PARENT SESSION
Oral Session 16: Statistics and Biometrics.
Presiding: E Garton and T Simons
Monday, August 2, 8:00 AM to 11:30 AM, Meeting Room D 139.
Assessing conservation status from time series data: Incorporating complex models through computer-intensive algorithms.
Lee, Danny1, 1 Pacific Southwest Research Station, Arcata, CA
ABSTRACT- Many existing methods of assessing conservation status or extinction risk use time series data on measures of abundance or other indices of population size. The conventional statistical approach is to fit a parsimonious model that best describes the data in hand. While statistically elegant, the simplification inherent in such models often does not allow examination of factors such as age structure or partitioning of environmental stochasticity. More complex models are available, but parameter estimation is often done piecewise, i.e., one component at a time with the hope that the overall prediction will be consistent with observed data. Increasingly, more complex models are being developed using a hierarchical Bayesian approach and computer-intensive methods such as Markov chain Monte Carlo (MCMC) algorithms. These developments have two important implications. First, the use of Bayesian methods implies an epistemological shift with inferential significance (which often is unrecognized). Second, the application of more complex models rarely leads to reduction of uncertainty. Instead, these models provide a richer picture of the sources and consequences of uncertainty in knowledge and unpredictability in nature. I illustrate these points using data on California spotted owls (Strix occidentalis), and an age-structured population model fitted using a computer-intensive approach.
Key words: time series, Bayesian methods, PVA, MCMC