| Re: How to find out the best fitted function ? [message #35783] |
Fri, 18 July 2003 05:25 |
robert.dimeo
Messages: 42
Registered: November 2001
|
Member |
|
|
dinhnq@yahoo.com (Dinh Huong) wrote in message news:<b5f3d715.0307170152.2338aae7@posting.google.com>...
> Hi lists,
>
> How to find out the best fitted function to data before determine the
> best fitted parameters? Is there any routine in IDL or other programs
> can pick out the best fitted function directly instead of previous
> inputting? Since for most cases the parameters are rather easy to
> determine as long as the suitable function types are known.
> Thank for help in advance,
> Nguyen
If I understand your question correctly you are asking for a solution
to a tough problem, i.e. determining the best model that fits your
data. Yes, there are very reliable algorithms for refining a
particular model by varying the parameters (see the MPFIT routines by
Craig Markwardt for instance or LMFIT or CURVEFIT in the IDL
distribution). However determination of the best model is not so easy
unless you are happy making a simple comparison of chi-squared. In
some cases that might be appropriate but in many other cases it is
not. You might want to look into using Bayesian techniques for model
selection. There is a nice introductory book that can help get you
started called "Data Analysis: A Bayesian Tutorial" by Devinder Sivia
from Oxford Science Publications, 1996.
Sorry but I don't know of a simple answer to this one.
Hope this helps.
Rob
|
|
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
