Re: [Gretl-users] VAR/Granger Causality

Thanks. For my purpose, your function is actually superior to running the var function in that I can skip estimating the irrelevant equations. Logan -------- Original message -------- From: "Riccardo (Jack) Lucchetti" <r.lucchetti(a)univpm.it&gt; Date: 07/15/2013 5:49 PM (GMT-06:00) To: Gretl list <gretl-users(a)lists.wfu.edu&gt; Subject: Re: [Gretl-users] VAR/Granger Causality On Mon, 15 Jul 2013, Logan Kelly wrote: I suppose a function could do the job quite nicely. The following is an example of the many ways in which this could be done. <hansl> set echo off set messages off function scalar GCtest(series y, series x, list detvars, scalar order, scalar *pv[null]) ols y detvars y(-1 to -order) x(-1 to -order) --quiet omit x(-1 to -order) --silent pv = $pvalue return $test end function # -- open data ---------------------------------- open AWM.gdt --quiet # -- check the function works as intended ------- var 3 LTN STN --quiet # compare GC1 = GCtest(LTN, STN, const, 3) PV = NA GC2 = GCtest(STN, LTN, const, 3, &PV) print GC2 PV # -- apply the function in a loop --------------- smpl +12 ; matrix moo = zeros(12,2) loop for i=1..12 --quiet GC = GCtest(STN, LTN, const, i, &PV) printf "GC test (%2d lags) = %g, (p-value = %g)\n", i, GC, PV moo[i,] = {GC, PV} end loop print moo </hansl> Of course you could prefer incorporating the loop inside the function and have it spit out the matrix you want, it would just be a matter of style. ------------------------------------------------------- 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 -------------------------------------------------------