In this paper, we illustrate how a generalization of the techniques in association analysis for real-valued data can be utilized to extract coherent functional modules from large microarray data sets.