SVAR parameter standard errors

cjiang3

New Member
Hi everyone,

How to calculate the standard errors of SVAR parameters. I know how to estimate it for contemporaneous parameter. I want to know how about the others? Are SE for SVAR model going to be the same as these in VAR model (other parameters are the same in these two models)?

Thanks

GretaGarbo

Human
SVAR could mean almost anything. (It could mean "answer" in my language!)

Also the standard error could also be anything as long as you have not chosen an estimation method. (Suppose if I chose to use just random numbers and ignore the data. Estimators can be good or bad!) But what about maximum likelihood and the inverse of the information matrix?

But if svar is structural vector autoregression, then I would guess that the parameter estimates are influenced by the structural restrictions you put on the model. However, maybe this paper can be of help.

cjiang3

New Member
SVAR could mean almost anything. (It could mean "answer" in my language!)

Also the standard error could also be anything as long as you have not chosen an estimation method. (Suppose if I chose to use just random numbers and ignore the data. Estimators can be good or bad!) But what about maximum likelihood and the inverse of the information matrix?

But if svar is structural vector autoregression, then I would guess that the parameter estimates are influenced by the structural restrictions you put on the model. However, maybe this paper can be of help.
Thanks for the reply! Yes SVAR is structural autoregression. I only use two-variable system x & y, and restrict one contemporaneous effect (e.g xt on yt) to be zero. Then, in this case, yt will have an impact on xt. I only know how to solve the equations to get parameters of svar by using results from var but how can I get the standard errors for calculating the significance? I do use R but I can only get the standard errors for the contemporaneous parameter not for the others, and the paper didn't talk about it.