Sparse Methods for Analysis of Sparse Multivariate Data from Big Economic Databases
Sparse Methods for Analysis of Sparse Multivariate Data from Big Economic Databases

Summary/Abstract: In this paper we present a novel perspective dedicated for sparse highdimensionaldata sets, i.e. data which contain many zeros among coordinates ofobservations. Using jointly, selected sparse methods recently proposed inmultivariate statistics, and kernel density framework for discrete data, we outlinea general perspective for bringing out useful information from big economicdatabases. As a framework for our considerations we take the so-called functionaldata analysis, which originates from Ramsay and Silverman works. In particularwe use functional principal components analysis within 2D density estimationprocedure proposed by Simonoff.

• Details
• Contents