Message-ID: <alpine.LRH.2.01.1107041114410.15091@hymn31.u.washington.edu>
Date: 2011-07-04T18:14:41Z
From: Eric Zivot
Subject: semi-nonparametric density function
In-Reply-To: <1309774185.30956.YahooMailClassic@web27406.mail.ukl.yahoo.com>
The SNP density model is coded in in S+FinMetrics. However, George Tauchen at Duke has the source code in Fortran and in C++ on his website and you can try to compile and link it in R (good luck). It is essentially a Graham-Charile expansion (edgeworth type expansion) around the normal distribution involving Hermite polynomials and is a fairly standard expansion used in Statistics.
****************************************************************
* Eric Zivot *
* Robert Richards Chaired Professor of Economics *
* Department of Economics *
* Adjunct Professor of Finance *
* Adjunct Professor of Statistics
* Box 353330 email: ezivot at u.washington.edu *
* University of Washington phone: 206-543-6715 *
* Seattle, WA 98195-3330 * *
* www: http://faculty.washington.edu/ezivot *
****************************************************************
On Mon, 4 Jul 2011, stefan strunz wrote:
> Hi guys,
>
> I am trying to understand Bao's paper "Comparing Density Forecast Models (2007)". Specifically, page 6, link:
> http://www.faculty.ucr.edu/~taelee/paper/BaoLeeSaltoglu_KLIC.pdf
>
> There, he constructs an AR(p) process, with innovations which are from the semi-nonparametric density function of (Gallant and Nychka - "Semi-nonparametric maximum likelihood estimation" 1987). From what I understand, this density "nests" the standard normal distribution, so that if the data is really standard normally distributed, then the density degenerates to N(0,1).
>
> I tried looking for it on rseek.org and google, but couldn't find anything - even on.? Has anybody had some experience with it, or any idea where to find code for it?
> Would greatly appreciate any help!
>
> Best Regards,
>
> Stefan
>
> [[alternative HTML version deleted]]
>
>