Identifiability
Student Objectives for this Chapter
- To understand the problem of model identifiability
- To recognize some common examples and types of identifiability problems
- To use some of the techniques that could be used to recognize identifiability problems
In the previous Chapter (Chapter 19) we considered the selection of the best model. We were able to compare model using the results from analysis using non-linear regression programs. If the model is too big, too many parameters, then we expect increased uncertainty in the parameter values determined by the program. It is possible to build model that are big enough that some parameters values can not be determined at all. These parameters can be called non identifiable. Parameters that can be determined, even if with some considerable variability, are essentially identifiable.
This topic will be discussed by considering definitions, some examples, and numerical and analytical techniques that can be used to explore the identifiability of model parameters.
Next Page
Copyright 2001-3 David W. A. Bourne (david@boomer.org)
This file was last modified:
Wednesday 30 Jul 2003 at 12:24 PM