On Sun, 30 Nov 2008, Tim Cox wrote:
After estimating a VAR, Gretl offers a forecast variance
decomposition under the "analysis" menu. The dialogue box
requires the user to select at least two periods for the
forecast variance decomposition, which makes me wonder why two
periods are required ...
Hmm, under "Forecast variance decomposition" I don't see any
selection of "periods", only an integer horizon representing
how far out the decomposition should go.
Under plain "Forecasts" two periods are called for, namely the
start and end of the range over which a forecast should be
generated. If you have reserved some out of sample observations,
the range defaults to the full out-of-sample range, but you're
free to alter the starting and ending point of the forecast if you
wish.
Have you tried the context Help (i.e. the Help button in the
dialog box for defining the forecast range and type)?
Allin Cottrell