For α∈(0, 1) an α-trimming, P∗, of a probability P is a new probability obtained by re-weighting the probability of any Borel set, B, according to a positive weight function, f≤1/(1−α), in the way P∗(B)=∫Bf(x)P(dx).
¶
If P, Q are probability measures on Euclidean space, we consider the problem of obtaining the best L2-Wasserstein approximation between: (a) a fixed probability and trimmed versions of the other; (b) trimmed versions of both probabilities. These best trimmed approximations naturally lead to a new formulation of the mass transportation problem, where a part of the mass need not be transported. We explore the connections between this problem and the similarity of probability measures. As a remarkable result we obtain the uniqueness of the optimal solutions. These optimal incomplete transportation plans are not easily computable, but we provide theoretical support for Monte-Carlo approximations. Finally, we give a CLT for empirical versions of the trimmed distances and discuss some statistical applications.