Structural VAR: Home

Contents

Introduction

At this stage I would like to advertise a program that I'm currently working on. It's main objective is to estimate VAR models using various approaches. That is, the software is intended to support:

structural and reduced form VAR models,

estimated with classical Maximum Likelihood or using Bayesian methods,

with or without cointegration restrictions,

allowing for both long run and contemporaneous identifying restrictions.

For the classical approach, bootstrapping is available for most test statistics and estimated parameters as an alternative to relying entirely on asymptotic theory.

In contrast with the programs listed on the the Software Code page, Structural VAR (or SVAR, which is also the Swedish word for answer) is a GUI (Graphical User Interface) based program.

The upper part of the Structural VAR main program window.

The latest beta version, 0.45, has just been released and it's licensed under the GNU General Public License. In its original MatLab code, Structural VAR relies on some of the functions in James P. LeSage's Econometric Toolbox to compute p-values from the normal, F, and chi-square distributions. In addition, it makes use of Christian Beardah's Kernel Density Estimation Toolbox for estimation of density functions. For simplicity and completeness, these functions are already included in the MatLab and source releases which are available from the Download - Source page.

NOTE: If you're using Windows NT4, 2000, XP, 2003, Vista, 9x, or ME, MatLab is not a pre-requisite to use Structural VAR. Just download the Setup Program for Windoze instead, run the file, and follow the installation instructions on the dialogs. Windows XP (and probably 2003 Server and Vista) users are well advised to read the Known Issues page first.