APPLIED BAYESIAN DATA ANALYSIS USING STAN
OCTOBER 24 – 25, 2014
Daniel Lee, Department of Statistics, Columbia University, New York
Michael Betancourt, Department of Statistical Sciences, University College London
Location: Swiss Ornithological Institute, Sempach, Switzerland
Daniel Lee and Michael Betancourt are members of the core development team of STAN. Both are excellent software engineers. They work with Prof. Andrew Gelman on applied Bayesian statistics, modelling and software development.
Daniel is doing research on Monte Carlo Markov chains (MCMC) and Bayesian analyses (http://linkd.in/12xLZYK). Michael studies the mathematical foundations of Bayesian methods in order to motivate efficient practical techniques (http://www.homepages.ucl.ac.uk/~ucakmjb/).
Stan is an open-source, general Bayesian inference tool with interfaces in R, Python, Matlab, and the command line. Stan was developed to address the speed and scalability issues of existing Bayesian inference tools, BUGS and JAGS, while maintaining the ability to write models easily through a statistical language. The default algorithm is an auto-tuned variant of Hamiltonian Monte Carlo, which is a more efficient MCMC algorithm for general problems than Gibbs sampling or random-walk Metropolis Hastings. This exciting new tool is now open to everybody and has the potential to be very useful in the daily life of a data analyst that use comparably complex models in a Bayesian framework.
The course will start with a short introduction to Bayesian inference and how Stan works. However, the main goal of the course is the practical application of Stan to different models starting with ordinary linear regression and ending with more complex models such as generalized linear mixed and hierarchical models.
Download the attached course announcement for full details and information on how to registration.
course_announcement_STAN 2014.pdf 337.4KB 61 downloads