Speaker
Mr
Antonie de Beer
(Student)
Description
In data analysis it is often difficult to understand which parameters
in a model are relevant and which are not. The number of parameters
must also be optimised in a systematic way.
In the Bayesian framework, the optimal set of parameters is determined
by maximising the model evidence (eg via the Savage-Dickey density
ratio), which is made up of a likelihood and a prior probability.
Commonly encountered orthogonal series expansions are, however,
necessarily oscillatory and hence cannot describe positive definite
likelihood probabilities. Expanding square roots of the likelihood
permits us to achieve positivity. The argument for using square roots
is reinforced by the constraints to the model by the Savage-Dickey
density ratio and the Jeffreys prior.
Level for award<br> (Hons, MSc, <br> PhD, N/A)?
Msc
Main supervisor (name and email)<br>and his / her institution
Prof. Hans Eggers
eggers@physics.sun.ac.za
Stellenbosch University
Would you like to <br> submit a short paper <br> for the Conference <br> Proceedings (Yes / No)?
No
Apply to be<br> considered for a student <br> award (Yes / No)?
Yes
Primary author
Mr
Antonie de Beer
(Student)
Co-authors
Hans Eggers
(Stellenbosch University)
Dr
Michiel De Kock
(Stellenbosch University)
