trying to be somewhat creative with linear regressions

Hello friends,

I recently got stumped by a linear regression interview question I saw online, and that bruised my ego. So in between my other learnings, I decided to lightly brush up on my linear regression knowledge.

Last year, I used this pdf as my main reference. But recently I went through this pdf just to change it up a bit. I like both actually.

I do quite a bit of regressions, so I'm always on the hunt of methods/techniques to understand and fit better models. Stuffs like: what drives my predictions? Which data points are more valuable than others out-of-sample? Should I use rolling or expanding window? Why is my regression underperforming in XYZ?

I already wrote a bit about diagnosing ML models here - that was my attempt to be somewhat creative when diagnosing ML models, so this article is gonna be somewhat related to that, but focuses on linear regressions. Let's get it.

diagnosing ML models for noob modelers

all models are wrong, but some are hot

quantymacroquantymacro