Challenges and Advances in Statistical Software For Ecology

Wednesday, August 13, 2014: 8:00 AM-11:30 AM
Gardenia, Sheraton Hotel
Perry de Valpine, University of California - Berkeley
Matthew Smith, Microsoft Research; and David S. LeBauer, University of Illinois
Christopher J. Paciorek, University of California, Berkeley
This symposium will address the design and development of statistical software for ecology as a research topic in its own right as well as providing overviews of current tools and recent advances in a range of application domains. Expertise with software development is rapidly growing in the ecological community, and software development is an increasingly recognized aspect of research productivity as part of the "data science" movement. The opening talk will set the stage for the session with an overview on the interplay between statistical methods, software, and scientific advances. The other speakers will draw upon their experience developing and applying new statistical software tools for ecological problems. Speakers will give overviews of software in their application domains, which include spatial models, species distribution models, ecosystem flux models, generalized linear mixed models, and general hierarchical models. Then they will discuss approaches, pitfalls and successes for developing new software and statistical methods based on their own projects. Issues include the feedbacks between advances in scientific questions and the development of new methods and software; the role of statistical software in scientific workflows that include increasingly automated and standardized data streams; different approaches to specifying models and data; the pros and cons of making available advanced methods that require expert interpretation; and the challenges of combining multiple models and algorithms modularly. The projects discussed include a range of software design choices, including web-based tools, R packages, C++ libraries, programmable domain specific languages, and scientific workflow pipelines. While these talks will include overviews of the tools, their goal is not simply to give demonstrations but rather to discuss concepts and challenges of putting cutting-edge analysis methods at ecologists’ fingertips. The symposium is intended to advance dialogue within the ecological community on different approaches to designing and building software for advancing science, as well as to bring attendees up to date on recently developed tools.
8:30 AM
 Developing spatial modelling methodology for complex ecological data sets with INLA - the story of a symbiosis
Janine B. Ilian, University of St. Andrews and Norwegian University of Science and Technology
9:00 AM
 Beyond BUGS: Flexible programming of statistical methods for BUGS-compatible models using NIMBLE
Perry de Valpine, University of California - Berkeley; Daniel Turek, University of California - Berkeley; Christopher J. Paciorek, University of California, Berkeley; Rastislav Bodik, University of California, Berkeley; Duncan Temple Lang, University of California, Davis
9:30 AM
9:40 AM
 Combining heterogeneous data and process understanding using the Predictive Ecosystem Analyzer
David S. LeBauer, University of Illinois; Michael Dietze, Boston University; Shawn P. Serbin, Brookhaven National Laboratory
10:10 AM
 Technical, design, and cultural aspects of developing general-purpose software for mixed models
Benjamin M. Bolker, McMaster University; Steven C. Walker, McMaster University
See more of: Symposia