Controlling risks under different loss functions: The compromise decision problem (Q5899835)

Let A denote a \(\sigma\)-algebra of subsets of a set X and let M(X) denote the Banach lattice of all bounded signed measures on A. Consider an experiment \(E=(X\), A, P), where P denotes a set of probability measures on A. Let L(E) denote the band generated by P in the L-space, M(E) denote its topological dual, Y be a locally compact, separable, metrizable decision space, C(Y) be the Banach lattice of all real valued continuous functions on Y which vanish at infinity, D be the set of all ``sub- decision'' rules, L(C(Y), M(E)) be the space of all continuous linear operators from C(Y) to M(E) and T be a locally convex Hausdorff topology on L(C(Y), M(E)). For a given mapping J: \(D\to L(C(Y)\), M(E)) it is shown that an experiment E is weakly dominated if and only if J(D) is T-compact, giving a characterization of weakly dominated experiments.