Local mixtures of the exponential distribution

A new class of local mixture models called local scale mixture models is introduced. This class is particularly suitable for the analysis of mixtures of the exponential distribution. The affine structure revealed by specific asymptotic expansions is the motivation for the construction of these models. They are shown to have very nice statistical properties which are exploited to make inferences in a straightforward way. The effect on inference of a new type of boundaries, called soft boundaries, is analyzed. A simple simulation study shows the applicability of this type of models.