Marginal models and pruning of association rules

Association rules are a well established tool in data mining techniques which are nowadays used to detect empirical associations in large data sets. To reduce the mass of discovered rules to a manageable number of patterns, a number of selection and pruning methods have been proposed. The use of statistical tests have also been advocated to asses the "interestingness" of an association rule. Along these lines, here we outline how recent developments in the analysis of frequency data, in particular the theory of marginal models, can be applied to this context