Testing a predictor-corrector integral method for estimating parameters in complex kinetic systems described by ordinary differential equations

An iterative predictor-corrector integral method based on the Himmelblau-Jones-Bischoff technique was tested for estimating kinetic rate constants in complex reaction networks. This approach differs from the original method because the numerical integrals are calculated using the computed concentrations instead of the experimental data. The estimates from the original and the modified techniques were compared and it is shown that the latter gives more precise and accurate parameters even when only short size experimental data are available, the data are affected by large errors, or very complex reaction networks are to be tested. In addition, the confidence bands estimated by applying a rough error analysis show reasonable agreement with the dispersion directly calculated by statistical analysis of the different estimates. (C) 1997 Elsevier Science Ltd.