two cultures

[hardin47]

Nick keeps pushing us to think about expanding IMS to Stat 2. I don't know. But, one of the pieces I find most interesting is model building with two cultures and how the code makes the process easier or harder.

Model as variables: if you really care about the variable interpretation, then you want to hone in on which coefficients should be kept in the model. add1() and drop1() are super useful for that.

Model as prediction: if you really care about the model as a way to make predictions, then you mostly care about how the model does in the wild. And tidymodels is incredibly powerful for working with data in principled ways so as to get accurate assessments.

There could be some good stuff in here. But man, there would be a lot of other stuff to fill in the gaps, and it feels like a GINORMOUS task to do it well.

[mine-cetinkaya-rundel]

To me as well this sounds both super intellectually engaging, and like a huge task. I can see this content being in IMS2, where IMS1 stays as this book (or maybe even loses a chapter or two at the end) and IMS2 starts where it leaves off and goes deeper. Def a v3 convo, but a great one to have 😄