A Second Course on Linear Models

Introduction

These notes are written for students on MAS61004. I have called this a second course because to get onto the MSc you will almost certainly have learned something about linear models already! These notes will still cover topics from the beginning, without assuming you have studied the content before, but we will work through some of the earlier topics fairly quickly.

Students who have taken the Graduate Certificate in Statistics to gain entry onto this MSc have studied linear models in the MAS5052 module. The Graduate Certificate lecture notes on linear models are available here. and you may find them helpful for revision.

I will assume some knowledge of likelihood and maximum likelihood estimation, but will revise this topic in the first lecture. Specifically, you should know how to construct a likelihood function for a statistical model, and understand the rationale for using maximum likelihood to estimate model parameters. This topic is also covered in the MAS5052 module: the notes are available here.

We will be using R for implementing linear modelling methods, and I will not assume you have used R before. You will be learning R in the EDA with R part of this module, which is taught in parallel with this part.

Acknowledgements

Some parts of the notes have been written by me, and others were written and modified by various colleagues who taught this content over the years: Eleanor Stillman, Jonathan Jordan, Kostas Triantafyllopoulos, Kevin Walters.