| Re: chi-squared of poly_fit and polyfitw results [message #34924] |
Fri, 02 May 2003 20:00  |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
"Chris Boshuizen" <chrisb@newsgroup.usyd.edu.au> writes:
> Hi group,
>
> I was wondering how I relate the sigma values from these routines to a
> chi-squared value. Is it a simple statistical relation? I tried a google
> search and didn't find out how, so I hope someone can offer a suggestion.
As Sergey says, there is not a direct formulaic relation between the
the "sigma" values and the chi-squared value. Generally speaking, the
chi-squared statistic measures goodness of fit. The sigma values
establish confidence regions for each parameter.
It is worth checking the discussions of statistics and fitting in
either Numerical Recipes or Bevington to understand this better. When
estimating useful confidence regions, it is common to vary parameter
values until the chi-squared value changes by a set threshold. So
there is indeed a connection between the those two sets of quantities,
but it is usually not a formula you can write down explicitly.
Happy fitting,
Craig
--
------------------------------------------------------------ --------------
Craig B. Markwardt, Ph.D. EMAIL: craigmnet@cow.physics.wisc.edu
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
------------------------------------------------------------ --------------
|
|
|
|
| Re: chi-squared of poly_fit and polyfitw results [message #34933 is a reply to message #34924] |
Fri, 02 May 2003 04:41  |
Sergey Koposov
Messages: 7
Registered: April 2003
|
Junior Member |
|
|
Hello Chris,
In principle, chi-square is not directly related with sigma.
The function poly_fit have the keyword
CHISQ=variable, which allow to put the chi-square value of fit into the
variable.
You want the chi-squared value of parameters , or of fit?
Cheers,
Sergey
Chris Boshuizen wrote:
> Hi group,
>
> I was wondering how I relate the sigma values from these routines to a
> chi-squared value. Is it a simple statistical relation? I tried a google
> search and didn't find out how, so I hope someone can offer a suggestion.
> Thanks.
>
> Best regards,
> Chris Boshuizen
>
> To reply via email, replace 'newsgroup' with 'student'.
>
>
|
|
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
