Copula component analysis for dependence modelling

A copula function can be employed to decompose the information content of a multivariate distribution into marginal and dependence components, with the latter quantified by the mutual information. From this statement, it is possible to state that a link between information and copula theories is valid. On the basis of these results, in the paper we show as it is possibile to use the independent component analysis to estimate the mutual information of a multivariate random sample and, then, to select the model of copula which better interprets the dependence in sample data.