Optimized Geometric Attribution

摘要

To address the problem of geometric performance attribution in a recently proposed framework of qualitative characteristics and quantitative properties, the approach reported here is to solve the problem in a two-step fashion. The first step is to define a set of attribution effects in their "pure" geometric form. Because these pure effects will not aggregate in a residual-free manner, however, the second step is to perturb the geometric attribution effects in such a way that they deviate as little as possible, subject to the constraint that there be no residual, from their pure values. The resulting optimized attribution effects represent the most accurate formulation of geometric attribution.