event
Thursday 26 Oct 2017: Statistical Science - INLA - A gentle introduction
Stefan Siegert -
H101 14:30-15:30
The Integrated Nested Laplace Approximation (INLA) method has recently become a
popular tool for its efficiency in solving high-dimensional spatio-temporal
inference problems. INLA is essentially a collection of standard mathematical
approximations that, in combination, yield a framework for Bayesian computation
that is much faster than alternatives such as MCMC. In this talk, I will walk
you through the mathematical foundations of INLA, deriving the essential steps
from scratch as much as possible. I will also say a few things about the INLA
R-package. The goal of the talk is to teach you the basics of the method so we
can start a critical discussion if and how we can use INLA to solve more
complicated inference tasks.