The roughness of the model function f（x） to the basis functions has been identified. When the model function is continuous segment, its roughness does not depend on the behavior of the first segment, but depends on ...The roughness of the model function f（x） to the basis functions has been identified. When the model function is continuous segment, its roughness does not depend on the behavior of the first segment, but depends on ＂h＂, the shift in the slope of two consecutive segments. If the distribution of design is uniform, f（x） is continuous segment function, and h is constant, then the maximum roughness is h2/192 obtained at the midpoint of the observations. Suppose that we have a sequence of designs {Pn（x）} then its corresponding distribution {Fn （x）} converges weakly to some distribution F（x）. Let D（f） be a set of discontinuous points off（x）, it is possible to take the limit of the roughness if D（f） has zero （dF）-measure. The behavior of maximum roughness of the discontinuous segment function has been studied by using grid points.展开更多
文摘The roughness of the model function f（x） to the basis functions has been identified. When the model function is continuous segment, its roughness does not depend on the behavior of the first segment, but depends on ＂h＂, the shift in the slope of two consecutive segments. If the distribution of design is uniform, f（x） is continuous segment function, and h is constant, then the maximum roughness is h2/192 obtained at the midpoint of the observations. Suppose that we have a sequence of designs {Pn（x）} then its corresponding distribution {Fn （x）} converges weakly to some distribution F（x）. Let D（f） be a set of discontinuous points off（x）, it is possible to take the limit of the roughness if D（f） has zero （dF）-measure. The behavior of maximum roughness of the discontinuous segment function has been studied by using grid points.
