A method for simulating stable random vectors (Q1965940)

The article deals with simulating a class of multidimensional \(\alpha\)-stable random vectors with dependent components on a computer. The main results of the article are the following: 1. Any \(\alpha\)-stable random vector with discrete spectral measure has the same distribution as a linear combination of vector multiples of one-dimensional i.i.d. \(S_{\alpha}(1,1,0)\) random variables. 2. There exists a sequence of \(\alpha\)-stable random vectors with discrete spectral measure that converges in distribution to a given \(\alpha\)-stable random vector. A FORTRAN subroutine is given which implements the method.