par Liska, Roman
Président du jury Melard, Guy
Promoteur Hallin, Marc
Publication Non publié, 2008-09-10
Thèse de doctorat
Résumé : The main objective of this work is to propose new procedures for the general dynamic factor analysis

introduced by Forni et al. (2000). First, we develop an identification method for determining the number of common shocks in the general dynamic factor model. Sufficient conditions for consistency of the criterion are provided for large n (number of series) and T (the series length). We believe that our procedure can shed

light on the ongoing debate on the number of factors driving the US or Eurozone economy. Second, we show how the dynamic factor analysis method proposed in Forni et al. (2000), combined with our identification method, allows for identifying and estimating joint and block-specific common factors. This leads to a more

sophisticated analysis of the structures of dynamic interrelations within and between the blocks in suchdatasets.

Besides the framework of the general dynamic factor model we also propose a consistent lag window spectral density estimator based on multivariate M-estimators by Maronna (1976) when the underlying data are coming from the alpha mixing stationary Gaussian process.