Analysis of parameter error in a nonlinear model

Autor: J.B. Mankin, Robert V. O'Neill, Robert H. Gardner
Rok vydání: 1980
Předmět:
Zdroj: Ecological Modelling. 8:297-311
ISSN: 0304-3800
Popis: Solutions of a two-variable nonlinear model were studied by randomly selecting parameters from independent triangular distributions, with maximum and minimum parameter values equal to 10% of the expected value. Small coefficients of variation on parameter values can result in coefficients of variation on model output that are as much as an order of magnitude larger. Establishing limits of acceptable system behavior (error criteria at four points in time reduced the total error and revealed significant variance-covariance relationships between parameters. With parameters drawn from a prespecified variance-covariance matrix, total error over the entire simulation was reduced by a factor of 4.8. Our results can be applied to improve the design of experiments used to estimate parameter values and validate model output.
Databáze: OpenAIRE