Abstract

CoCo 1 , a highly advanced program for analysis of complete and incomplete contingency tables, is presented. In the paper a short presentation of CoCo is given. Incremental search by backward elimination and forward selection and the global search procedure from Edwards & Havr'anek (1985) is considered. By incremental search a single minimal acceptable model is identified. By the principles of weak acceptance and weak rejection the class of minimal acceptable models are found in the global search procedure. In CoCo each of the model searches can be done by a single command, or CoCo can be guided through the search in a highly user controlled model selection. 1 CoCo CoCo works especially effectively on graphical models, and some of the commands in CoCo are designed to handle graphical models. Graphical models are for contingency tables log-linear interaction models that can be represented by a simple undirected graph with as many vertices as the table has dimension (Darroch, Lauritz...

Citations