Week 3: Re-sampling and regularisation

Main reference

ISLR 5.1, 5.2, 6.2, 6.4

What you will learn this week

  • Common re-sampling methods: bootstrap, cross-validation, permutation, simulation.
  • Cross-validation for checking generalisability of model fit, parameter tuning, variable selection.
  • Bootstrapping for understanding variance of parameter estimates.
  • Permutation to understand significance of associations between variables, and variable importance.
  • Simulation can be used to assess what might happen with samples from known distributions.
  • What can go wrong in high-d, and how to adjust using regularisation methods.

Lecture slides

Tutorial instructions

Instructions:

Assignments

Assignments