Spherical logistic distribution (Q2191435): Difference between revisions

The authors define a spherical logistic distribution on the unit sphere \(\mathbb{S}^{d-1}\) in \(d\)-dimensional Euclidean space, with density function \[ f(y)=C_d(b,\kappa)\frac{e^{\kappa y^T\mu}}{(b-1+e^{\kappa y^T\mu})^2}\,, \] for \(y\in\mathbb{S}^{d-1}\), where \(\mu\in\mathbb{S}^{d-1}\), \(b\geq1\) and \(\kappa>0\) are parameters, and \(C_d(b,\kappa)\) is a normalizing constant. They consider method-of-moments and maximum likelihood estimation of these parameters, illustrating their results with simulations, with special attention given to the case \(d=3\). The paper concludes with an application of this spherical logistic distribution to a real data set.