1: \begin{abstract}

2: This paper provides the basis for new methods of inference for

3: max-stable processes $\xi$ on general spaces that admit a certain

4: incremental representation, which, in important cases, has a much

5: simpler structure than the max-stable process itself. 

6: A corresponding 

7: peaks-over-threshold approach will incorporate

8: all single events that are extreme in some sense

9: and will therefore rely on a substantially larger amount

10: of data in comparison to estimation procedures based on block

11: maxima. \\ Conditioning a process $\eta$ in the max-domain of

12: attraction of $\xi$ on being \emph{extremal}, several convergence

13: results for the increments of $\eta$ are proved.  In a similar way,

14: the shape functions of mixed moving maxima (M3) processes can be

15: extracted from suitably conditioned single events $\eta$.  Connecting

16: the two approaches, transformation formulae for processes that

17: admit both an incremental and an M3 representation are identified.

18: 

19: \keywords{extreme value statistics; incremental representation; max-stable process; mixed moving maxima; peaks-over-threshold; weak convergence on function space}

20: \ams{60G70}{62G32; 62E20}

21: \end{abstract}

22: