ABSTRACT
Any model will have some function that defines that model. For example, in statistics this could be a likelihood, or in compartmental modelling this could be a set of differential equations. An obvious cause of non-identifiability is having too many parameters in a model. When this happens, the model can be reparameterised in terms of a smaller set of parameters, which is termed parameter redundancy. This chapter gives formal definitions of parameter redundancy and identifiability as well as definitions used in this area. It provides general symbolic methods for detecting parameter redundancy and non-identifiability. The chapter demonstrates how one can distinguish between local and global identifiability. It proves identifiability results directly and compares to the general symbolic methods. The chapter also discusses near-redundancy, sloppiness and practical identifiability.
