objective of the summer school is to provide students with both, a wide
overview of the body of statistical and data mining techniques widely
applied to astronomical problems, and a specialized primer to the
latest developments in this field, occurred in the past decade.
A preliminary program of the school is as follows:
- Classical statistics: basic concepts, parameter estimation,
statistical inference, error analysis, hypothesis testing, confidence
intervals. The frequentist aproach versus Bayesian
statistics. Priors. The problem of model selection. Sampling
techniques (MCMC, nested sampling...). Latest developments in Bayesian
inference for Astronomy
- Advanced statistical techniques: Time series analysis. Wavelet
analysis. Statistical techniques for astronomical image
processing. Spherical statistics.
- Supervised Classification and Regression: the problem of feature
selection, the curse of dimensionality. Regression methods. Assessment
of regression models. Continuous and categorical
variables. Classification models (artificial neural networks, support
vector machines, bayesian networks...). Model evaluation (n-fold cross
validation and variants; statistical tests). Feature selection
revisited. The construction of training and test sets in astronomical
applications of Data Mining.
- Unsupervised classification: alternative methodologies, the problem of feature selection for clustering, evaluation.
- Data Mining and Statistics in the era of the petabyte databases.
Technical aspects, database architecture, intelligent access,
distributed computing, efficient software
Basic knowledge of statistics is desirable. The lectures will be
interspersed with practical exercises. Students will be asked to bring
their laptops with appropriate software installed.