Statistical Learning in Practice, Lent 2018

Statistical learning is the process of using data to guide the construction and selection of models, which are then used to predict future outcomes. In this course, which consist of 12 lectures and 12 practical classes, we will examine some of the most successful and widely used statistical methodologies in modern applications. The practical classes will deal with an introduction to R, exploratory data analysis and the implementation of the statistical methods discussed in the lectures. We aim to cover a selection of the following topics:

• Generalised linear models
• Model selection and regularisation
• Mixed effect models and quasi-likelihood methods
• Linear discriminant analysis and support vector machines
• Introduction to neural networks
• Time series and spatial statistics

Pre-requisites: elementary probability theory, maximum likelihood estimation, hypothesis tests and confidence intervals, linear models. Previous experience with R is helpful but not essential.

Resources

The current version of the lecture notes can be found here. It will be updated after each lecture. Practical texts and solutions can be found in the timetable below. Datasets used in the practicals can be found here.

We will use the statistical programming language R in this course. It is recommended that you also use RStudio , an integrated development environment (IDE) for R. Both R and RStudio have been pre-installed on all machines in the CATAM Room (GL.04). You can download them for your own computer from http://cran.r-project.org/ (R programming language) and http://www.rstudio.com/ (RStudio).

We will be moving to MR5 for both lecture and practicals due to high demand for the course. Please bring your own laptop with R and RStudio preinstalled from Practical 2 onwards. If MR5 becomes too crowded, we will use the CATAM room for overflow, where desktop machines are available.

Timetable