[Ifeffit] Ifeffit Digest, Vol 49, Issue 20

[Scott Calvin]

Hi Ricardo,

I'll add one thing to Bruce's comment:

Do you care what the sigma^2 is? I mean, other than to hope it's 
"good," is it part of the information you want to know about your sample?

If it's not, and if the uncertainties and correlations reported by 
Artemis suggest that the best fit is simply finding a value of S02 
and sigma2 that are both too low, and if the parameters you are 
interested in do no correlate strongly to S02 or sigma2, then it's 
not a particularly serious problem. You could just set S02 to the 
value found for a standard, or something like that, and as long as 
the fit is still decent in other respects, not worry about it further.

I realize I strung together 3 "if's" there, but it's still a pretty 
common set of circumstances. In your case, the first "if" looks OK 
from your graph--an S02 of 0.8 or so shows error bars from the three 
k-weights that nearly overlap and correspond to a reasonable value of sigma2.

--Scott Calvin
Sarah Lawrence College

At 06:18 PM 3/29/2007, you wrote:
>On Thursday 29 March 2007 16:58, Ricardo Faccio wrote:
> > I am performing a refinement for a  La2CuO4 in powder.
> > I followed Bruce's course, and I tried determine S02 hopping its
> > correlation with sigma^2.
> > I am attaching a graph of this determination. Please see enclosed the file.
> > My question is: Is common to obtain a S02 of 0.625? and is common to obtain
> > a nearly sigma^2=0?
>
>That doesn't seem right.  So2 and sigma^2 are correlated such that a
>low value of one can be compensated by a low value of the other.  I'd
>guess that is happening here.