In econometrics, getting a deep understanding of concepts often requires learning some abstract linear algebra. For example, the mathematical properties of ordinary least squares are easier to understand once you know the projection theorem.
Economic research requires writing documents, and often writing code. How can economists avoid drowning in a mess of different versions? How can they keep everything organized, and keep track of the changes they made? How can coauthors work constructively on the same set of files without going crazy?
I wrote notes on demand estimation in industrial organization that survey a selected set of discrete choice techniques, from multinomial logit to Berry, Levinsohn, and Pakes (BLP)-type methods.