Introduction to data analysis for astronomers

Menu Start Research Publications Statistics playground Teaching CV Contact

Lecture dates: Friday 18. April: no lecture (Easter) Friday 25. April 09.15-10.45: Lecture 1 (Introduction) Friday 2. May 09.15-10.45: Lecture 2 (Regression I) Friday 9. May: no lecture (room blocked) Friday 16. May 09.15-10.45: Lecture 3 (Regression II) Friday 23. May 09.15-10.45: Lecture 4 (Numerical Methods) Friday 30. May 09.15-10.45: Lecture 5 (Model Comparison) Friday 6. June 09.15-10.45: Lecture 6 (Classification) Friday 13. June 09.15-10.45: Lecture 7 (Learning Theory) Friday 20. June 09.15-10.45: back-up date Location: Seminar room H-310 in basement of House of Astronomy at Konigstuhl. Requirements: basic analysis (differentiation, integration) basic matrix algebra no prior knowledge of statistics required undergraduates (Bachelor/Master) need to solve exercises or take an (oral or written) exam to get the 1 credit for this course PhD students get the 1 credit for this course for attendance (documented by a list with signatures) Syllabus Problem sheets: Exercises 1 (submission due 2. May) Exercises 2 (submission due 16. May) Exercises 3 (submission due 30. May) Exercises 4 (submission due 6. June) Literature and further material: Barlow 1999: Readible introduction but far too basic. Frequentist statistics, i.e., no big picture. Gelman, Carlin, Stern, Rubin 2004: Thorough introduction into Bayesian statistics. Hastie, Tibshirani, Friedman 2009: Very good/classic textbook. Also includes classification. MacKay 2003: A classic and well written. Sometimes a bit philosophical and takes several detours on side-topics that are intersting but distract from the red line. The book is freely available here. Vapnik 1999: Insightful and illuminating read written by one of the founders of Learning Theory, though later chapters become increasingly mathematical. Youtube has many nice lecture courses on data analysis, e.g.: Andrew Ng, Stanford, Machine Learning, 2008 Yaser Abu-Mostafa, CalTech, Introductory Machine Learning Course, 2012 Paper on Bayes factors by Kass and Raftery. Sections 1, 3 and in particular the discussion in Section 8 are very interesting. The reversible-jump MCMC is an MCMC that can change the number of fit parameters during iterations. RJMCMC is described, e.g., on Wikipedia.