example_ms_ar_fit_with_mex.m

Autoregressive Markov Switching Model函数用于评估、仿真及预测自回归的马尔可夫转换模型。可以选择用于模型估计的分布函数。用于研究时间序列结构性变化

% Example Script for MS_AR_Fit.m with mex version of Hamilton's
% Filter. The gain in speed is quite impressive (up to 5x when comparing
% against non mex version).
%
% The mex function was created using a c++ API provided by NR
% (http://www.nr.com/). This makes it easier  by using matrix notation (and
% not pointers notation in the usual mex enviroment).
%
% The mex function was deleloped using matlab 2008a and C++ MS Visual Express 2008 compiler
% (you can get it free at microsoft site).
%
% WARNING. It will NOT compile under LCC. You'll need to recompile the
% files with a proper compiler for it to work on your pc. The name of the
% target file is mex_MS_Filter.cpp (it also uses nr3matlab.h).

clear;

addpath('m_Files');

load Example_Data.mat; % load .mat file

ar=4;                       % Number of lags in autoregressive component
k=2;                        % Number of states 
x=ret;                      % Time series from example.mat
advOpt.distrib='Normal';    % Distribution to use ('Normal' or 't' - default = 'Normal')
advOpt.std_method=1;        % Method for standard error calculation
advOpt.useMex=1;            % Use mex version of likelihood?

[Spec_Output]=MS_AR_Fit(x,ar,k,advOpt); % fit the model

rmpath('m_Files');