On Wed, 30 Jan 2013, Henrique Andrade wrote: > Em 27 de janeiro de 2013, Henrique escreveu: > > Em 25 de janeiro de 2013, Allin escreveu: >> >> On Fri, 25 Jan 2013, Henrique Andrade wrote: >>> >>>> I think there's something "strange" with the special >>>> accessor $["roots"]. Please take a look at the following >>>> Hansl code: >>>> >>>> <hansl> >>>> # Gretl doesn't save the "r" matrix >>>> open data9-7 >>>> loop i = 1..2 >>>> arima $i 1 2 ; QNC >>>> matrix r = $["roots"] # overwrites "r" >>>> print r >>>> endloop >>> >>> Yes, that sort of thing ought to work in loops, but it didn't. >>> This should now be fixed in CVS and snapshots. >>> >> >> Now it is working fine. Thanks Allin! > > Now it is not working anymore :-( Using the "test case" I provided > everything works: > > <hansl> > open data9-7 > loop i = 1..2 > arima $i 1 2 ; QNC > matrix r = $["roots"] # overwrites "r" > print r > endloop > </hansl> > > But with my "real world" example it doesn't [...]
> loop for P=0..max_ar_lags --quiet > loop for Q=0..max_ma_lags --quiet > matrix especificacoes = especificacoes | {$P, 1, $Q} > endloop > endloop The trouble is that the first model estimated in your "real world" example doesn't have any ARMA roots (P = Q = 0). For all the other specifications you'll get the $["roots"] matrix OK. Allin Cottrell