Re: [Gretl-users] (no subject)

On Wed, 23 Jan 2013, Manny B wrote: Testing for Granger causality in non-stationary systems is not as easy as one may think. The obligatory reference here is Toda & Phillips (93) http://ideas.repec.org/a/ecm/emetrp/v61y1993i6p1367-93.html A different solution has been proposed independetly by Toda & Yamamoto (95) and Dolado & Lütkepohl (96) http://ideas.repec.org/p/fth/cemfdt/9424.html, which wouldn't be difficult to code (unsurprisingly, this test is what JMulTi offers). However, I vaguely remember this kind of tests to have disappointingly low power. Morever, practitioners' interest in Granger causation seem to have faded in recent years, so my question is: in the absence of a clear choice on how to perform such tests, wouldn'it be better if someone other than the core developers implemented GC tests in VECMs as a function package? Volunteers welcome! ;) ------------------------------------------------------- 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 -------------------------------------------------------