A fast procedure for model search in multidimensional contingency tables

摘要: SUMMARY A procedure to select the simplest acceptable models for a multidimensional contingency table is proposed. It based on two rules: first, that if model accepted, then all include it are considered be and secondly, rejected, its submodels rejected. Two versions described, one class of graphical models, other hierarchical log linear models. Application both six-dimensional illustrated. The can regarded as an alternative fitting possible made computationally feasible by application rules. generalization proposed Havr'anek (1984), but in many cases considerably faster.