ESAC DATA analysis and Statistics Workshop 2015

    Following the successful ESAC Data Analysis and Statistics Workshop in October 2014, we held a second edition from Monday to Wednesday on October 19 to 21, 2015. This workshop was intended primarily for ESAC employees, it was funded by the ESAC Faculty, and no prior background knowledge in statistics was assumed or necessary. The confirmed lecturers are Luis Manuel Sarro and Eric Feigelson. 

    TABLE OF CONTENTS

    TUTORS, NOTEBOOKS & NOTES

    AGENDA

    • Monday 19 October 2015

      09:00 - 09:30: Welcome and Installations troubleshooting

      09:30 - 11:00: Introduction to R (Eric Feigelson)

      11:00 - 11:30: Coffee/tea break

      11:30 - 13:00: Basic concepts in statistics: Random variables and their distributions; and introduction to information theory: entropy, mutual information and related measures; statistical inference: a conceptual presentation of the kind of questions that arise in astrophysics: parameter inference (regression, classification), model comparison and selection, hypothesis testing (Luis Manuel Sarro)

      13:00 - 14:00: Lunch @ ESAC canteen

      14:00 - 15:30: Bayesian inference I: Fitting a straight line. The main characters in bayesian inference: priors, likelihood and posteriors. Conditional probabilities and their interpretation. The concept of marginalization. Model selection: the bias-variance trade off (Luis Manuel Sarro)

      15:30 - 16:00: coffee/tea break

      16:00 - 17:30: Bayesian inference II: Analysing a spectrum. We will apply the topics outlined above to the problem of analysing a spectrum. Hands-on session, with solutions made available at the end (Luis Manuel Sarro)

    • Tuesday 20 October 2015

      09:30 - 11:00: Bayesian inference III: Estimating distances from parallaxes. A mixture of theory and hands-on session based on the recent paper by Coryn Bailer-Jones (Luis Manuel Sarro)

      11:00 - 11:30: Coffee/tea break

      11:30 - 13:00: Bayesian inference IV: Hyperdimensional parameter spaces. The problem of estimating large numbers of parameters. Optimization techniques based on MCMC techniques. Posterior sampling, model selection and a very brief digression on hierarchical models (Luis Manuel Sarro)

      13:00 - 14:00: Lunch @ ESAC canteen

      14:00 - 15:30: Statistics and astronomy: Role of statistics in science; astrostatistics past & present; statistical computing (Eric Feigelson)

      15:30 - 16:00: coffee/tea break and group picture

      16:00 - 17:30: Parametric vs. nonparametric inference: Problems with histograms and power laws; nonparametric density estimation and local regression techniques (Eric Feigelson)

    • Wednesday 21 October 2015

      09:30 – 11:00: Common statistical problems in the astronomical literature (Eric Feigelson)

      11:00 - 11:30: coffee/tea break

      11:30 - 13:00: Censoring & truncation: Non-detections & flux-limited surveys; survival analysis; Kaplan-Meier & Lynden-Bell-Woodroofe estimators (Eric Feigelson)

      13:00 - 14:00: Lunch @ ESAC canteen

      14:00 - 15:30: Clustering & classification: Difficulties of unsupervised clustering; methods of multivariate classification (Eric Feigelson)

      15:30 - 17:00: Data visualization: 2D & 3D plotting, multidimensional visualization (static & dynamic), color schemes (Eric Feigelson)

    SOFTWARE & INSTALLATION

    Compared to previous years, and to what was previously announced, the hands-on sessions for the 2015 Workshop edition will be based on R and not on Python.

    LINUX

    1. Download R from here or alternatively use your Linux package management system
    2. Install the package

    Mac OS X

    1. Download R from here

      There are 2 packages, so be sure to pick the one that matches your version of Mac OS X: R-3.1.1-snowleopard.pkg for Mac OS X 10.6 (Snow Leopard), 10.7 (Lion), 10.8 (Mountain Lion), R-3.1.1-mavericks.pkg for Mac OS X 10.9 (Mavericks)

    2. Double-click to install

    Windows

    1. Download R from here
    2. Install the package following the wizard and accepting all defaults.

    ORGANIZING COMMITTEE

    • Michele Armano
    • Guillaume Belanger
    • Hervé Bouy
    • Uwe Lammers
    • Jean-Christophe Leyder
    • Bruno Merín
    • William O'Mullane
    • Álvaro Ribas
    • Pablo Riviere-Marichalar (chair)
    • Celia Sánchez
    • Luis Manuel Sarro
    • Roland Vavrek

    FUNDING

    The SOC warmly thanks the ESAC Science Faculty for fully funding this workshop.