Home > Contribution List
We will start from the beginning with a discussion of the sorts of statistical problem that arise in scientific analysis, and then review basic probability theory for completeness. We will discuss briefly the difference between the Bayesian and Frequentist approaches. Priors play a special role in Bayesian analysis, so we will pay attention to when priors matter and when they don't. We discuss ... More
Presented by Prof. Alan HEAVENS on 8 Mar 2015 at 10:00 AM
Here we address on how MCMC can be improved for certain classes of problem, using Hamiltonian (or Hybrid) Monte Carlo (HMC). This technique can be used for some problems with huge numbers of parameters (millions).
on 8 Mar 2015 at 2:15 PM
First practical session, to cement the ideas of MCMC, to understand how to design or use an MCMC code.
on 8 Mar 2015 at 12:00 PM
Finish numerical exercises. We may also pick up on topics that arise during the day.
on 8 Mar 2015 at 4:45 PM
Here we ask the higher-level question of which model is more probable, given the data (regardless of the parameters), and introduce the notion of Bayesian Evidence.
Presented by Prof. Alan HEAVENS on 8 Mar 2015 at 4:00 PM
We discuss parameter inference - i.e. within a model, how do we determine the most probable parameters and their uncertainties. We consider simple grid searches (<4 parameters) to more flexible sampling approaches that can deal with larger (~10) dimensional parameter spaces. We concentrate on a standard method - Markov Chain Monte Carlo (MCMC), which forms the basis for the hands-on exercise in t ... More
Presented by Prof. Alan HEAVENS on 8 Mar 2015 at 11:30 AM