A Distance-based Approximate Bayesian Computation Procedure for Time-Series Data
Speaker: Dr Hien Duy Nguyen
Affiliation: University of Queensland
Abstract
Approximate Bayesian Computation (ABC) is a useful framework for conducting inference when likelihood functions are computationally infeasible or unknown. Common ABC procedures require assumptions of independence or exchangeability to generate high-quality inference. In this work, we consider the use of distance-based ABC methods, based on U-statistics, that can feasibly be used when data are dependent. We show a consistency result regarding such procedures and demonstrate their usefulness on some example time-series experiments.
About Maths Colloquium
The Mathematics Colloquium is directed at students and academics working in the fields of pure and applied mathematics, and statistics.
We aim to present expository lectures that appeal to our wide audience.
Information for speakers
Information for speakers
Maths colloquia are usually held on Mondays, from 2pm to 3pm, in various locations at St Lucia.
Presentations are 50 minutes, plus five minutes for questions and discussion.
Available facilities include:
- computer
- data projector
- chalkboard or whiteboard
To avoid technical difficulties on the day, please contact us in advance of your presentation to discuss your requirements.