Tsilaeri

Duke and CMU belongs in a different tier than the rest. I don't know why anyone would consider the other options than these two. You can view data science as a collaboration of statistics and computer science. Even the traditional statistics programs are slowly integrating with CS, and it would be imperative for you to take couple of CS courses while in stats MS regardless of whether you'll go for employment or PhD. In that sense, CMU that has no.1 CS department will have stronger synergy affect every year. It already has strongest statistics program among your list. Duke is also strong choice, with top 10 statistics program, but since you don't like the locality I'd say CMU before Duke. Then, there's the rest. Columbia DS is only a year or two old, I wouldn't risk going there considering that you already see its limits. Moreover, you have no idea what their job placements will be like, and their stats department barely makes top 20. Look if they offer financial aid other than standard FASFA aid, such as TA/RA ship or fellowship/scholarship. If they don't, it is more likely that this program could be a cash cow for PhDs or its more research-focused degrees. The other three look like they're terminal degrees that would be more suitable for someone who would pursue PhD afterwards. In summary, CMU. No brainer. If not, choose Duke even if you don't like its locality.

Uh, I was gonna edit my post, but somehow hid it. Let me add more here. If you got an admission from top programs like those two, I assume you already have fair amount of statistics education from undergrad? I will assume that, correct me if I'm wrong. Consider a business undergrad going for an MBA. Total waste of time in terms of gaining additional knowledge, but people still do that for the value of networking. Yes, top tier MBA such as HBS will give you a splendid network that can far outweigh any form of academic knowledge. You are basically doing the same thing with this guy, but without the networking advantage. As Pleaaa put it nicely, the role you will get as a master's student in the industry will be mostly validation or programming - which will not be far different from the role that you can get after a BS. PhD belongs to another league. The best path you can take, in my opinion, is to defer UCLA admission, work in the industry for 1 or 2 years to have some practical experience, then go for a PhD. Data science is a very practical field. Having an industry experience will make a big difference in terms of your future choices and paths. (But... to be honest, in the end of the day, a smart guy like you can never go wrong, whichever path you take. Congratulations!)