“…Indeed, from the speed of approximation of smooth functions [5, p. 52, 156] these two types of wavelets are equivalent, and orthogonality to polynomials provides a locally maximal "likeness" to a better rootmean-square approach. At the same time in statistical applications, orthogonality for polynomials of high degree is not required at all [6], in contrast, for example, to the problem of numerical differentiation [7].…”