Variation diminishing properties and convexity for the tensor product Bernstein operator (Q4027211)

[For the entire collection see Zbl 0745.00065.] About ten years ago, G. Chang and P. J. Davis gave a necessary and sufficient condition for the convexity of Bernstein polynomials over the triangle and let T.N.T. Goodman studied their variation diminishing properties. In the paper, the authors consider corresponding results for tensor products of Bernstein polynomials. In particular, they prove first a variation diminishing theorem and then a convexity theorem for this generalized situation. The variation diminishing theorem states that for any function \(f\) on the unit square \(R\), and for any integers \(m,n\geq 1\), the variation (as introduced by Goodman) of the Bernstein operator \(B_{n,m}\) applied to \(f\) on \(R\) is not bigger than the variation of the ``net'' \(\widehat f\) on \(R\): \(V(B_{n,m}f;R)\leq V(\widehat f_{n,m};R)\). The convexity theorem states that if \(\widehat f_{n,m}\) is convex on \(R\), so is \(B_{n,m}f\).