Coverage Problems in Geometrical Probability: Moments and Probability of Coverage for Random Arcs on a Circle

Andrew F. Siegel
University of Washington Business School and Statistics Department

Abstract: Suppose n arcs, each of length a, are placed at random on the edge of a circle of circumference 1. The measure of the circumference that is covered will be a random variable with a very intriguing asymptotic distribution as the number of arcs grows while decreasing their length so that the probability of complete coverage remains fixed. I will review historical background on coverage problems including (1) a manuscript fragment discovered in 1994 at the former estate of Buffon, who created the field of geometrical probability with his needle problem, (2) a duality due to Ailam that links the k-th moment when n arcs are placed to the n-th moment when k arcs are placed, and (3) Stevens' formula for the probability that the circle is covered. Then I will present my own work, including (1) an exact characterization of the distribution of the amount covered by providing the moments and cumulative distribution function in closed form, (2) an application of this distribution to a time-series problem in which the goal is to discover whether apparent multiple sine waves can be distinguished from background noise (extending Fisher's test for a single frequency), and (3) the emergence of the previously unpublished noncentral chi-squared distribution with zero degrees of freedom as the asymptotic coverage distribution.