In the spirit of providing guidance for the person who asked about minimum AICc, I would quote Burnham and Anderson (2002) "In general, there are situations where choosing to make inferences based on other than the AICc best model can be justified". Such situations & justifications might include 1) having a simpler model which has a small delta AICc (e.g, <2), or 2) difficulty in estimating parameters for the AICc best model. I do not doubt that multimodel averaging is a good way of dealing with such situations, but it is not the only way. Often a clear exposition of model selection criteria will do the trick.
Dr. Seth W. Bigelow Biologist, USDA-FS Pacific Southwest Research Station 1731 Research Park Drive, Davis California [email protected] / ph. 530 759 1718
