A new statistical model for random unit vectors

This paper proposes a new statistical model for symmetric axial directional data in dimension pp. This proposal is an alternative to the Bingham distribution and to the angular central Gaussian family. The statistical properties for this model are presented. An explicit form for its normalizing constant is given and some moments and limiting distributions are derived. The proposed density is shown to apply to the modeling of 3×3 rotation matrices by representing them as quaternions, which are unit vectors in ℜ4. The moment estimators of the parameters of the new model are calculated; explicit expressions for their sampling variances are given. The analysis of data measuring the posture of the right arm of subjects performing a drilling task illustrates the application of the proposed model.