Abstract:
The problem of regression curve approximation is regarded as the problem of mean risk minimization on the knowledge of empirical data. Spline — functions are shown to insure asymptotically uniform approximation of a continuous regression curve while polynoms do not have this property.