CONICET | Buscador de Institutos y Recursos Humanos

gdpc is an R package for the computation of the generalized dynamic principal componentsproposed in Peña and Yohai (2016). In this paper, we briefly introduce the problemof dynamical principal components, propose a solution based on a reconstruction criteriaand present an automatic procedure to compute the optimal reconstruction. This solutioncan be applied to the non-stationary case, where the components need not be a linearcombination of the observations, as is the case in the proposal of Brillinger (1981). Thisarticle discusses some new features that are included in the package and that were notconsidered in Peña and Yohai (2016). The most important one is an automatic procedurefor the identification of both the number of lags to be used in the generalized dynamicprincipal components as well as the number of components required for a given reconstructionaccuracy. These tools make it easy to use the proposed procedure in large datasets. The procedure can also be used when the number of series is larger than the numberof observations. We describe an iterative algorithm and present an example of the use ofthe package with real data.