Abstract Data smoothing or regression kernels based on locational entropy embody the principle that observations towards the extremes of the chosen data window should provide less information than those at the midpoint. Weight patterns can be flexible, depending on the choice of prior information density.