Hello, if there is no cointegration among variables, it means they have no long run association. In this case the only option we use to explore short run association is VAR. In this case is it possible to use variance decomposition analysis and impulse response function?
Please share some useful resources in this regard.
Of course it is possible. As long as there is some association among your endogenous variables, their variances will contain a mixture of influences: a part of those influences will come from own shocks, and another part will come from shocks on other variables. Recall that the VECM form of a VAR model is just a different reparametrisation of that same model, nothing more.
Yes of course! variance and historical decomposition are applicable after VAR estimation.
It is possible to decompose the forecast error variance into the proportions due to each shock, where the moving average matrices are determined recursively taking into consideration the mean effects of shocks. The elements of coefficient matrices are interpreted as the impulse response.The forecast variance decomposition determines the proportion of the variation of the time series due to the shock.
- Badges
- Science topic