谢谢大家的建议，

已经有些眉目了。

/* ==========================================================================================================================*/
/* MSVARlib Version 2.0 - December 2004 */
/* First Version - May 2004 */
/* Benoit BELLONE */
/* Copyright (C) 2004 by Benoit BELLONE. All rights reserved - http://bellone.ensae.net - e-mail: benoit.bellone@ensae.org */
/* To install and adapt this program : MSVARlib_readme.txt */
/* See for more detail : "Classical estimation of Multivariate Markov Switching Model with MSBVARlib " */
/*===========================================================================================================================*/

This version has been deeply enhanced .It introduces a general Multivariate Markov-Switching Regression framework and enables estimates dealing
with 5 different family models :

1, MS-Mean-Variance model : y[t,.]= 1.Beta_S(t)+ u(t) =mu(S(t)) +u(t)

2, MS-VAR(_p) - General model : y(t)=[1 yt-1,..yt-p]*Beta_S(t) +u(t) =x(t).Beta(S(t))+ u(t)

3, MS-VAR(_p) - Switch Intercept model : y(t)= mu(S(t)) + [ yt-1,..yt-p]*Delta(t) +u(t) = 1.*Beta_S(t)+z(t).Delta(t)+ u(t)

4, MS-OLS Mix model : y(t)=x(t)*Beta_S(t) +z(t)*Delta+u(t),

5, MS-OLS General model : y(t)=x(t)*Beta_S(t) +u(t) ,

notice that intercept is included in X(t) see setsample.prg for modifications

I also allows a full description of Covariance regime dependent (resp. independent) Matrices with tree options :
Heteroskedastic variance 1, Homoskedastic variance 2 , Full variance 3.

Automatic or manual parameters initialization are possible. A_priori initialiazing datations can be loaded if needed.

To a complete description of the programs, related tutorials and example programs, the reader should
refer to the attached article : 'Classical estimation of Multivariate Markov Switching Model with MSBVARlib "