Re: simultaneous fitting

From: John Houck <houck_at_email.domain.hidden>
Date: Mon, 2 Feb 2009 20:56:58 -0500
On Mon, Feb 02, 2009 at 17:49 -0500, Herbert Pablo wrote:
> Hey,
> 
> Thanks for the reply.
> 
> First to make things a little more clear. All the combining is done in one 
> script and then a fitting script is run. To that end, this is what I'm curious 
> about. From what I understand, once I combine two data sets into one, there is 
> only 1 piece of information now and that is what I want to fit. If that is the
> case why would I have to specify wavelength ranges for both sets of data if 
> they are already combined?
> 

The combine_datasets function merely labels the datasets to
indicate how they are to be combined later, during the fit.

The sums are computed "on-the-fly" when the fit-statistic is
computed and not when you invoke combine_datasets.  The sums
are not saved -- they are recomputed as needed and then
discarded.  (To be more precise, isis computes the dataset sum
once at the beginning of the fit because that sum won't change
during the fit.  The sum of models is recomputed with each
iteration).

Because the sums are computed on-the-fly, it is important that
the datasets being combined have matching grids and have the
same bins noticed.

> 
> On a secondary note, I think there is an issue with my script to combine 
> datasets. If I run the same script using counts everything is fine and I get 
> what I would expect. However, if i run the script with flux, something strange 
> is happening. If I plot the plus and minus orders before I combine them, then 
> they look fine. But, if I plot the plus and minus orders after I combine, one
> of them is fine and the other is empty (horizontal line at 0). I'm sure this
> is affecting my fit, but honestly I don't know what the problem is. For
> brevity, I'll just write the combining part of the code below:
>
[...]
> 
> If you see any obvious errors here let me know.
> 

I cannot reproduce the problem using the code you sent.

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 Mon Feb 02 2009 - 20:57:15 EST

This archive was generated by hypermail 2.2.0 : Wed Feb 18 2009 - 16:33:02 EST