Astrostatistics Image Penn State University Eberly College of Science Center for Astrostatistics Center for Astrostatistics

2005 Summer School | Biographies | Travel, Visa and Lodging | Contact | Registration 

 

Summer School in Statistics for
Astronomers & Physicists
June 5-17, 2005

 

Overview: Local accommodation provided at no cost to the admitted participants during the sessions. The 2005 summer school offering has three sessions. Participants may register for one session or any combination of sessions. Enrollment is limited to 25 participants per session.

  • Statistical inference for astronomers  (June 5-10, 2005)
    A 5-day course in fundamental statistical inference designed to provide physical scientists, with little or no prior exposure to statistics, with a strong conceptual foundation in modern statistics and to develop a repertoire of well-established techniques applicable to cutting-edge research in astronomy and astrophysics. These statistical techniques include hypothesis testing and parameter estimation, recent developments in confidence interval estimation, non-parametric methods, maximum likelihood methods, Monte Carlo methods, and many other procedures.  
  • Spatial processes and image analysis  (June 12-14, 2005)
    A 3-day specialized course in image analysis and spatial point processes, developing sophisticated tools for image restoration and noise reduction, feature extraction, and characterization of spatial processes such as the clustering of galaxies in 3-dimensional redshift space or Galactic stars in 6-dimensional phase space. 
  • Computational algorithms for astrostatistics  (June 15-17, 2005)
    A 3-day specialized course in computational algorithms which provide insights into problems too complex for standard analytic approaches. These methods include EM-algorithms, genetic and evolutionary algorithms, and homotopy continuation methods. The emphasis will be on algorithms which can find confidence interval to nested parametric models, nonlinear relationships in high-dimensional space, and summations over hypothesis spaces needed for Bayesian approaches. General formulation of EM algorithms, convergence properties, and computation of standard errors. Extensions of EM for accelerating convergence. Monte Carlo algorithms (Rejection Sampling, Importance Sampling), Markov chain Monte Carlo algorithms, Metropolis-Hastings, Gibbs Sampling, and Monte Carlo versions of EM algorithms.  

In the spring of 2006 and summer of 2007, CASt plans to offer courses in Bayesian inference, multivariate analysis, and time series analysis.

Forthcoming  2006 Penn State Conference:

    Statistical Challenges in Modern Astronomy IV. The fourth in a series of interdisciplinary international conferences, Statistical Challenges in Modern Astronomy, will be held during June 11-14, 2006 at the Pennsylvania State University, University Park, USA.
     

Motivation for the Summer School in statistics: The volume and complexity of data in astronomy and physics have increased enormously in recent years. Statistics provides the means for extracting physical insights from such data. A vast range of methodological challenges emerge from optical & X-ray astronomy, analysis of Virtual Observatory databases, the spatial distributions of galaxies or the cosmic microwave background, the spectra of stars or quasars, the variability of accreting black holes, and a vast range of other datasets. Analogous problems, sometimes methodologically similar and sometimes different, arise in fields of observational physics such as cosmic ray and neutrino studies, gravitational wave detection and high energy particle physics. 

To meet these challenges, modern statistics has mushroomed with vast capabilities: survival analysis, spatial point processes, wavelet analysis, multivariate analysis, time series analysis, likelihood methods, and computational techniques such as bootstrap resampling, the EM Algorithm, and MCMC. While some physical scientists know some of these methods, traditional statistical pedagogy is poorly adapted to the needs of physical scientists.
The newly formed Center for Astrostatistics builds upon an experienced interdisciplinary group at Penn State to provide a range of services for both the scientist and statistician. Our first project is to inaugurate an annual summer school in modern statistics for astronomers and physicists. To train scientists in appropriate advanced statistical methodology, we are developing new curricula, extracting concepts and methods from a wide range of modern statistics, and applying them with readily available software to datasets from the physical sciences.  The participants are taught by leading experts and experienced educators of advanced statistics.      




NSFDepartment of StatisticsEberly College of ScienceDepartment of Astronomy and Astrophysics