Bayesian analysis made easy
Alexandre Santerne, Michaël Bazot
CAUP
This series of lectures will be divided in two parts. In the first, I will present some applications of Bayesian Statistics to astrophysics. An emphasis will be put, on one hand, on the use of Poisson statistics for supernova-neutrino counting experiments and, on the other hand, on periodogram analysis. The second part of these lectures will be concerned with problems often encountered in the practice of Bayesian Statistics. We will first discuss the computation of the so-called "Bayesian evidence", through the study of some existing algorithms. To close this series, I will present some theoretical considerations on the nature and use of prior distributions in Bayesian Statistics.
Lecturer: Alexandre Santerne
In this single course, I will present some recipes of Bayesian analysis of data in an easy way. I will start with the main principle of Bayesian analysis of data and the notion of priors (and why they are important). Then, I will present the recipes of the Markov Chain Monte Carlo and the Metropolis–Hastings algorithm. I will also discuss some limitation of this approach and some tricks to skirt them. This course is not a pratical course (no need to bring your laptop). To attend this course, it is not requested to be a world-wide expert in Bayesian statistics, but some knowledge about what a probability is would be more than useful.
The slides from this lecture can be downloaded here.
Lecturer: Michaël Bazot
Lecturer: Michaël Bazot
Lecturer: Michaël Bazot
Lecturer: Michaël Bazot