<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
</head>
<body bgcolor="#ffffff" text="#000000">
<tt>Dear all-<br>
<br>
</tt><tt>Package <i class="moz-txt-slash"><span class="moz-txt-tag"></span>fGarch<span
class="moz-txt-tag"> </span></i>version 2100.78 in R version 2.8.1
(2008-12-22) running on linux 2.6.22.9-91.fc7</tt><br>
<tt><br>
I am trying to fit garch models in the above environment using
garchFit. I fail to get convergence for any of the examples on the
garchFit help page (or for any other simulated or real data for that
matter). Here are the examples from that page where all fail to
converge giving a "singular convergence (7)" message<br>
</tt><tt><br>
R> ## UNIVARIATE TIME SERIES INPUT:<br>
R> ## A numeric Vector from default GARCH(1,1) - fix the seed:<br>
R> N = 200<br>
R> x.vec = as.vector(garchSim(garchSpec(rseed = 1985), n = N)[,1])<br>
R> fit <- garchFit(~ garch(1,1), data = x.vec, trace = FALSE)<br>
R> fit@fit$message<br>
[1] "singular convergence (7)"<br>
R> <br>
R> ## MODELING HIGH/LOW RETURN SPREADS FROM MSFT PRICE SERIES:<br>
R> X.timeSeries = as.timeSeries(data(LPP2005REC))<br>
R> X.mat = as.matrix(x.timeSeries)<br>
R> X.mts = ts(X.mat)<br>
R> fit <- garchFit(100*(SPI - SBI) ~ garch(1,1), data =
X.timeSeries,trace=FALSE)<br>
R> fit@fit$message<br>
[1] "singular convergence (7)"<br>
R> <br>
R> # MODELING HIGH/LOW RETURN SPREADS FROM MSFT PRICE SERIES:<br>
R> X.timeSeries = as.timeSeries(msft.dat)<br>
R> fit <- garchFit(Open ~ garch(1,1), data =
returns(X.timeSeries),trace=FALSE) <br>
R> fit@fit$message<br>
[1] "singular convergence (7)"<br>
R> fit <- garchFit(100*(High-Low) ~ garch(1,1), data =
returns(X.timeSeries),<br>
+ trace=FALSE)<br>
R> fit@fit$message<br>
[1] "singular convergence (7)"<br>
</tt><br>
<tt>I am able to get convergence for these examples and other data
using rgarch::ugarchfit. Using tseries::garch I got "False Convergence"
</tt><tt>even using grad ="numeric" </tt><tt>for all but the 100*(SPI
- SBI) ~ garch(1,1) fit. <br>
</tt><br>
<tt>I tried fiddling with the garchFitControl() but since I do not
understand these setting I failed to find a combination that lead to
convergence in the fits. </tt><br>
<br>
<tt>Perhaps there is a bug in the garchFit code, or perhaps an expert
can suggest suitable control settings so that, at least, the examples
converge. </tt><br>
<br>
<tt>Thank you all for your consideration,</tt><br>
<br>
<tt> Ron Burns</tt><br>
<br>
<pre class="moz-signature" cols="72">--
R. R. Burns
Physicist (Retired)
Oceanside, CA
</pre>
</body>
</html>