Bootstrap confidence cones for directional data
Given a sample from a distribution F supported on a p-dimensional sphere, several confidence cones for the ‘mean directional vector’ are proposed. Some methods have been available, but they depend on parametric assumptions. In this paper, nonparametric confidence cones based on the bootstrap are considered, and consistency of these methods is obtained under minimal conditions. We compare the performances of bootstrap methods with the others by Monte Carlo simulations.