 | | |
| ENVT0016-1 | Bayesian statistics applied to environment
|
|
| Duration : | 24h Th, 6h Pr |
|
| Number of credits : |
| Master in sciences and environment management | | Second semester | | 2 |
|
| Master en sciences et gestion de l'environnement, à finalité spécialisée en développement durable, 1st year | | Second semester | | 2 |
|
| Master en sciences et gestion de l'environnement, à finalité spécialisée en énergies renouvelables, 1st year | | Second semester | | 2 |
|
| Master en sciences et gestion de l'environnement, à finalité spécialisée en intervention technique, 1st year | | Second semester | | 2 |
|
| Master en sciences et gestion de l'environnement, à finalité spécialisée en interfaces sociétés-environnements, 1st year | | Second semester | | 2 |
|
| Master en sciences et gestion de l'environnement, à finalité spécialisée en surveillance de l'environnement, 1st year | | Second semester | | 2 |
|
| Master en sciences et gestion de l'environnement, à finalité spécialisée en procédés biologiques de valorisation des déchets, 1st year | | Second semester | | 2 |
|
|
|
| Lecturer : | Jean‑Jacques Boreux |
|
Language(s) of instruction :
|
| French language |
|
Course contents :
|
| This course consists of two parts.
The first concerns the finite mixture models which all components belong to the same member of the so-called exponential family. Indeed, many real phenomena are too complex to be captured from any standard model such as the normal law. The inference of the mixture model is carried out under the Bayesian paradigm using MCMC methods like Gibbs sampler or Metropolis-Hastings algorithms, sometime preceded by an EM step (expectation-maximization). Various applications in environmental science and economics illustrate the suitability of mixture models in the decision support.
The second part is dedicated to extreme value models. Indeed, protection against damages caused by extreme events is at the heart of the decision making. GEV (generalized extreme value) and POT (peak over threshold) models are analyzed, resolved and illustrated. |
|
Learning outcomes of the course :
|
| The purpose of the course is to provide powerful tools to people interested by the help of decision making in uncertain future, not only in science and environmental management, but also in all sciences where the decision-maker wants to set up its choices on right basis. |
|
Prerequisites and co-requisites/ Recommended optional programme components :
|
| An elementary course of Bayesian statistics and whether to use a software such as R or Matlab. |
|
Planned learning activities and teaching methods :
|
| The theoretical parts (presence of the Professor) and practices (exercises in computer room or at home) are roughly equivalent in time. Examples treated in the course are deliberately simplified to go to the essential. The personal works are more realistic or even real problems. |
|
Mode of delivery (face-to-face ; distance-learning) :
|
| The course is given in modules, each covering a half day of 4 hours (1 credit = 3 half-days), over a period concentrated in time. Personal and practical works in the computer room (Arlon campus) or home are included in the schedule and planned credits. |
|
Recommended or required readings :
|
| Boreux, JJ, Parent, E., Bernier, J. (2010). Pratique du calcul bayésien. Springer.
Marin, JM, Robert, C. (2006). Bayesian Core : a practical approach to computational Bayesian statistics. |
|
Assessment methods and criteria :
|
| The student must make a personal work by placing in a professional context. In this way he proposes a problem, if possible illustrated with real data, and provide a final report which will enable to judge his scientific background and professionalism. |
|
Training(s) :
|
| None. |
|
Organizational remarks :
|
| None |
|
Contacts :
|
| Jean-Jacques Boreux
jj.boreux@ulg.ac.be |
|
