Speaking of time-varying stuff, and in case anyone's interested, attached is a script that estimates Stock & Watson's unobserved-component stochastic volatility (UC-SV) model from their 2007 JMCB paper. Also attached is a sample data file with a spot commodity price index.
It's pretty much a straight translation of Watson's Gauss code (including his comments), so there are no doubt 'gretl efficiencies' that could be exploited. Comments welcome! PS -----Original Message----- From: gretl-users-bounces(a)lists.wfu.edu [mailto:gretl-users-bounces(a)lists.wfu.edu] On Behalf Of Riccardo (Jack) Lucchetti Sent: Monday, February 14, 2011 10:51 AM To: Gretl list Subject: Re: [Gretl-users] Time Varying Parameters Model On Sun, 13 Feb 2011, Riccardo (Jack) Lucchetti wrote: > On Sun, 13 Feb 2011, Henrique Andrade wrote: > >> Dear Gretl Community, >> I trying to estimate a time-varying parameter model (TVP) using the Kalman >> filter but I'm getting no success [...] > > Henrique, > your state-space model is not correctly specified; in fact, you have no > "regressors": the Phillips curve parameters are your states and the > explanatory variables (output gap etc) form a time-varying H matrix. > > I was going to send you an example, but the script I prepared for you > unearthed a bug in CVS gretl, so please leave us a couple of days to fix this > and then I'll send you a (hopefully) working example. Ok, example attached. It uses some data from Stock & Watson's textbook and it shouldn't be difficult for you to make sense of it and adapt it to your needs. Be sure to use the current CVS version to avoid bugs. Riccardo (Jack) Lucchetti Dipartimento di Economia Università Politecnica delle Marche r.lucchetti(a)univpm.it http://www.econ.univpm.it/lucchetti
uc_ngfilt.inp
Description: Binary data
CRBspot.gdt
Description: Binary data
