PARENT SESSION
Wednesday, August 9, 8:00-11:30 am
COS 57 - Ecological modeling II
L-12, Lobby Level, Cook Convention Center
Presiders: A Edwards and K Sakai

Testing stochastic ecological models using AIC.

Richards, Shane^*,1, ¹ Durham University, Durham, Durham County, UK

ABSTRACT- Ecologists are increasingly using ecological theory to develop stochastic models, and then testing the associated theory using model selection techniques. Many ecological studies result in data that is overdispersed. Recognising overdispersion is an important problem when performing model selection because it typically results in overly complex models being selected, which can lead to innapropriate support for a theory. Using simple examples, I show how the stochastic processes that lead to overdispersion can be modeled, and how it can improve model selection performance. I also demonstrate the benefits and costs of dealing with overdispersion using the quasi-likelihood approach, which is becoming increasingly common. Based on my findings I shall provide some general suggestions for dealing with overdispersed data.

Key words: model selection, AIC, stochastic models