Re: confidence limits

From: John Houck <houck_at_email.domain.hidden>
Date: Thu, 23 Sep 2004 12:33:13 -0400
On Thu, Sep 23, 2004 at 10:59 -0400, Saku Vrtilek wrote:
> 
> 
> Is there some way to limit the incremental change for reduced
> chisq when seeking confidence limits?  I keep getting ``found
> improved fit'' when the change in reduced chisq is minimal, e.g.:
> 
> 1.091 down to 1.090
> 

Yes.  To increase the change-tolerance when searching for 90%
confidence limits, try

  (min, max) = conf (index, 1, 0.01);
  
The last parameter is the delta-chisquare tolerance.  For
details, see the help page:

  isis> help conf
  
One can also adjust the tolerance used in fitting by setting 
a parameter of the minimization algorithm. For the default
Levenberg-Marquardt algorithm, the tolerance is set like so: 
 
  set_fit_method ("marquardt;tol=1.e-3");
  
A list of supported fit-method parameters can be obtained using
the 'help' option.  For example: 

  isis> set_fit_method ("marquardt;help");
  Valid options for subsystem "marquardt" include:
         tol=value     Fractional chisqr acceptance tolerance
       delta=value     Initial numerical derivative step size
    jump_factor=value     Lambda adjustment factor
    max_loops=value     Max number of iterations

Thanks,
-John  
----
You received this message because you are
subscribed to the isis-users list.
To unsubscribe, send a message to
isis-users-request_at_email.domain.hiddenwith the first line of the message as:
unsubscribe
Received on Thu Sep 23 2004 - 12:33:31 EDT

This archive was generated by hypermail 2.3.0 : Fri May 02 2014 - 08:35:44 EDT