Statistical Modelling 11 (2011), 523–534

On the reduced-rank model with leading index

Yu-Pin Hu
Department of International Business Studies,
National Chi Nan University
No. 1 University Road
Puli, Nantou, Taiwan 545
Republic of China
eMail: huyp@ncnu.edu.tw

Abstract:

The index model and the reduced-rank model are two useful models which adopt different views to explore a vector autoregressive (VAR) model structure. However, we expect that there will be more information when the two models are implemented simultaneously. This paper combines these two models to propose a VAR model with constraints. The model extracts the leading index and explores the reduced-rank structures simultaneously. In this study, theoretical results are presented, which include a necessary and sufficient condition for a transformation for parsimonious parameterizations and model identification, the maximum likelihood estimations and the likelihood ratio test. The empirical results for the analysis of famous US hog data using the proposed model display the unification of several results in the literature, revealing that the novel model is reasonable and works quite well.

Keywords:

index model; reduced-rank model; vector autoregressive model

Downloads:

Gauss code and example data in zipped archive
back