The RATS Software Forum

That first column in the FEVD table is completely irrelevant to this. Ignore it. It has some potential value in choosing a VAR for forecasting (particularly with one target variable), but that's really it.

The first column is the horizons, I think you wanted to say that first row i.e. the median results? Right?

Once again just to make sure, when we run errors command we get VDCs for each variable, these confidence bands are relevant to those right?

ateeb wrote:So the basic question that i wanted to know answer to is that:

When we run the errors command, it computes the variance decompositions for each variable in the system?

At the point estimates, yes.

ateeb wrote: The @MCFEVDTABLE procedure only computes the lower, median and upper bounds for the estimates?

It computes three percentiles of the FEVD's across simulations of the system.

ateeb wrote: e.g. if forecast error variance as per the Errors instruction is 4.5 for variable x1 due to shock in x2, then the error bands by @MCFEVDTABLE for x1 with a shock to x2 is what should be looked at? correct?

Not really. There's a reason these aren't routinely done. They are what I just described and in practice (with a five variable system) are impossible to interpret in any meaningful way---the median number often has little to do with the point estimate of the FEVD (for reasons I described previously). They simply don't work the same way that IRF error bands do (because they involve squaring the IRF's).

If you want to get a feel for why this doesn't really produce anything useful, look at the behavior of the chi-squared densities as you increase the degrees of freedom. Chi-squareds are squared and summed N(0,1)'s (forecast variance is squared and summed IRF's). Even though the individual Normals are centered at zero, by the time the degrees of freedom reaches 3, the density at 0 is 0.