A Bivariate Double Exponential Distribution and its Generalizations

A bivariate (multivariate) double exponential distribution can be generated using differences of cross products of bivariate (multivariate) normal random variables. This distribution possess desirable dependence properties in that it has linear regressions and is indexed by a single correlation parameter whose extreme values correspond to the Frechet bounds. Also, it is easily generated on the computer and offers a useful heavy tailed alternative to the bivariate (multivariate) normal when doing simulations.