YGGDRASIL-A statistical package for learning Split Models

There are two main objectives of this paper. The first is to present a statistical framework for models with context specific indepen­ dence structures, i.e. conditional independen­ cies holding only for specific values of the con­ ditioning variables. This framework is consti­ tuted by the class of split models. Split mod­ els are an extension of graphical models for contingency tables and allow for a more so­ phisticated modelling than graphical models. The treatment of split models include estima­ tion, representation and a Markov property for reading off those independencies holding in a specific context. The second objective is to present a software package named YG­ GDRASIL which is designed for statistical in­ ference in split models, i.e. for learning such models on the basis of data.