Spatial Career Guide – Spatial Statistician
This is my fifth post in the Spatial Career series. In previous posts I’ve written about how to prepare for a career as a GIS Software Developer, as a Geospatial Analyst, as a Cartography/Visualization Specialist, and as a Geographic Information Scientist. In this post I will describe how to prepare for a career as a Spatial Statistician. This career path is similar to the GIScience path in that it’s as critical for you to want to be a statistician as it is for you to want to work with geographical data. So this guide might be reasonably helpful for anyone interested in a career as a statistician (but you’ll want to substitute specifics from the relevant application area and/or data universe, i.e., instead of geographic/spatial you’ll have to find out about the nuances in health, financial, etc). So, here are 5 steps that should put you in good position for a career as a spatial statistician.
Step 1. Get an advanced degree. I would say for this path you really should get at least a Master’s degree and a Ph.D wouldn’t hurt. In fact, one nice thing about the statistician route is that a Ph.D can open up options in private industry as well as in government and academics. Businesses aren’t afraid to hire a Ph.D type for a quantitative position whereas in other roles they will shy away from a Ph.D. So this could be a good option for someone who has a PhD but no longer wants to pursue an academic career. Also, this is another career path where you might consider at least one degree in a field other than geography. Unfortunately the market place is such that people won’t value your degree in geography in terms of the quantitative preparation you would’ve received in school. So I think you have to round out your training, ideally with a degree in math or statistics. There are several combinations that would work. A bachelor’s in math or statistics with a Master’s in Geography would work. A bachelors’ in geography and a Master’s in statistics would probably be better (you would want to minor in math or at least take a full set of courses in Calculus and Linear Algebra as an undergrad). Or, you might consider a joint Master’s in both geography and statistics. This would require some administrative effort and more time but it would be a nice combination and open a lot of doors.
Step 2. Become an expert in spatial analysis methods. Unfortunately, it’s fairly difficult to acquire this type of training. Most geography programs have a few offerings and one or two people on the faculty with appropriate expertise but most programs don’t have enough depth in this area, at least not in my opinion. There are a handful of programs that come to mind where you’ll get all the spatial analysis training you could want and more but they are few and far between. Top on my list is Ohio State because you might get a chance to study with Noel Cressie, who is pretty much the godfather of spatial statistics. He is in the Statistics department where they offer a special program in Spatial/Environmental Statistics but I suspect there are solid ties to Ohio State’s outstanding geography department. Other key programs (and key people) to consider would include UC Santa Barbara (Michael Goodchild), Arizona State (Luc Anselin) and SUNY Buffalo (Peter Rogerson). Each of these programs has a number of talented faculty and an Associate Professor might make a better advisor since they’re a bit less likely to be globe-trotting than the super stars. If you’re Canadian or interested in studying in Canada, take a look at Ryerson University – they offer a good applied graduate program in Spatial Analysis. If you’re in the UK, check out the University of Leeds. There are probably others that I’m leaving out. Feel free to suggest others in the comments section below.
Step 3. Learn the most popular statistical software programs. I would try to obtain some exposure to all of the top statistical software tools including SAS, SPSS and R. The more the merrier. Add Stata and Minitab and others if you can and you might check out free trials of newer technologies like Alteryx. You’ll probably have one go-to tool that you like best but you don’t want to forfeit a job opportunity just because you aren’t familiar with the software the target organization has already licensed. Spend enough time to feel comfortable doing basic analysis in a few different statistical computing environments. You need to be able to get past the gatekeepers in Human Resources who might toss out your resume if it doesn’t have the right buzzwords listed. Also, learn how to automate tasks in these computing environments. Let a scripting language allow you to reproduce results at a click of the button. You don’t need to take hard-core computer sciences classes, not that it would hurt, but some basic programming concepts will certainly help.
Step 4. Learn to communicate effectively. Read bullets 3 and 4 in the GIS Software Developer post and read the Cartography/Visualization specialist post. Being a statistics nerd is cool but don’t be a statistics nerd who can’t communicate with real people. Try to explain things so that a sophomore in high school would be able to understand what you’re saying. And, try to be really good at data visualization. The ability to communicate through visualization, narrative and anecdote is what will set you apart from just the regular statistician who writes equations on the white board when asked to present something to a non-mathematical audience. Instead, help your organization and your clients visualize their data in a way that leads to better understanding and collaboration. For more on this read my post on Geospatial Visualization in Business.
Step 5. Get some applied experience. You will learn a ton of great stuff from your statistics professors and other faculty. They will be able to teach you all about theory and methods. But, when it comes to applying these methods you will want to find a practitioner who has been forced to deliver answers to difficult questions with messy data under tight time constraints (think days or weeks instead of months or years). The rare professor will be good at both but don’t count on such luck. Instead, plan to volunteer to do data entry or whatever you have to do in exchange for the opportunity to interact and learn from statisticians who work on applied problems. Building statistical models that work well doesn’t involve particularly elegant mathematics, at least not in my experience. You have to be able to use a bit of “art” to get some models to work. This doesn’t make the model less useful, it will do the opposite if you do it well, but it won’t be something you can publish in an academic journal and most of your professors won’t like it. You have my permission to ignore them (unless you need their signature on your dissertation or want tenure). Learning how to do a bit of brute force modeling is a great skill to add to your tool belt. It just might allow you to actually suggest a real solution to a thorny problem. And that’s what brings in the big bucks and for some, if you’re like me, the big fun.
The opportunities in statistics/analytics are huge. If you haven’t heard the term “Big Data” by now you’ve been hiding under a rock. Use of the term Big Data may be a passing phase but a world in which all organizations are trying to make sense of enormous volumes of data has just begun with no end in sight.