the Climate Corporation (he also interned there as a geospatial statistician the summer before he went to go work there full-time). Since he had a Masters in Stat already, he was able to complete his PhD in four years. I'm sure that you would be well-positioned to do so as well if you can bypass having to repeat Masters-level courses.

I would look at the job requirements for jobs that interest you. If most of the job descriptions say "PhD preferred," then it might be worthwhile to go straight to the PhD. If you decide to go into industry in your final years, you can always do another internship the summer before you graduate. I would also assess how important it is for you personally to earn a PhD. If this is very important to you, you should go straight into it from your Masters. If you're on the fence, it might be better to go make some money.

I have (naively) not considered programs other than Statistics, so that is a helpful comment.  

Just making sure I understand the part on finishing in 4 years since I am still learning; PhD sequence is roughly (1) Take Masters Courses (2) Pass qualifying exams (3) Enter Candidacy - Research/ take PhD courses (4) Comprehensive Exams (5) Thesis defense. Your saying that with adequate preparation I can validate (1), (2) thus reducing the entire length of the program (like your friend did)? My guess is that this is program specific but speaking generally probably best to load up on electives that prepare me for the qualifying exam(s) (Applied Stats, Stat Theory, complimentary math classes - real analysis, linear algebra, etc.)?

 

[bayessays]

Course requirements differ from program to program. If you come in with a master's degree, you'll often be able to skip the first year Casella/Berger courses and maybe the intro to linear regression. But you'll still have to take qualifying exams if the program has them, so a lot of people end up taking the courses anyways just to make sure they are prepared for the quals, because the courses differ slightly from program to program.  If you already have a master's from somewhere else, you might be able to to cut off up to a year of your PhD, but it will vary from program to program. Very rarely can you bypass more, and often it doesn't save you much if any time at all.

[Stat Assistant Professor]

54 minutes ago, Space+TimeStats said:

I have (naively) not considered programs other than Statistics, so that is a helpful comment.  

Just making sure I understand the part on finishing in 4 years since I am still learning; PhD sequence is roughly (1) Take Masters Courses (2) Pass qualifying exams (3) Enter Candidacy - Research/ take PhD courses (4) Comprehensive Exams (5) Thesis defense. Your saying that with adequate preparation I can validate (1), (2) thus reducing the entire length of the program (like your friend did)? My guess is that this is program specific but speaking generally probably best to load up on electives that prepare me for the qualifying exam(s) (Applied Stats, Stat Theory, complimentary math classes - real analysis, linear algebra, etc.)?

 

Like @bayessays mentioned, it will probably be program-specific. But I do know that a lot of programs, you have the option of taking the qualifying exam upon arrival if you already have a Masters degree. This is true for University of Michigan, NCSU, and University of Florida, for instance (I know students at these programs who already had Masters degrees and took the first-year exam before the start of their first semester in the PhD program -- if they passed the qualifier, they were allowed to skip the first-year classes). I think this is likely the case at a lot of other schools as well. However, you probably can't skip all the coursework entirely, so you would still need to take the second-year classes and some electives.   

Some very elite schools (like Stanford and UPenn) start off their first-year PhD students in measure-theoretic probability and advanced statistical inference right away, so they don't even start with Casella/Berger. However, the students they admit are typically fairly advanced and have already taken Casella/Berger or measure theory before entering.  But the majority of programs would start off with Casella/Berger and not introduce measure theory until the second year.

Edited May 3, 2020 by Stat Postdoc Soon Faculty

[Space+TimeStats]

Thank you @bayessays and @Stat Postdoc Soon Faculty, both very helpful! It is nice to have realistic expectations prior to applying but it sounds like the best thing to do is simply confirm with the programs that I actually get admitted to (if that happens lol) and then make a decision.

 Not to completely change subject, but I noticed schools in UK have much shorter program lengths (3-4 years) and seem to expect candidates already having masters degrees. It seems like this assumption automatically shaves a year off the program (whereas in the US it seems to vary), I wonder if there is a tradeoff associated with this benefit?

[Stat Assistant Professor]

1 hour ago, Space+TimeStats said:

Thank you @bayessays and @Stat Postdoc Soon Faculty, both very helpful! It is nice to have realistic expectations prior to applying but it sounds like the best thing to do is simply confirm with the programs that I actually get admitted to (if that happens lol) and then make a decision.

 Not to completely change subject, but I noticed schools in UK have much shorter program lengths (3-4 years) and seem to expect candidates already having masters degrees. It seems like this assumption automatically shaves a year off the program (whereas in the US it seems to vary), I wonder if there is a tradeoff associated with this benefit?

As I understand it, UK programs have a strict deadline for how long a PhD student can receive guaranteed funded (four years max -- if they take more than four years, then they have to self-fund). Even in the U.S.A., I think it would be perfectly reasonable for someone with a previous Masters who skips the first-year courses to finish a PhD in four years. Students at Duke routinely finish in four years. 

This also depends on your goals, though. If your only goal is to finish and get an industry/non-academic job, then you may not have to worry that much about publishing and can focus on just getting the dissertation done. Then it would be reasonable to finish in four years. If you are interested in an academic career, then it may not be advisable to to finish that quickly, because you need publications to be competitive in the academic job market and publishing can take a long time. Some students will stay in their PhD longer just to get more publications on their CV. In my PhD program, one of my cohort classmates could have easily graduated in 3-4 years, but he stayed for the whole five years so he could get an Annals of Statistics paper on his CV. He then spent his fifth year on the job market and ultimately got a sweet TT job at University of Minnesota. 

The students I know who already had Masters degrees but took five years either: a) repeated the Masters-level coursework, or b) wanted to pursue academia and opted to spend a fifth year in grad school so they could go on the job market in their fifth year. 

[StatsG0d]

As always, @bayessays and @Stat Postdoc Soon Faculty gave some very good points and advice.

I think many programs are starting to shy away from master's coursework -> PhD coursework -> electives. I believe PSU and TAMU are two such programs where the qualifying exam is taken after the first year, and the first year curriculum includes measure theoretic probability. If you have the requisite background, they might let you take the qualifying exam as soon as you arrive.

Regarding UK programs (programmes?), based on LSE and UCL, it does not really seem like funding is the norm, whereas I feel like in the US, no funding is the exception.

[Space+TimeStats]

Wow, great information, really allows me to narrow down my course selection for next year to best accomplish my goals. So glad that I joined GradCafe and posted the question!

[player-tracking-data]

Hello, I have the same research interest, so I PM'ed you!