Lecturer(s): Dr Jeremie Houssineau
Prerequisite(s): ST218/219 Mathematical Statistics A&B
This module runs in Term 2.
Availability: Only available to students who have not taken ST337
Commitment: 3 lectures per week
Rationale: Forecasting is a vital prerequisite to decision making. This course is concerned with the theory and practice of short-term forecasting, using both data and subjective information. The course focuses on Dynamic Linear Models (DLM). DLM’s are a class of Bayesian Forecasting Models which generalise linear regression models and static statistical linear models. The course offers a very powerful fundamental probabilistic approach to forecasting, controlling and learning about uncertain commercial, financial, economic, production, environmental and medical dynamic systems.
- State space modelling
- Bayesian updating of beliefs
- Specifying Dynamic Linear Models
- Updating Dynamic Linear Models, forecasting
- Building Dynamic Linear Models, accommodating external information
- ARIMA models, stationarity
The theory will be illustrated by real examples from industry, marketing, finance, government, agriculture etc.
Students will be given selected advanced research material for independent study and examination.
Books: Printed course notes will be available. Mike West & Jeff Harrison, “Bayesian Forecasting and Dynamic Models”, 1997 (2nd edn.) Springer - Verlag. Andy Pole, Mike West & Jeff Harrison, “Applied Bayesian Forecasting and Time Series Analysis”, 1994 Chapman and Hall.
Research material on some advanced topics will be made available.
Assessment: 100% by 2-hour examination.
You may also wish to see: