fastVAR-package: Compute large VAR and VARX models

In jeffwong/fastVAR: fastVAR

Description

This package is designed for time series data. Uses fast implementations to estimate Vector Autoregressive models and Vector Autoregressive models with Exogenous Inputs. For speedup, fastVAR can use multiple cpu cores to calculate the estimates. For very large systems, fastVAR uses Lasso penalty to return very sparse coefficient matrices. Regression diagnostics can be used to compare models, and prediction functions can be used to calculate the n-step ahead prediction.

Includes Canada data set from package vars to validate results.

Details

Package:	fastVAR
Type:	Package
Version:	1.9.9
Date:	2012-09-30
License:	GPL
LazyLoad:	yes
Depends: glmnet Suggests: multicore

Very few functions:

• VAR - estimate a standard VAR model

• VARX - estimate a standard VARX model with exogenous inputs

• SparseVAR - use lasso penalty to estimate a VAR model

• SparseVARX - use lasso penalty to estimate a VARX model

• VAR.diag - get regression diagnostics for a VAR or VARX model

VAR, VARX, SparseVAR, and SparseVARX have accompanying prediction methods, which can be invoked by predict(one of the four VAR models, n.ahead)

Author(s)

Jeffrey Wong <[e-mail]>

References

Robert Tibshirani <http://www-stat.stanford.edu/~tibs/lasso.html>

See Also

glmnet, vars

jeffwong/fastVAR documentation built on May 19, 2019, 4:02 a.m.