The statistical combination procedure in measures for risk in financial systems

In the literature of risk analysis different synthetic indices are built on the bases of some indicators and in this work we propose to use, alternatively to PCA, a combination statistical procedure. The univariate indices that we use are those proposed by V-lab using a nonparametric combination of dependent rankings. The combination technique may also be considered to perform nonparametric inference, suitable to the treatment of non gaussian distributions as in the case of indices. So we propose to highlight systemic risk in a network of companies performing a nonparametric test to reveal heterogeneity behaviour; in this case the rankings may be used to create different behavioural groups.