I am estimating a time-series model, which includes 7 time-series (6 "exogenous" and a main ("endogenous") variables). I wanted to estimate a model which allows me to examine impulse response functions.

My theoretical model postulates that the 6 time-series models are correlated in a system with themselves and with "main" (I know the term "endogenous" might be inappropriate here, so let me use "main" instead) variable, as with a traditional set of variables in a VAR-model. However, I also expect there to be a contemporaneous effect between the main variables and the rest, which goes from the 6 variables to the main variable and not the other way.

I am wondering how this system should be estimated. My initial response was to create a SVAR model, where the the restrictions (the Amat matrix in R) only finds an effect from the 6 variables on the main variable. Does this make sense?

Further, I am struggling to think of fitted and predicted values in such a system.

  • $\begingroup$ Probably a SVAR like that is fine. I am not an expert in SVAR, though. $\endgroup$ – Richard Hardy Sep 12 '17 at 15:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.