1: \begin{abstract}

2:   The classical approach to multivariate extreme value modelling

3:   assumes

4:  that the joint distribution belongs to a

5:   multivariate domain of attraction. This requires 

6:   each marginal distribution be individually attracted to a univariate

7:   extreme value distribution.  

8: %The domain of attraction condition may

9:  % be phrased conveniently in terms of regular variation of the joint

10:   %distribution on an appropriate cone. 

11: An apparently more flexible extremal model for multivariate data

12:    was proposed by Heffernan and Tawn

13:    under which not all the components are required to belong to

14:   an extremal domain of attraction but assumes instead the

15: %Such a model accomodates varying

16:  % degrees of asymptotic dependence between pairs of

17:   %components. 

18: %Instead of starting from the joint distribution, assume

19:   existence of an asymptotic approximation to the conditional

20:   distribution of the random vector given one of the components

21: is extreme. Combined with the knowledge that the conditioning

22:   component belongs to a univariate domain of attraction, this leads

23:   to an approximation of the probability of certain risk regions.

24: The original focus on conditional distributions had

25: technical

26: drawbacks

27:  but is natural in several contexts.  We place this approach in the

28:  context of the more

29: general approach using convergence of measures and multivariate

30: regular variation on cones.\end{abstract}