WITHDRAWN: Comparison of Some Dimension-Reduction Techniques in Multivariate Nonstationary Time Series: A Case Study

• Activity Number: 423
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Yennyfer Johana Feo
• Companies: Universidad Nacional de Colombia
• Keywords: Cointegration ; Common factors ; Multivariate time series

Abstract

The main objective of this paper is to compare three dimension reduction techniques for non-stationary time series that are linear transformations of the initial vector arising from the proposals of Gonzalo and Granger (1995), the transformation of Johansen (1991) and methodology in Peña and Poncela (2006). Finally, the procedures are compared through a time series vector of the Colombian economy. We generated vector autoregressive processes VAR(1) and VAR (2) for vectors of size (px1) with p=2,3,4,5 and 9 in the statistical package R. It was noted that both in simulation and in practice, the reduction by all three methods are very similar. However, the similarity between the series obtained by the Johansen method and the method of Peña and Poncela is more evident. This would lead us to believe that the trends of Johansen, is equivalent to linear transformations from nonstationary model