Hi,
in the Stata svar command short run restrictions are imposed in the form of the A and B matrix: A*u=B*e, where u is the vector of reduced form innovations and e is the vector of uncorrelated structural shocks normalized to have a unit variance.
Now I always thought that the B matrix is basically where I can get the estimates of the shock standard deviations, if I impose zeros off-diagonal and leave the diagonal elements as free parameters. However, this way the sign of these diagonal elements is not defined. I think the likelihood function takes the same value if you change the sign of any of the diagonal elements. But the standard deviations should be positive.
Question 1: Am I wrong?
Question 2: Can I force Stata to choose the positive solution?
Thanks
in the Stata svar command short run restrictions are imposed in the form of the A and B matrix: A*u=B*e, where u is the vector of reduced form innovations and e is the vector of uncorrelated structural shocks normalized to have a unit variance.
Now I always thought that the B matrix is basically where I can get the estimates of the shock standard deviations, if I impose zeros off-diagonal and leave the diagonal elements as free parameters. However, this way the sign of these diagonal elements is not defined. I think the likelihood function takes the same value if you change the sign of any of the diagonal elements. But the standard deviations should be positive.
Question 1: Am I wrong?
Question 2: Can I force Stata to choose the positive solution?
Thanks