I worked as a data scientist for 3 years then left my company to get a MS in statistics. I'm currently finishing up a program at an ivy league university. Honestly, even as I'm finishing up my degree I agree with skadamat.
I learned way more about how to program (in many languages), build models, apply machine learning techniques, etc etc, on the job in 3 years than I did in my undergrad or thus far during my graduate degree. The only way to really learn data science is to do it and the fact of the matter is that school just doesn't provide enough doing. Jump on a few Kaggle comps and survey the multiple lists of online resources. If you're looking into more focused topics, Coursera is a great resource.
These online incubator programs seem to me to be sort of joke honestly. They are all basically 6-11 week survey courses of CS, Statistics, NLP, HIVE, PIG, Python programming, etc. Long enough to learn the very very basics (in some cases just the buzz words) but not long enough to learn anything really useful. Plus they are expensive in some cases.
I'm not saying that a graduate degree isn't useful. I've learned a ton about theoretical statistics and CS and it has certainly provided me with a great new prospective and augments what I learned on the job. But, if I hadn't had that prior work experience I would still be ill equipped to jump into a Data Science job.
