Measurement error and density dependence models.
DENNIS, B.* 1,2
Dept Fish and Wildlife Resources, University of Idaho 1
Division of Statistics, University of Idaho 2
This presentation surveys various methods of incorporating measurement or sampling error into models for analyzing time series of population abundances. Two promising approaches are identified, based on a "state space" formulation that combines process noise and sampling variability in a stochastic population growth model. First, lognormal sampling error can be incorporated into a stochastic, discrete-time Gompertz growth model. The resulting nonlinear model can be transformed into a well-known linear state space model (the "Kalman filter") that has easy formulas and straightforward computing. Second, Poisson sampling error (approximately characteristic of mark-recapture surveys, sighting indexes, and transects) can be incorporated into any of the familiar stochastic population growth models (stochastic Ricker, Gompertz, etc.). The resulting nonlinear state space models require numerically-intensive computing methods for fitting to time series data. Results of data analyses and simulations under both approaches will be discussed.
Keywords: density dependence, sampling error, measurement error, stochastic population growth, time series
This abstract is being presented at: 9:30 AM in session:
Symposium # 15: Measurement Error in Ecological Data.