Approximation of the yolk by the LP yolk

If nn points are sampled independently from an absolutely continuous distribution with support a convex subset of ℜ2ℜ2, then the center and radius of the ball determined by the bounding median lines (the LP yolk) converge with probability one to the center and radius of the yolk. The linear program of McKelvey (1986) is therefore an effective heuristic for computing the yolk in large samples. This result partially explains the results of numerical experiments in Koehler (1992), where the bounding median lines always produced a radius within 2% of the yolk radius.