Appendix D — Further reading
Optional. Nothing here is required, nothing here is examinable, and the course is complete without any of it. This is the list for the reader who finishes a chapter and wants the longer treatment — or who meets, in their own work, one of the problems the conclusion admits this book does not cover.
Where a book is legally free online, that is the link given. Where it is not, the publisher’s page is, so you can find it in a library rather than guess at an edition.
D.1 The two nearest neighbours of this book
An Introduction to Statistical Learning — James, Witten, Hastie, Tibshirani. statlearning.com · free PDF
The closest thing to a companion volume. Covers chapters 2 through 8 of this book at a similar level with more examples and less econometrics, and its treatment of the bias–variance trade-off and of regularisation is the standard one. Editions exist for both R and Python. If you read one book on this list, read this.
Data Science for International Business with R — Warin. warin.ca/ds4ibr
The R volume of this course. Exploratory analysis, regression and panel data, logistic models, unsupervised learning and structural equation modelling, in the other language of the field. Reading the two side by side is the fastest way to see which difficulties belong to the method and which belong to the tooling.
D.2 Going deeper on the methods
The Elements of Statistical Learning — Hastie, Tibshirani, Friedman. hastie.su.domains/ElemStatLearn · free PDF
The graduate version of ISLR, and the reference for chapters 5, 7 and 8. Harder, complete, and the place to look when you want the derivation rather than the result. Its chapter on model assessment is the fuller treatment of what chapter 8 demonstrates by simulation.
Categorical Data Analysis — Agresti. doi.org/10.1002/0471249688
Chapter 4 in book form: logistic regression, ordinal and multinomial models, and the exact-inference machinery for the small samples where the asymptotics this book relies on stop being safe.
Econometric Analysis of Panel Data — Baltagi. link.springer.com
Chapter 6 at length. Fixed and random effects, the Hausman and Mundlak formulations, dynamic panels and the Nickell bias, and the estimators that address it.
Python Data Science Handbook — VanderPlas. jakevdp.github.io · free online
Not a statistics book. It is the reference for the tools this course uses — NumPy, pandas, matplotlib, scikit-learn — and the right place to go when the obstacle is the library rather than the method.
D.3 Causal inference, where this book stops
Chapters 10 and 11 are an introduction to a field with an unusually good literature, most of it free.
Causal Inference: The Mixtape — Cunningham. mixtape.scunning.com · free online
Potential outcomes, matching, difference-in-differences, regression discontinuity and instrumental variables, with code and an unusually direct voice. The natural next book after chapter 11.
The Effect — Huntington-Klein. theeffectbook.net · free online
Starts further back than the Mixtape and spends longer on what a research design is before reaching estimators. The chapters on causal diagrams are the clearest short treatment of the good-and-bad-controls problem chapter 10 raises.
Mostly Harmless Econometrics — Angrist and Pischke. press.princeton.edu
The book that made design-based inference the standard in applied economics. Terse, opinionated, and assumes more econometrics than this course does. Mastering ’Metrics by the same authors (Princeton) is the gentler version, and a better first read.
Causal Inference for Statistics, Social, and Biomedical Sciences — Imbens and Rubin. cambridge.org
The formal treatment of the potential-outcomes framework chapter 10 introduces, by the people who built it. Where to go when you need the assumptions stated precisely rather than described.
D.4 The gaps the conclusion admits
Statistical Rethinking — McElreath. xcelab.net/rm
The Bayesian course this book is not. Rebuilds inference from probability rather than from sampling distributions, and its treatment of multilevel models is the one to read after chapter 6. Lectures are free on the author’s site.
Regression and Other Stories — Gelman, Hill and Vehtari. avehtari.github.io/ROS-Examples · free PDF
Sits between this book and McElreath’s. Applied, Bayesian by default, and better than anything else on the list at the part this course keeps insisting on: what a model does and does not license you to say.