Study CS or Statistics to work in Data Science

[tonym9428]

I was just wondering what would be a better undergraduate degree to prepare for work in "data science".

Statistics, computer science, applied maths, physics, etc.

I asked this question on flowing data also:

http://forums.flowingdata.com/topic/data-science

[newms]

I think Nathan Yau gave a pretty good answer in that thread. That said though, there's a lot of overlap between computer science and statistics, particularly in machine learning, where many statisticians are studying machine learning and many machine learning researchers are studying statistics.

[tonym9428]

In my other post, I mentioned that I recently finished my MA and was considering returning to school in order to take some math courses so that I can apply for graduate programs in statistics.

Do you think it would be beneficial to just head back and work on a second bachelor degree in computer science?

While I know a few programming languages, I'm not interested in programming as a career. I just want to work with statistical modeling and data visualization of large data sets. I'm also intrigued by some of these computational math and computational science programs.

[newms]

In my other post, I mentioned that I recently finished my MA and was considering returning to school in order to take some math courses so that I can apply for graduate programs in statistics.

Do you think it would be beneficial to just head back and work on a second bachelor degree in computer science?

While I know a few programming languages, I'm not interested in programming as a career. I just want to work with statistical modeling and data visualization of large data sets. I'm also intrigued by some of these computational math and computational science programs.

What's your BA in? Do you have a background in maths? If not it might be a good idea to go back to get a bachelors in either computer science (with some statistics classes) or math/statistics if you want to get into data science as a profession. If you already have a bachelors in math or a related field then I wouldn't think that you needed a second bachelors.

[tonym9428]

What's your BA in? Do you have a background in maths? If not it might be a good idea to go back to get a bachelors in either computer science (with some statistics classes) or math/statistics if you want to get into data science as a profession. If you already have a bachelors in math or a related field then I wouldn't think that you needed a second bachelors.

I have a BA and MA in political science.

The only math that I took as an undergrad was calc 1, calc 2, and elementary linear algebra...so yeah, I need to take a lot of math courses.

In grad school, I had stats courses on linear regression, logistic regression, statistical computing with R, multilevel modeling, etc...but the mathematical theory was often ignored. For example, in one of my classes, we used John Fox's "Applied Regression Analysis and Generalized Linear Models", which is pretty dense, but we rarely touched on the mathematical theory behind dummy variables, interaction terms, statistical modeling, etc.

Unfortunately, I'll be pursuing my second Bacheor's at a school which doesn't have a stats department (Univ of Kansas). They have a math department and supposedly offer some mathematical stats courses, but it doesn't seem to be a priority.

[kowtown]

While I know a few programming languages, I'm not interested in programming as a career. I just want to work with statistical modeling and data visualization of large data sets. I'm also intrigued by some of these computational math and computational science programs.

I disagree with nathany's post. I doubt you need the dozen or so computer science courses that constitute a major. You might want to take a handful of courses including introductory programming (if you're not proficient in C or C++), data structures, algorithms, etc. You've had more than enough applied stat courses. You probably don't need another major at all, never mind a degree, since admissions committees only care about what courses you took and how well you did in them. (Rarely does one return to undergrad after obtaining an advanced degree, but if you have the time and can get some sort of financial support for pursuing a second bachelor's, you may want to consider it.) What you need, as you seem to already know, is a good foundation in mathematics and a year sequence in mathematical statistics, which I'm sure you can get at U. Kansas.

I originally majored in computer science. I didn't finish the degree at the time but I was just an elective or two from completing the major. I left with about the same math background you have now: two semesters of calculus and a basic linear algebra course (as well as discrete math). I returned to school a number of years later and completed an applied math major in two years, taking multivariate calc, two semesters of advanced calculus (real analysis), numerical linear algebra (to both brush up on linear algebra and introduce myself to numerical methods), a year sequence in applied statistics, and a year sequence in mathematical statistics. Of course I had some additional courses, like differential equations and stochastic processes, but these were more to just reinforce concepts in calculus and probability theory, and I would consider these optional. It may be premature to say so, but I feel this background prepared me very well for my first semester in a statistics PhD program. I regret not taking a more advanced theoretical linear algebra course, but I won't know until next semester how much that would have helped. I would recommend you take one since large data sets and visualization will each require vector and matrix manipulation and good grounding in associated theory.

Of course one key token in the above paragraph is "PhD." You may not need to take all of the more theoretical courses like real analysis, mathematical statistics, or theoretical linear algebra, but taking at least a subset will make you a better thinker and improve your chances of getting into a good program. The program that you eventually want to get into will dictate what courses you actually take. I think I've said this before, do look at graduate programs you're interested in and see what they require and recommend, and do talk to professors and professionals in the field. Academics love to give advice! Also, once you begin your studies, you yourself will have a better idea of what you're doing and want to do. Best of luck.

Edited October 15, 2010 by kowtown

[weninger]

I am a PhD candidate in Data Sciences at a Top 5 CS school. Most of the researchers here write their code in Matlab or C++/Java.

For our purposes there is no need for software engineering, language semantics, arch design, etc. Math Theory and stats is important, but the Bachelors CS stuff isn't necessary so long as you have a cursory understanding of programming.

While I didn't read everyones post before mine, I would tend to say that you can apply to Data Sciences PhD without a CS degree. My office mate has her BS in Math, and she does fine.

Good luck to you!