ABSTRACT
The purpose of maximum likelihood estimation is to find the most probable parameter estimates for parameters in a model, given a set of data. For example, in an ecological population, maximum likelihood estimation can be used to find an estimate of the probability of survival. However in a parameter redundant model there is not a single parameter estimate for every parameter. This chapter illustrates why parameter redundancy is a problem in modelling and inference. It explores the issues caused by parameter redundancy in maximum likelihood parameter estimation and shows the practical problems with fitting parameter redundant models. The chapter also illustrates that standard errors do not exist for parameter redundant models and shows how model selection is effected by parameter redundancy. Most model selection methods are based on the various candidate models being identifiable. The chapter discusses two model selection methods: information criteria and score test.
