[Statlist] Workshop Announcement: Fitting Hierarchical Models Using the Laplace Approximation and Automatic Differentiation

[Mollie Brooks]

Workshop Announcement: Fitting Flexible State-Space and Hierarchical Models Using the Laplace Approximation and Automatic Differentiation
Applications Due Friday 1 August 2014
 

1-3 September 2014, University of Zurich, Irchel
 
Prof. Hans Skaug, Dept of Mathematics, Univ of Bergen
 
Dr. Kasper Kristensen, Dept of Applied Maths and Comp Sci, Technical Univ of Denmark
 
Dr. Mollie Brooks, Institute of Evolutionary Biology and Environmental Studies, Univ of Zurich                                    
 
Workshop Leaders:
Hans Skaug and Mollie Brooks are members of the core development team of ADMB (Automatic Differentiation Model Builder), a powerful software package for fitting nonlinear statistical models including latent variables (http://admb-project.org). Kasper Kristensen is developing a similar package in R (TMB, Template Model Builder, https://github.com/kaskr/adcomp). Hans Skaug has also participated in developing interfaces to use ADMB from R (R2admb and glmmADMB).
 
Target Audience: PhD students, postdocs, professors, and researchers in applied math/stats, finance, quantitative ecology, biostatistics
                                                                                                       
Workshop Outline:
ADMB and TMB are free and open-source software for fitting statistical models using maximum likelihood estimation (MLE), automatic differentiation, and the Laplace approximation. They are as flexible as Bayesian methods for fitting hierarchical models, but faster and more robust. ADMB is 10 times faster than Bayesian methods and TMB is often 100 times faster.
 
The workshop will include introductions to Laplace approximation and automatic differentiation and their usefulness for MLE. We'll also learn about a new algorithm for detecting the sparseness of the Hessian matrix. The majority of class time will be spent on practical applications of software. We will begin by estimating a random walk time series with observation error. Examples will come from both ecology (e.g. population growth) and finance (e.g. stochastic volatility) and include spatial and temporal patterns.
 
Prerequisites:
Participants should have prior exposure to basic programming (esp. relevant: R and C++), maximum likelihood estimation, and probability distributions and link functions commonly used in generalized linear models.
Workshop participants should bring their laptop with the most recent ADMB and R installed. Installation instructions will be emailed prior to the workshop.
 
Time, Date, and Location:
9-17h, 1-3 of Sept (Monday-Wednesday), UZH Irchel, Y25-H79
 
Registration/ Application:
The course is free, thanks to a GRC Grant from the UZH Graduate Campus.
Due to space limitations (40 participants), interested persons are asked to submit an application, describing your background and the course�s usefulness for your research, to mollieebrooks using gmail.com. Include �state-space workshop� in the subject.

------------------------
Mollie Brooks, PhD
Postdoctoral Researcher, Population Ecology Research Group http://www.popecol.org
Institute of Evolutionary Biology & Environmental Studies, University of Z�rich




	[[alternative HTML version deleted]]