Tuesday, August 7, 2007: 8:00 AM-11:30 AM
C1&2, San Jose McEnery Convention Center
OOS 10 - What is the right size ecological model? Views on model complexity and parsimony from different statistical paradigms
Recent innovations in model fitting technology allow ecologists to fit an impressive array of complex, hierarchical and parameter-rich models to data. Bayesians and frequentists alike can harness modern computational power to entertain and fit models of complexity that was out of practical reach less than a generation ago. With this model fitting capability in hand, a necessary next question is: how does one choose the right level of model complexity for a given problem? This session brings together ecologists and applied statisticians who have grappled recently with questions about model complexity or parsimony in an ecological context. Model complexity and parsimony is a particularly interesting axis on which to frame a discussion of contemporary statistical frontiers because it is an issue confronted by all data analysts, and it impacts the interpretation of scientific results in major ways. However, different statistical paradigms generate different perspectives concerning how model complexity affects the interpretation of an analysis and how one interprets model parsimony. This session includes speakers from a variety of statistical paradigms and asks each to emphasize how their different vantage points may lead to different viewpoints on parsimony in statistical modeling. The session will also emphasize practical strategies for data-driven model selection with real data analysis problems. Speakers will illustrate viewpoints on model complexity in the context of real ecological data analysis problems. We hope that this session will interest a practical audience, and provide a useful roadmap of the different philosophies and methodologies regarding model complexity and model selection.
Organizer:Kevin Gross, North Carolina State University
Co-organizer:E. E. Holmes, Northwest Fisheries Science Center
Moderator:Kevin Gross, North Carolina State University
8:00 AMOOS 10-1Complementary Bayesian, frequentist, and cross-validation approaches to model selection and testing
Perry de Valpine, University of California - Berkeley
8:20 AMOOS 10-2Structural equation models: Perspectives on and challenges to model selection
James B. Grace, United States Geological Survey
8:40 AMOOS 10-3Bayesian multimodel inference: Suggestions for handling the potentially profound effects of priors on parameters
William A. Link, United States Geological Survey
9:00 AMOOS 10-4Assessing models of complex ecological systems using Pareto optimality
E. David Ford, University of Washington, Maureen A. Kennedy, University of Washington, Joel H. Reynolds, US Fish & Wildlife Service, Marianne C. Turley, Bureau of Land Management, R. Komuro, The Bioengineering Institute
9:20 AMOOS 10-5Proper and improper use of AIC
Shane A. Richards, University of Durham
9:40 AM Break
9:50 AMOOS 10-6Parsimony for ecological models of multivariate response data based on dissimilarity matrices
Marti J. Anderson, University of Auckland
10:10 AMOOS 10-7Parsimony and complexity in mixed models for longitudinal data
Philip Dixon, Iowa State University
10:30 AMOOS 10-8Easy maximum likelihood estimation for complex ecological models using Bayesian Markov chain Monte Carlo methods
Brian Dennis, University of Idaho, Subhash R. Lele, University of Alberta, Frithjof Lutscher, University of Ottawa
10:50 AMOOS 10-9Bayesian model selection and trend analysis of the Florida manatee via aerial surveys
Bruce A. Craig, Purdue University
11:10 AMOOS 10-10Model complexity affects predicted transient population dynamics: A case study with Acyrtosiphum pisum
Brigitte Tenhumberg, University of Nebraska, Lincoln

See more of Organized Oral Session

See more of The ESA/SER Joint Meeting (August 5 -- August 10, 2007)