| I'm a graduate student struggling with an important decision as to whether I want to pursuit phd/research work after my master's in economics, specifically econometrics & data analysis in association with consumer behavior and derived demand. My question is that for years while learning very sophisticated and practical methods of analysis, I had to force myself to ignore a lot of the machine learning frontier. I only recently started learning these practices and see definitely the source of mutual benefit. I had to ignore it because from what I saw, people were solving similar questions with almost zero of the same sophisticated methods. My question for someone with maybe experience with machine learning and high level methodologies of statistical forecasting is: is the mathematical practice of these methods going away with automation and machine learning? Or similarly is mathematics becoming less and less useful in practice? If that's the case, then I'm wasting my time at school learning more and more sophisticated layers of mathematics to incorporate into analysis. When I often see things on Kaggle for instance, people solving complex problems with almost zero mathematical modeling. Of course this was at first frightening, but after delving deeper there is reason for the models to exist although some people may disagree. I'm less concerned with this as I am with future obsolescence of the high level mathematical econometrics that's currently taking up so much of my time. Often feels as if these machine learning frontiers are more important. I of course would like to learn both but I won't be able to do that till I finished. My work now involves almost exclusively with R. Some work with SAS and Matlab but that's pretty much it. So after years of practice, and I've yet to delve into any deep learning, it's very nerve racking. Any help or suggestions would be very much appreciated. |