Exploring simple methods for the sensitivity analysis of functions in dynamic models
by
John Hearne
School of Mathematical and Geospatial Sciences, RMIT University

While it is standard practice to perform a sensitivity analysis of the parameters used in any model the same cannot be said for model functions. This is partly due to the lack of any automated method with which to explore the sensitivity of functions. In this talk a method is proposed which is both automated and simple to implement. The method involves perturbing each function by a multiplier function. By varying the parameters of the multiplier function the shape of the model functions can be distorted in various ways. Using standard optimisation methods the most sensitive shape distortion of a model function can be identified. Furthermore the functions can be ranked in terms of their sensitivity. The method will be illustrated with two examples.

Date received: December 8, 2009