Encompassing in Finite Parametric Spaces

Jean-Pierre Florens, Jean-Francois Richard
Toulouse University, Duke University

Nov 30 1989

The Bayesian notion of encompassing essentially represents a concept of sufficiency among models. Analytical results can only be derived for a limited class of models. In the present note we analyse the case where the parametric spaces associated with the competing models are finite. This restriction greatly simplifies the analysis and, in particular, enables us to study the asymptotic properties of encompassing transitions by means of elementary probability tools.


PDF icon 1989-03.pdf