Folks,
I’ve had another look at this, and couldn’t find any major discrepancies between my code and Watson’s. Attached are 2 script files – one in Gretl, the other a modified version of Watson’s ‘uc_ngfilt.gss’ (I cut out Watson’s data-reading and calendar routines to keep things focused on the UC filter part). I’ve also attached a spreadsheet showing the resulting versions of figures 2a and 2b from each script. Note that these correspond to the CPIU measure of inflation. They’re both based on 5100 total draws with 100 burn-in, and are well within Monte Carlo error of each other. I can’t see any major differences (apart from MC variation) with Stock & Watson’s corrected figures (also attached; see p. 8).
I’m using Gretl 1.9.8 cvs as of 4/6/12, and Ox for the Gauss replication.
Please let me know if you’re still finding significant differences, or have any other questions or comments.
Best,
PS
From: gretl-users-bounces@lists.wfu.edu [mailto:gretl-users-bounces@lists.wfu.edu] On Behalf Of artur tarassow
Sent: Tuesday, May 15, 2012 3:31 PM
To: Gretl list
Subject: Re: [Gretl-users] unobserved component model with stochastic volatility (UC-SV)
Hey Allin,
I should have been more precise.
Actually I am trying to replicate figure 2a (std. deviation of the permanent component) and 2b (std. deviation of the transitory component) from the S/W paper.
If one compares figures 2a and 2b with the results I've obtained with Peter Summer's gretl code (pdf, attached before) you see that they highly deviate from Stock and Watson's results. But using the original code I can replicate their figures quite well (see attached pdf files).
I thought that maybe somebody on the mailing list has already tried to replicate the S/W paper using their data and Peter Summer's code.
Best,
Artur
2012/5/15 Allin Cottrell <cottrell@wfu.edu>
On Mon, 14 May 2012, Artur Tarassow wrote:
> Dear gretl mailing list
>
> some while ago, Peter Summers posted his translated code from Stock/Watson's
> 2007 paper (originally written in Gauss)
> (URL: http://lists.wfu.edu/pipermail/gretl-users/2011-February/005897.html).
>
> I tried to replicate some work, among them Stock/Watson's paper, but
> unfortunately I do not get the same or at least similar results using Peter
> Summer's code. Even though, the input parameters are as in the original Gauss
> code.What exactly is it that you're comparing between the Stock &
Watson paper and gretl output? And are you using the corrected
S & W output?
http://www.princeton.edu/~mwatson/papers/jmcb_07_stockandwatson_corrected_tables_figures.pdf
Allin Cottrell
_______________________________________________
Gretl-users mailing list
Gretl-users@lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
_______________________________________________
Gretl-users mailing list
Gretl-users@lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users