Some estimation theory on the sphere (Q1086943)

This paper looks at distributions on the unit sphere in \({\mathbb{R}}^ q\) which generalize the von Mises-Fisher distribution, with densities of the form f(\(\mu\) 'x) where \(\mu\) and x are unit vectors. Also studied are generalisations of the Scheiddegger-Watson distribution, with densities of the form \(f(\| x_ V\|)\) where \(\| x_ V\|\) is the norm of the projection of x onto some subspace \(V\subset {\mathbb{R}}^ q\) of dimension s, \(1\leq s<q\). Under uniform priors on \(\mu\) and V and under various loss functions, Bayesian estimators of \(\mu\) and V are derived and are found to be Pitman estimators. Also, some infinitesimal robustness characteristics are calculated.