Hidden regular variation, second order regular variation and asymptotic independence (Q1424686)

This paper deals with a survey of several related asymptotic properties of multivariate distributions connected with the notion of regular variation. The author gives the definition of multivariate regular variation in terms of vague convergence of measure. Such an orientation is more natural when considering changes of coordinate systems. The author defines precisely hidden regular variation and gives a characterization which roughly says that a random vector \(X\) in \(E\) has a distribution which possesses hidden regular variation if both \(\bigvee_{i=1}^{d}a^{(i)}X^{(i)}\) and \(\bigwedge_{i=1}^{d}a^{(i)}X^{(i)}\) have regularly varying tail probabilities for any positive vector \(a\). Examples are given and it is shown that the hidden regular variation property transforms manageably under change of coordinate system. It is shown that the second order regular variation implies the hidden regular variation. The author interprets the analytical second order condition as convergence of a sequence of signed measures and discusses when this condition can be reduced to convergence of the positive and negative parts of the sequence in the Jordan decomposition of a signed measure. Conditions which guarantee that the second order regular variation condition can be transformed via polar coordinates are obtained.