Presentation by: Peter Verheijen (Delft University of Technology)
Mathematic approach of kinetic parameter estimation from experimental data.
The error model is the starting point for a properly weighted regression. Uncorrelated errors can be estimated by repeated experiments –> Lack-of-fit tests. The normal order used is second order, because that relates to normal distribution. Other orders: choice of weight of extreme residues or in case of non-normal distributions. Also the Maximum-Likelihood method is used. Parameter estimation is effectively an optimization problem, and therefore optimization techniques can be well applied. Some tips and tricks for optimization are given. The use of (confidence) intervals at all levels is a measure of quality and criterion for decision making.
- Download the Abstract: Parameter estimation - Downloaded 604 times