||Cubic Spline Smoothing: A Useful Tool for Curve Estimation.
DeHaan, M. S. ;
||Northrop Services, Inc., Corvallis, OR.;Corvallis Environmental Research Lab., OR.
Curve fitting ;
Regression analysis ;
Least squares method ;
Mathematical prediction ;
Spline functions ;
||Most EPA libraries have a fiche copy filed under the call number shown. Check with individual libraries about paper copy.
Data analysis frequently involves fitting curves to data. Often the investigator has no idea what the underlying functional relationship is and ordinary functions or polynomials fit poorly. This is especially common with experimental data that has a lot of noise and/or measurement error in it. In these cases, spline smoothing can be used to estimate and fit curves with excellent results. Nonparametric cubic spline smoothing is a remarkably accurate and widely applicable approach to curve estimation that has been inexplicably underutilized. This powerful tool can be used in the exploratory, descriptive, and predictive stages of bivariate data analysis. Many examples of curve fitting using spline smoothing are given. Smoothing splines are compared to several other common methods of curve fitting. Methods are detailed for implementing spline smoothing on SAS software.