/* MSVARlib Version 1.1 - August 2004 /* First version - May 2004 */ /* Benoit BELLONE */ /* Copyright (C) 2004 by Benoit BELLONE - /* See for more detail the pdf file : "MSVarlib : a new Gauss library to estimate Multivariate Hidden Markov Models" */ /*==========================================================================================================================*/

1. Introduction

MSVARlib "Makov switching Vector Autoregression library" is an open-source basic package designed to model univariate or multivariate time series subject to shifts in regime. This work is fully related to the MSVAR library developed by Krolzig (1998) which provides statistical tools for the maximum likelihood estimation and model evaluation of Markov-Switching Vector Autoregressions.

2. Copyright and Conditions of Use

The GAUSS programs available on the website are written and Copyrighted (c) 2004 by Beno顃 BELLONE, all rights reserved.

The code in this archive is licensed gratis to all third parties under the terms of this paragraph. Copying and distribution of the files in this archive is unrestricted if and only if the files are not modified. Modification of the files is encouraged, but the distribution of modifications of the files in this archive is unrestricted only if you meet the following conditions: modified files must carry a prominent notice stating (i) the original author and date, (ii) the new author and the date of release of the modification, (iii) that there is no warrantee for the code, and (iv) that the work is licensed at no charge to all parties.

If you use the code extensively in your research, you are requested to provide appropriate attribution and thanks to the author of the code.

3. Disclaimer

The code is posted as submitted and is subject to no performance tests. There are absolutely NO GUARANTEES OR WARRANTEES WHATSOEVER, not even the implied warrantees of merchantability or fitness for a particular purpose. No representation is made or implied as to the accuracy or completeness of the programs which may indeed contain bugs or errors unknown to the author. Beno顃 Bellone takes no responsibility for results produced by MSVARlib programs which are used entirely at the reader's risk. This package is by no means finished yet, a comprehensive enhancement "to do list" remains to be done. If you have extended the library, found any problems or have suggestions for improvement please inform the author

4. Installation and Main structures Before installing the package and execute the applications, you need to be ale to run a GAUSS program.

4.1 Version and contributions This package has been developed thanks to the programs of Roncalli (1995), the theoretical work of James Hamliton (1994), and of Krolzig (1997).For the time being, MSVARlib is written in the GAUSS (r) matrix language. It requires the GAUSS(r) 3.2 or a later version. It has been tested on the 3.2, 5.0 and 6.0 versions and has been designed in a WINDOWS 95/NT framework.

4.2 Installation instructions To install and run these programs, you must absolutely create a directory C:\GAUSS\MSVAR where you should unzip the MSVARlib package. This package includes two directories, a readme file and this paper:

- the MSLIB directory includes programs and saved output results, - the DATA directory includes input data files and data sample spreadsheets with convenient templates. - Readme.txt exhibits basic installation instructions

Once unzipped, you should put the first two directories in the following subdirectories:

C:\GAUSS\MSVAR\MSLIB

C:\GAUSS\MSVAR\DATA

To start estimation, open the Gauss(r) program, select C:\GAUSS\MSVAR\MSLIB as your working library. Three programs are available: "MSVAR.prg" which is the main program used to launch a generic multivariate markov switching estimate and two specific programs (rec.prg and IPI.prg) written on this template to provide examples.

4.3 Main files and data organization

In the DATA directory, some sample files are available. They are all ASCII files, saved in the ".txt" format with tab-spaces separators. These data are based upon American survey and quantitative time series starting from January 1960 to December 2004. The first two columns of an input file should start with a month series (m format:1 to 12) and a year series (yyyy format), the following columns should refer to data. No label should be referenced A missing value is represented by a dot ".".