Re: fit_flux

On Thu, Feb 22, 2007 at 13:20 -0500, John E. Davis wrote:
> John Houck <houck_at_email.domain.hidden> >On Thu, Feb 22, 2007 at 10:47 -0500, John E. Davis wrote:
> >> What does "fit_flux" do?
> >> 
> >> Suppose I load a grating spectrum C(h) grating ARF A(E), and RMF
> >> R(h,E), where in this context h represents the grating wavelength.
> >> If I flux_corr the spectrum I will get some quantity denoted by F(h).
> >> Then does fit_flux assume that F(h) related to the model s(E) by
> >> 
> >>    F(h) = \int dE R(h,E) s(E) ,
> >> 
> >> independent of the kernel?
> >
> >Yes and no.  The standard kernel is linear, so in that case,
> >your interpretation (which is the one isis uses) seems
> >reasonable.
> 
>   I realise that "flux_corr" depends upon the kernel.  However, is
> "fit_flux" kernel-dependent?


It's kernel-dependent because the kernel's 'compute_kernel'
method is used to compute F(h) from s(E).

However, the intent was that fit_flux should normally be used
with the standard kernel.

Using the standard kernel, the compute_kernel method with
A(E)=1 yields F(h) defined above.


> 
> [...]
> >  isis> flux_corr(1);            
> >  Not Implemented:operation not supported by assigned kernel
> >  
> >That should strongly suggest that it's not a good idea to
> >use fit_flux with piled-up CCD data.
> 
>   Why?  

Are you asking why it's a bad idea to use fit_flux with
piled-up CCD data?  It seems like a bad idea because the
relationship between flux and counts is non-linear. (note that
when I say "piled-up CCD data" I mean imaging-mode CCD data,
not dispersed spectra).

>   Couldn't I use "put_data_flux" to assign a flux to a dataset
>   and then fit that, independent of the kernel?

Yes.  In ambiguous cases, isis mostly assumes the user knows
what they're doing.

>   
> >The isis distribution does also contain the gpileup kernel 
> >(no doubt you're familiar with it, since you wrote it).
> >However, because that kernel has remained experimental, the
> >issue of how it should interact with 'fit_flux' has never been
> >fully resolved.  Is this a good time to revisit that?
> 
> I think fit_flux should be independent of the kernel and assume
> 
>     F(h) = \int dE R(h,E) s(E)
> 
> is the relation between the model-flux s(E) and the data-flux F(h).
> Of course R(h,E) should not include the ARF.

ok.

> 
> [...]
> >Did you try looking at the help page?:
> >
> >   isis> help fit_flux
> 
> Yes, but that did not tell me anything about the kernel or provide an
> equation such as the above one.

I'll work on that.

> 
> [...]
> >> 
> >> Finally, I noticed that the mailinglist archives are not publically
> >> available.  Are there any plans to make them accessable?
> >> 
> >
> >I hadn't considered it, but it seems like a good idea.
> 
> I think that would also give isis a somewhat larger web presence.
> 

It would.  I'll raise the issue with our support staff.

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