|Fees||Look up fees|
Supervised statistical learning is based on the widely used linear models that model a response as a linear combination of explanatory variables. Initially this subject develops an elegant unified theory for a quantitative response that includes the estimation of model parameters, hypothesis testing using analysis of variance, model selection, diagnostics on model assumptions, and prediction. Some classification methods for qualitative responses are then developed. This subject then considers computational techniques, including the EM algorithm. Bayes methods and Monte-Carlo methods are considered. The subject concludes by considering some unsupervised learning techniques.
Intended learning outcomes
On completion of this subject students should be able to:
- Understand the underlying statistical theory of linear models and the limitations of such models.
- Fit linear models to data using a standard statistical computing package and interpret the results.
- Be able to predict (or classify) qualitative responses.
- Understand the underlying theory of Bayesian statistics.
- Understand the underlying theory and be able to apply the EM algorithm in simple settings.
- Be able to use a computer package to perform statistical computing and data analysis
Last updated: 29 October 2019