Estimating copula densities through wavelets (Q1017760)

This paper presents a rank-based estimator of a copula density. It is a smoother version of a 3D-histogram. The approach uses a wavelet decomposition of the copula density, starting from an histogram of pairs of standardized ranks. The graph is gradually made smoother through successive applications of the wavelet filter.Some optimality results (in the minimax sense) are proved. Two applications of the proposed methodology are presented.