Abstract
The main framework of multivariate extreme value theory is well-knownin terms of probability, but inference and model choice remain an active researchfield. Theoretically, an angular measure on the positive quadrant of the unit spherecan describe the dependence among very high values, but no parametric form canentirely capture it. The practitioner often makes an assertive choice and arbitrarily fitsa specific parametric angular measure on the data. Another statistician could come upwith another model and a completely different estimate. This leads to the problem ofhow to merge the two different fitted angular measures. One natural way around thisissue is to weigh them according to the marginal model likelihoods. This strategy, theso-called Bayesian Model Averaging (BMA), has been extensively studied in variouscontext, but (to our knowledge) it has never been adapted to angular measures. Themain goal of this article is to determine if the BMA approach can offer an addedvalue when analyzing extreme values.