On Sat, 4 May 2013, Gabriela Nodari wrote: Dear all, > > I have notice a great difference between the fevd matrices of the Var > estimated via model -> time series -> var and the Svar estimated via > console. > > Someone could help me to understand why is it so? > Weird. Try this. Do the FEVD matrices look different? <hansl> set echo off set messages off include SVAR.gfn open sw_ch14.gdt genr infl = 400*ldiff(PUNEW) rename LHUR unemp list X = unemp infl var 3 unemp infl F0 = $fevd mod = SVAR_setup("plain", X, const, 3) SVAR_estimate(&mod) F1 = FEVD(&mod) print F0 F1 </hansl> ------------------------------**------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/**servizi/hpp/lucchetti<http://www2.econ.un... ------------------------------**------------------------- _______________________________________________ Gretl-users mailing list Gretl-users(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-users

On Sat, 4 May 2013, Gabriela Nodari wrote: > In this way they are the same.. that is a bad news for me! So via gui the > fevd is not right? The _presentation_ of the FEVD is different in the GUI, in three ways: 1) for a VAR with p variables, the results are divided into p blocks, as opposed to one big matrix. 2) the forecast standard errors for each variable are shown in the first column of each block (these are not present in $fevd); and 3) the shares attributable to each variable are given as percentages rather than as decimal fractions. But a quick check, running the VAR below via the GUI, shows that the underlying FEVD values are the same as those in the $fevd matrix. Allin Cottrell > On 04/05/2013 1:18 PM, "Riccardo (Jack) Lucchetti" < r.lucchetti(a)univpm.it&gt; > wrote: > >> On Sat, 4 May 2013, Gabriela Nodari wrote: >> >> Dear all, >>> >>> I have notice a great difference between the fevd matrices of the Var >>> estimated via model -> time series -> var and the Svar estimated via >>> console. >>> >>> Someone could help me to understand why is it so? >>> >> >> Weird. >> >> Try this. Do the FEVD matrices look different? >> >> <hansl> >> set echo off >> set messages off >> include SVAR.gfn >> >> open sw_ch14.gdt >> genr infl = 400*ldiff(PUNEW) >> rename LHUR unemp >> list X = unemp infl >> >> var 3 unemp infl >> F0 = $fevd >> >> mod = SVAR_setup("plain", X, const, 3) >> SVAR_estimate(&mod) >> F1 = FEVD(&mod) >> >> print F0 F1 >> </hansl> _______________________________________________ Gretl-users mailing list Gretl-users(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-users

On Sat, 4 May 2013, Gabriela Nodari wrote: > Thank you for your answer Allin. I know that the exposition form is > different. But I get different numbers as well.. the same shock explains > different shares. I don't know why is this so.. Well, can you give us an example where you're seeing different results? In the example Jack gave I'm seeing exactly the same output (script and GUI), aside from the mode of presentation. (If you change the Cholesky ordering via the GUI then of course you'll get different results.) Allin > On 04/05/2013 2:13 PM, "Allin Cottrell" <cottrell(a)wfu.edu&gt; wrote: > >> On Sat, 4 May 2013, Gabriela Nodari wrote: >> >>> In this way they are the same.. that is a bad news for me! So via gui >> the >>> fevd is not right? >> >> The _presentation_ of the FEVD is different in the GUI, in three >> ways: >> >> 1) for a VAR with p variables, the results are divided into p >> blocks, as opposed to one big matrix. >> >> 2) the forecast standard errors for each variable are shown in the >> first column of each block (these are not present in $fevd); and >> >> 3) the shares attributable to each variable are given as percentages >> rather than as decimal fractions. >> >> But a quick check, running the VAR below via the GUI, shows that the >> underlying FEVD values are the same as those in the $fevd matrix. _______________________________________________ Gretl-users mailing list Gretl-users(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-users