PARENT SESSION
Symposium 24: Bridging the gap between theory and empiricism in ecology
Organized by: P Amarasekare and M Neubert
Friday, August 12, 8:00 AM - 11:30 AM, Meeting Room 517 C, Level 5, Palais des congrès de Montréal

Linking mechanistic models with epidemiological data: Parameter estimation in the face of incomplete information.

King, Aaron^*,1, Lele, Subhash², Rohani, Pejman³, Pascual, Mercedes⁴, ¹ University of Tennessee, Knoxville, TN² University of Alberta, Edmonton, AB, Canada³ University of Georgia, Athens, GA⁴ University of Michigan, Ann Arbor, MI

ABSTRACT- The connection of ecological time-series data with mechanistic models poses severe statistical challenges. Typically, the dynamical models themselves are stochastic and nonlinear, only a few state variables can be observed, and these only with error. To deal with these problems, researchers have often been forced to incorporate unpalatable assumptions in their models for the sake of the statistics. We describe the method of simulated composite likelihood, a general approach that requires no such assumptions. We apply the method to measles, whooping cough, and cholera incidence data, using mechanistic continuous-time models for transmission dynamics to obtain insight into the dynamics of epidemics.

Key words: model-fitting, observation error, time-series, epidemiology