On the monotonicity of positive linear operators (Q1379564)

The main result of this paper concerns positive linear approximation operators of the so-called Feller type \[ K_n(f,x): =Ef(T_{n,x}) =\int_I f(t)dG_{n,x} (t),\;n\in\mathbb{N}, \] where the random variable \(T_{n,x}\) is the arithmetic mean of identically distributed random variables \(X_{i,x}\), \(I=1, \dots, n\) taking values in an interval \(I\) and where \(G_{x,n}\) is the distribution function of \(T_{n,x}\). Besides convolution operators this class contains many classical operators as e.g. the beta, Bleimann, Butzer and Hahn, Meyer-König and Zeller, Picard operator. Theorem 2 says that the inequalities \(K_n\{f,x) \geq f(x)\), \(x\in I\), \(n\in \mathbb{N}\) are necessary and sufficient for \(f\) to be convex. This provides a converse of a well known result of R. A. Khan.