Both can have fairly serious correlation with temperature factors unless
there are regions without reflections at high angles. I do not use fixed
background values as I usually have no idea of the background when the
peaks start getting very dense. Use of fixed background points
artificially removes the correlation, but your choice of the background at
high angle will strongly influence the temperature factor anyway, so the
temperature factors are no more reliable.
The reason why your Chi changes as you add fixed background points is
because GSAS treats them as additional observations. Though valid from an
information theory perspective, it is probably not a wise choice in
practice, since one knows less about a pattern not more when background
values are fixed. Use of a background function is more appropriate from
a statistical point of view as well, because you are treating the problem
by including it in your model rather than adding to your set of
observations.
Brian
! Brian H. Toby, Air Products & Chemicals, Allentown, PA 18195-1501 (USA) !
! net: tobybh@ttown.apci.com voice: (610) 481-4198 FAX: (610) 481-6517 !
! Opinions expressed herein are mine and do not represent my employer. !
! Any resemblance between these opinions and fact is purely accidental. !
On Mon, 14 Nov 1994 AOKDL@ACVAX.INRE.ASU.EDU wrote:
> What's the best way to get rid of the background from a quartz capillary?
> I've tried hand-fitting the background and I didn't like the way chi**2
> depended on my choice of sigmas for the background points (although it
> meant that I could make chi**2 equal 1.00 by "judicious" choice of the
> sigmas!).
> I'm using GSAS. Any suggestions would be appreciated.
>
> - Kurt Leinenweber (AOKDL@ACVAX.INRE.ASU.EDU)
>