Let the joint density be (1/pi) exp[-(x^2 + y^2)/2] if xy > 0 and
0 if xy < 0; that is, double the joint density of two independent unit
Gaussian random variables in the first and third quadrants, and zero
density in the second and fourth quadrants.
non penserai mica che abbia copiato, no?
mi pare una buona idea, e a occhio direi che funge
ciao