VAR estimation in Matlab and STATA

don

2012 9월 8

0 답변

조회 수: 5 (30일)

0 개 추천

I am running VAR(2) model with two variables, say A and V, with some restrictions on the estimating matrix. Following is the code that I am running in Matlab.

if true
restriction = [1,0;1,1];
Spec = vgxset('n',2,'ARsolve',{logical(restriction),logical(restriction)});
Y = [ A, V ];
[EstSpec,EstStdErrors,LLF,W] = vgxvarx(Spec,Y,[],[],'CovarType','full');
end

The results in Matlab and STATA by comparison are

Model  : 2-D VAR(2) with No Additive Constant
         Conditional mean is AR-stable and is MA-invertible
         Standard errors without DoF adjustment (maximum likelihood)
     Parameter   Matlab Value   Stata Value
-------------- -------------- --------------
    AR(1)(1,1)        1.03565      1.092692
         (1,2)              0     -8.70e-17                         
         (2,1)       0.126034      .0322713
         (2,2)        1.30409      1.768418
    AR(2)(1,1)     -0.0622634     -.1178665  
         (1,2)              0     -4.37e-15
         (2,1)      -0.129925     -.0368959
         (2,2)      -0.322161     -.7805202
        Q(1,1)    0.000275459     .00010679                 
        Q(2,1)   -0.000123201    -7.526e-06                 
        Q(2,2)   9.06242e-005     8.522e-06                
-------------- -------------- --------------

I am using MATLAB Version 7.11.0.584 (R2010b) with Econometrics Toolbox Version 1.4 and Stata 11.1 for Windows. Can anybody help me why they are different.