[Ifeffit] Question about the R_factor in artemis/ifeffit

[Matt Newville]

Hi Wayne,

The R-factor is being calculated with both the real and imaginary
parts.  If you're looking at the source code, the R-factor is
calculated in fitfun.f (in src/lib).  There, the sum is over the
elements of the arrays thifit and chifit.  For R-space fits, thifit
contain alternating real,imaginary elements of Delta chi(R)  
(theory-data) within the fit range and chifit contains alternating
real, imaginary elements of chi(R) for the data.  For k-space fits,
thifit contain the elements of the k-weighted Delta chi(k)
(theory-data)  and chifit contais k-weighted chi(k) for the data.

But I also believe that it doesn't matter much, and that due to the
nature of doing FT for a purely real function, an R-value calculated
for the Real part only should be equal to one calculated for the
Imaginary part only and should be very close to the total R-factor.

Hope that helps,

--Matt

On Sat, 1 May 2004, Bruce Ravel wrote:

> On Friday 30 April 2004 06:33 pm, Wayne Lukens wrote:
> > First, thank you for the great software!  I am a former EXAFSPAK
> > user and have been really impressed by Artemis/ifeffit.  It's easier
> > to use, more flexible, and gives more reasonable errors for the fit
> > parameters.
> 
> Well, welcome to the "family" and thanks for the kind words!
> 
> > My question is about the value of the R_factor for R-space
> > fitting.   This seems to be the R_factor for the real component of
> > the Fourier transform only although the fitting is performed on both
> > the real and imaginary components.  Is there a reason that the
> > contribution of the imaginary component to the R_factor is not
> > included?
> 
> A quick perusal of Matt's source code does not suggest that what you
> say is true.  Could you be more specific as to why you think that the
> R-factor is computed incorrectly?
> 
> B
> 
>