Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!news.duke.edu!concert!sas!mozart.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: Trading on the Edge
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <CyGJIr.CrL@unx.sas.com>
Date: Sat, 29 Oct 1994 23:20:51 GMT
References:  <38orks$1r2@pipe3.pipeline.com>
Nntp-Posting-Host: hotellng.unx.sas.com
Organization: SAS Institute Inc.
Lines: 44


In article <38orks$1r2@pipe3.pipeline.com>, anwar@pipeline.com (Anwar Shaikh) writes:
|> ... One question I have for you is: how
|> important do you think is the claim that alternate error
|> function s have to be implemented and used in order to deal
|> with financial data (which is often thick-tailed in its
|> distribution)? Several developers with whom I have spoken in
|> the attempt to find software to do this insist that it is
|> irrelevant (which is why their software does not offer this as
|> an option), and that their  own practical experience as
|> consultantts to financial analysts proves it.

If the question is "Are the data really heavy-tailed?" then the
appropriate thing to do is get some data and look at them.

If the questions is "Is least-squares training suitable for
heavy-tailed noise distributions?" then the answer is an emphatic
"No", as demonstrated in the statistical literature on robust
estimation. However, I suspect that robust error functions may be
more liable to overfitting, something I hope to investigate in
coming months.

A few references on robust statistics:

   Hampel, F.R., Ronchetti, E.M., Rousseeuw, P.J., and Stahel, W.A.
   (1986), _Robust Statistics: The Approach Based on Influence Functions_,
   NY: Wiley.

   Hoaglin, D.C., Mosteller, F., and Tukey, J.W., eds. (1983),
   _Understanding Robust and Exploratory Data Analysis_, NY: Wiley.

   Peter J. Huber (1981), _Robust Statistics_, NY: Wiley.

   Kaufmann, L. and Rousseeuw, P.J. (1990), _Finding Groups in Data_,
   NY: Wiley.

   Lawrence, K.D. and Arthur, J.L., eds. (1990), _Robust Regression:
   Analysis and Applications_, NY: Marcel-Dekker.

-- 

Warren S. Sarle       SAS Institute Inc.   The opinions expressed here
saswss@unx.sas.com    SAS Campus Drive     are mine and not necessarily
(919) 677-8000        Cary, NC 27513, USA  those of SAS Institute.
