Some curvature measures of nonlinear regression

ABSTRACTResidual analysis of linear regression has been studied for nonlinear regression analysis. Toevaluate the significance of the results, curvature measures have been investigated.In application, it was seen that the relationship between the model and the data sets seemed tobe linear. So first of all, linear analysis were done and residual analysis were searched for theappropriate model. Then, for the same data sets, nonlinear regression analysis were studiedand again residuals were analysied for the fitted model. It was seen that, the results of thelinear regression analysis and nonlinear regression analysis were parallel to each other. Toinvestigate the reliability of the residual analysis, curvature measures were calculated. At theend of these calculations, it was understood that the curvature measures were small and theydid not effect the results of the residual analysis. In these applications, SPSS 11.5, Excel2003 and Mathcad 6 were used.Consequently, it was seen that residual analysis of nonlinear regression analysis was noteffected by curvature measures for the attendant data set. This result was interpreted asresidual analysis of nonlinear regression is convenient because curvature is low.Key words: Nonlinear regression, linear regression, residual analysis, curvature measures