An embedded and charset-unspecified text was scrubbed... Name: not available Url: https://stat.ethz.ch/pipermail/r-sig-finance/attachments/20070714/942f02e3/attachment.pl
Rmetrics gpdFit & fitting an entire distribution
4 messages · Brian G. Peterson, Dale Smith, Eric Zivot
Dale Smith wrote:
Hi there, We are interested in fitting a generalized pareto distribution to the tails of our stock returns, with a kernel fit to the remainder. We looked at gpdFit, but it fits tails, instead of the entire distribution. We thought gpdFit worked like the corresponding function in FinMetrics ? gpd.tail. Alas, gpdFit in fExtremes works differently, by fitting tails only, instead of the entire distribution. I looked at the archives for information on this problem, but didn?t find anything. At this point, we plan to fit each tail separately, and fit the middle of the distribution with another non-parametric method. We are really looking for the quantiles of the fitted distribution. Has anyone done this before, and might be willing to share the algorithm they used?
On first inspection, it looks like the underlying function 'gpd' would
fit the entire distribution.
You may also wish to examine the code for the VaR.gpd function from
package VaR to get another log-liklihood gpd estimate
Cheers,
- Brian
Ok thanks very much, I will do that. Dale
On Jul 14, 2007, at 12:04 PM, Brian G. Peterson wrote:
Dale Smith wrote:
Hi there, We are interested in fitting a generalized pareto distribution to the tails of our stock returns, with a kernel fit to the remainder. We looked at gpdFit, but it fits tails, instead of the entire distribution. We thought gpdFit worked like the corresponding function in FinMetrics ? gpd.tail. Alas, gpdFit in fExtremes works differently, by fitting tails only, instead of the entire distribution. I looked at the archives for information on this problem, but didn?t find anything. At this point, we plan to fit each tail separately, and fit the middle of the distribution with another non-parametric method. We are really looking for the quantiles of the fitted distribution. Has anyone done this before, and might be willing to share the algorithm they used?
On first inspection, it looks like the underlying function 'gpd' would
fit the entire distribution.
You may also wish to examine the code for the VaR.gpd function from
package VaR to get another log-liklihood gpd estimate
Cheers,
- Brian
_______________________________________________ R-SIG-Finance at stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-finance -- Subscriber-posting only. -- If you want to post, subscribe first.
Rene Carmona has his EVENASCA splus library available for free download and this has the gpd function for nonparametrically fitting the middle of the distribution. See http://www.orfe.princeton.edu/~rcarmona/SVbook/evanesce.zip -----Original Message----- From: r-sig-finance-bounces at stat.math.ethz.ch [mailto:r-sig-finance-bounces at stat.math.ethz.ch] On Behalf Of Dale Smith Sent: Saturday, July 14, 2007 8:06 AM To: r-sig-finance at stat.math.ethz.ch Subject: [R-SIG-Finance] Rmetrics gpdFit & fitting an entire distribution Hi there, We are interested in fitting a generalized pareto distribution to the tails of our stock returns, with a kernel fit to the remainder. We looked at gpdFit, but it fits tails, instead of the entire distribution. We thought gpdFit worked like the corresponding function in FinMetrics gpd.tail. Alas, gpdFit in fExtremes works differently, by fitting tails only, instead of the entire distribution. I looked at the archives for information on this problem, but didnt find anything. At this point, we plan to fit each tail separately, and fit the middle of the distribution with another non-parametric method. We are really looking for the quantiles of the fitted distribution. Has anyone done this before, and might be willing to share the algorithm they used? Thanks very much, Dale Smith dtsmith at mindspring.com [[alternative text/enriched version deleted]]