International Max Planck Research School Course (joint with Dr. Claas Teichmann)

Winter term 2012/2013


Selected topics in climate data analysis: Bayesian statistics, extreme values, and pattern classification

The basic concepts of Bayesian statistics and their relation to statistical mechanics will be explained with a focus on the parameter estimation problem and Markov chain Monte Carlo. Moreover, the course covers an introduction to the statistical theory of rare events. Recently developed Bayesian hierarchical models for extreme events in climate science will be discussed. The third part will concentrate on learning algorithms (Bayesian and non-Bayesian), feature selection, and classification with a particular emphasis on classification and regression trees. Tree methods are an attractive tool not only for defining and classifying specific characteristics of data sets, but also for identifying influencing factors of these features. Despite their suitability for large data applications, they have not yet found widespread application in climate science.
The students will not only be familiarized with the basic concepts of the topics, but also be introduced to possible research fields like the application of classification tree methods to extreme values.
In the integrated hands-on sessions a short introduction to the open source statistics language "R" will be given.

The course will be as self-contained as possible, but familiarity with basic statistical notions like random variable, probability distribution, conditional probabilities etc. would be helpful.


Start: November 8, 2012, 10.15 a.m. in room 002 (ZMAW building)