Representation of the approximational error term and sharp estimates for some local splines (Q808353)

Several specific univariate spline methods are considered to approximate a given derivative of a sufficiently often continuously differentiable function. These spline approximants (mostly linear, cubic or quintic) are defined on a uniform mesh and use only local information about the function. After establishing an integral representation of the approximation error using Peano kernels, it is explicitly shown that the kernels for the spline approximants in question do not change sign on the given interval. Consequently, explicit error formulae and sharp estimates for these splines are derived by an application of the mean value theorem for integrals.