Dear Colleagues:
On 9 and 10 March 2009, the National Center for Ecological Analysis and
Synthesis (NCEAS) will host a workshop in the use of AD Model Builder
(admb-project.org), with attention to its applications that are general
to ecology. AD Model Builder is a well known modeling package in the
fisheries biology community, with applications that are general to
ecology and other sciences; NCEAS and the ADMB Foundation together have
recently purchased this package in order to make it free and open
source, and to make it accessible to a broader community of scientists.
It is now free to download.
Several seats are available in the March workshop, and the workshop
itself is free of cost, but we can not pay your travel expenses. We can
help to secure lodging within walking distance that is usually below
standard rates for Santa Barbara (e.g., $120/night or less). If you are
interested in attending, please send me a response
(hamp...@nceas.ucsb.edu) by 16 Feb 2009. Please include a CV and a short
statement about why you would like to attend - in the event that we
receive more interest than we can accommodate, we will select
participants who represent a breadth of disciplines.
A description of the 2-day ADMB workshop follows. A 1-day workshop will
also be offered at the ESA meeting this summer.
*AD Model Builder: a Free Tool for Parameter Estimation of Complex
Nonlinear Statistical Models
*Instructors: Mark Maunder & Anders Nielsen
This mini-course targets quantitative ecologists, and students who need
to handle complex nonlinear statistical models (both frequentist and
Bayesian). AD Model Builder (admb-project.org) is a highly efficient
freely available software for implementing non-linear statistical
models. The main reasons for preferring AD Model builder are: 1)
Flexibility. The user is free to define any desired model, and not
limited to choose between a set of predefined models. 2) Speed.
Automatic differentiation can make the difference between waiting hours
and seconds for a converging model fit. 3) Precision. Automatic
differentiation calculates the derivatives as accurately as if the
analytical derivatives were implemented. 4) Quantification of
uncertainties. With almost no extra effort AD Model builder produces
several different estimates of the uncertainties of model parameters and
selected derived quantities.
A beginners’ course in ADMB likely will include: 1) An overview of ADMB.
2) A refresher on model development and likelihood based inference. 3)
Installing and set up the software. 3) A use case. 4) Options for
importing data (the simple and the more exotic). 5) Definition of model
parameters (limits, phases, and some tricks). 6) Programming the
likelihood function. 7) Specification and formatting of output. 8)
Debugging, memory management, and other important implementation issues.
9) Estimation uncertainties (delta, profile, and MCMC methods). 10)
Random effects models in AD Model Builder.
The actual contents of the course will be customized to fit the
audience. The form will be a mixture between lectures and hands on
exercises.
--
Stephanie E. Hampton
Deputy Director
National Center for Ecological Analysis & Synthesis
University of California, Santa Barbara
735 State St., Suite 300
Santa Barbara, CA 93101-3351, USA
http://www.nceas.ucsb.edu
hamp...@nceas.ucsb.edu
Tel (805) 892-2505
Fax (805) 892-2510
