Regression for mixture models for extremes

In many practical problems the assumption that the data are generated by a single process does not hold. Mixture models can deal with this issue and represent a useful resource also in the context of extreme values. A typical scenario with extreme events is that there are two processes underlying the data: one that occurs with a much higher frequency and another which takes place more rarely but can lead to stronger magnitudes. However, simulation studies and applications to real data show that the rare type may not always correspond to the most extreme events and that the tails of the type-specific data-generating processes are not always identified by this information. Therefore, we aim at creating a new regression model that exploits the type of event as a covariate, together with other variables of interest, such as spatial characteristics.