A method for validation
Stroo, D. (1986). A method for validation. Interne verslagen Nederlands Instituut voor Onderzoek der Zee, 1986(1). NIOZ: Texel. 35 pp.
Part of: Interne verslagen Nederlands Instituut voor Onderzoek der Zee. Nederlands Insituut voor Onderzoek der Zee. , more
|  |
| Abstract |
Aim of this investigation was to give a contribution to the validation of the HOEDE simulation model. The HOEDE model is an ecosystem simulation model, simulating the carbonflux through the Ems-Dollard estuary system. There seems to be a variety of statistical tests to validate a simulation model but most of them have a restriction: autocorrelation. Most statistical tests assume the absence of autocorrelation. Still, autocorrelation is often present in sample data. In this report a description is given of a method to calculate a figure for the "measure of validity" ( usefulness) of a predictor compared to a set of observed values. There certainly is agreement between this figure and the ( subjective) judgement of the HOEDE researchers. Though this figure gives an indication of the usefulness, the judgement of people who are directly involved with the actual process is always better. The method consists of three parts: 1)A figure for the deviation between the actual and simulated data due to a random component. 2) A figure for the correlation. 3) A figure for the relative deviation. The final figure is the average of those three parts. |
|
