Transcription of Sparse Vector Autoregressive Modeling - arXiv
{{id}} {{{paragraph}}}
Example: tourism industry
Sparse Vector Autoregressive Modeling - arXiv
Sparse Vector Autoregressive ModelingRichard A. Davis, Pengfei Zang, Tian ZhengDepartment of statistics, Columbia UniversityJuly 1, 2012 AbstractThe Vector Autoregressive (VAR) model has been widely used for Modeling temporal de-pendence in a multivariate time series. For large (and even moderate) dimensions, the numberof AR coefficients can be prohibitively large, resulting in noisy estimates, unstable predictionsand difficult-to-interpret temporal dependence. To overcome such drawbacks, we propose a2-stage approach for fitting Sparse VAR (sVAR) models in which many of the AR coefficientsare zero. The first stage selects non-zero AR coefficients based on an estimate of the partialspectral coherence (PSC) together with the use of BIC. The PSC is useful for quantifying theconditional relationship between marginal series in a multivariate process. A refinement secondstage is then applied to further reduce the number of parameters. The performance of this2-stage approach is illustrated with simulation results.
Sparse Vector Autoregressive Modeling Richard A. Davis, Pengfei Zang, Tian Zheng Department of statistics, Columbia University July 1, 2012 Abstract
Loading..
Tags:
Information
Domain:
Source:
Link to this page:
Please notify us if you found a problem with this document: