Re: using tie-to

On Jul 27, 2009, at 11:37 AM, Manfred Hanke wrote:

> Btw, with the set_par_fun function, one can realize even more
> complicated parameter dependencies.


Yes, I am very proud of these:

  43  bbodyrad(1).norm       0     1         7.815466            
5         200
#=>  10.34615*(0.2023128/bbodyrad(1).kT)^4*4.68935/bband(0.5/ 
bbodyrad(1).kT,10/bbodyrad(1).kT)*(constant(1).factor/0.89702)
  44  bbodyrad(1).kT         0     0        0.2335175         
0.05        0.35  keV
  45  powerlaw(1).norm       0     1     8.033949e-06            
0       0.001
#=>  constant(2).factor*plratio(powerlaw(1).PhoIndex,0.5,10)/ 
0.338843*2.11778*1.497945e-5
  46  powerlaw(1).PhoIndex   0     0         1.591079         
-1.5           4
  47  constant(1).factor     0     0          1.31748          
0.5           3
  48  constant(2).factor     0     0        0.3800049            
0           3

Where bband is a S-lang function I wrote to calculate a normalized  
integral of the blackbody flux in a given keV band, and plratio is a S- 
lang function for a normalized integral of a powerlaw flux in a given  
keV band.  The constants 1 & 2 then become fit parameters for the  
0.5-10 keV blackbody flux and 0.5-10 keV powerlaw flux, referenced to  
the previously determined best fit value.  Thus for two separate  
observations I could do error contours for the fluxes in these  
components referenced to the best fit of a single observation, to  
convince a referee that the fluxes in these components did indeed  
statistically differ between the two observations.

Now, if some clever grad student could think of a way of simply using  
a finite set of chi^2 measurements to create an estimate of the fit  
covariance matrix, we could also program something up akin to XSPEC's  
Monte Carlo sampling of flux values.  (I've done it using just the  
single parameter confidence values, but I don't think that works  
particularly well for most situations.)

-Mike


----
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