A Pearson random walk with steps of uniform orientation and Dirichlet distributed lengths (Q5961872)

The author introduces a variant of the Pearson-Rayleigh random walk of \(n\geq 2\) steps which are random vectors whose orientations are independent and uniform in \(\mathbb{R}^d\) and whose lengths have a Dirichlet distribution whose parameters are all equal to a given positive number \(q\). Two families of walks, named ``hyperspherical uniform'', are obtained for values of the parameter \(q\) equal to \({d\over i}-1\), with \(d\geq i+1\), for \(i=1,2\) respectively. For any number of steps, the endpoint distributions of the latter walks are identical to the distributions of the projection in the walk space \(\mathbb{R}^d\) of a point randomly chosen on the surface of the unit hypersphere of a hyperspace \(\mathbb{R}^k\). The hyperspace dimension associated with each family is equal to \(n(d-i)+i\). The associated probability density function of the endpoint position \(\mathbf{r}\) is given by \(p_{d,n,q}(\mathbf{r})\propto (1-r^2)^\delta\) where \(2\delta=n(d-i)-(d+2-i).\) Four walks, two of two steps and two of three steps, whose endpoints are uniformly distributed in the inside of the unit hypersphere in \(\mathbb{R}^d\) are then found for \(d=2,3,4\) from \(\delta=0\). Three additional uniform walks are obtained from finite integrals of products of powers and Bessel functions of the first kind for Dirichlet distributions of step lengths whose parameters are not all the same.