What’s Working (and Missing) from Analytics Graduate Degrees?
A young colleague of mine is about a third of the way through an M.S. in Analytics program from a prestigious university. Though there are the inevitable startup frustrations, she likes her studies a lot, noting she’s learning a ton about predictive modeling, data mining and new-to-her software like SPSS, SAS and R.
She feels she’s among the better technically-trained students and that her programming and data chops make the program an easier go than for her less-techie peers. If she could change one thing, it’d be to introduce a big data focus to the curriculum.
I don’t think it’s an overstatement to note there’s been a proliferation of Master’s in Analytics programs over the last couple of years. Some of the programs are cohort-based, wherein a fixed group of students starts in the Summer/Fall and progresses together in a common curriculum throughout the year or so of instruction. Many are housed in business schools but, increasingly, engineering departments are entering the fray as well. And there are online programs, such as the popular Master of Science in Predictive Analytics Online from Northwestern.
As I now look back at my own Master’s in Stats entry of two years ago, I can immediately see changes I’d propose. Though this curriculum is quite strong in applied statistics, given the opportunity, I’d sacrifice a bit of statistics for more computation, “requiring” just two of the below, rather than all four as mandated then:
- Forecasting and Time Series Methods
- Multivariate Statistical Methods
- Elements of Statistical Learning
- Statistics and Experimental Design
The two “saved” course slots would be replaced by required new ones in Visualization and Exploratory Data Analysis, and Computation with Big Data. The program would remain 12 courses over three or four quarters.
Indeed, some of the newer M.S. in Analytics curricula appears to be evolving to more of a data science focus than their purely statistical primogenitors. This is a welcome development, reflecting the needs of the market and the understanding that data science is a broader discipline than pure analytics.
I like Anand Rao’s depiction of the data science role as “25 percent business knowledge, 25 percent analytics expertise, 25 percent technological capabilities and 25 percent visualization.” I might even be more granular, suggesting the modern data scientist be versed in computation, mathematical optimization, database management, big data handling, visualization, statistics/informatics/machine learning, business and science (social, physical). That’s quite a load, but I think acknowledges the reality that big data and computation are now sine qua non for the conduct of analytics.
The challenge for academia as analytics/data science matures is to offer an integrated curriculum that crosses disciplines – from statistics to computer science to engineering to business and the social sciences. As data scientist and Berkeley Ph.D. Cathy O’Neill noted at Strata Hadoop World, “academia is not currently aligned with the needs of data science, where the biggest challenges are determining the right questions and working with messy data … academia must work through its interdisciplinary politics and aggressively recruit industry experts to its faculties to become truly relevant for data science.”
I agree, but see progress. Among the many attractive analytics programs out there today, three especially catch my eye. The first is the M.S. in Analytics from the Institute for Advanced Analytics at North Carolina State University. This 10-month, cohort-based curriculum, started in 2007, has perhaps the most established track record of all such programs, with an enviable 90% placement record for graduates. It’s also shown the ability to adapt to changes in the analytics landscape. The IAA’s strong relationship with statistical software juggernaut SAS is touted as a major benefit by many.
The new 15-month Master of Science in Analytics from Northwestern’s McCormick School of Engineering looks to be a second generation program, combining “mathematical and statistical study with instruction in advanced computational and data analysis.” New student cohorts start in the Fall quarter and continue through the Fall of the following year, though they don’t necessarily take the same courses. I particularly like required classes in Data Visualization and Analytics for Big Data, as well as the Capstone Design Internship.
For those who wish to have a lot of say in their curricula, a program like Harvard’s new SM in Computational Science and Engineering is certainly worth a look. With only four required courses, the CSE will “provide rigorous training in the mathematical and computing foundations … Complementing the foundational coursework will be independent research projects and elective courses focusing on the application of computation to one or more domains.” This looks great to me – a roll-your-own curriculum built on a foundation of math/stats and computation and providing access to offerings from Harvard’s world-class research departments. I’d spend my elective time in the social sciences.
All told, pretty good stuff. Expect the continued growth of graduate programs in analytics which will become more data science-like. Expect also new programs built from the ground up for data science needs. Good news for business analytics and big data!