Textbooks are great.
Although I have some training in statistics and machine learning, for the most part I’m completely self-taught. A big part of that teaching has been to buy a great textbook on a given subject and to follow along until I really understood the subject. Coursera, YouTube, Stack Overflow, tutorials on the internet, etc are great, but textbooks are excellent too, and I think they are underutilized. Especially since there is a mental categorization of “that’s for when you’re in school” that people fall into.
It was at business school that I started doing some serious data science that my addiction to buying textbooks started. Right now I’m working through Design and Analysis of Experiments with R since Gradient is doing more (and more complex) survey-based experiments (I think at this very moment we are running four discrete choice experiments for clients)
I’ll be honest, this one is a bit of a slog. But getting experimental design right is crucial for successful projects so I’m working through it; I’ve had a few a ha moments already — like why they’re called “orthogonal” designs.
Most of the textbooks I use are written to work with R, although not all. Some of the textbooks that go into my hall of fame are:
- Applied Predictive Modeling (survey of lots of statistical and machine learning methods)
- Bayesian Networks
- Generalized Additive Models
- Data Analysis Using Regression and Multilevel/Hierarchical Models