hanna
Hi everyone,
I noticed that some of my Rietveld refinements had an Rwp (and Rwp') order of magnitude lower than Rexp (and Rexp'). From what I understand (from 10.1154/1.2179804), this means that uncertainties are overestimated or that I'm overparametrising---the latter being unlikely for powder.
What is the best way to approach this problem? The data is from a DAC experiment, so not great quality. I'm refining against an xy-file, so there is no information on errors, so I don't understand how they can be overestimated. How does TOPAS determine weights in the absence of errors on the data? Some general advice on these things would be appreciated.
Thanks in advance,
Hanna
alancoelho
Hi Hanna
The formula for Rwp is given in Technical_Reference.PDF in the section headed "Criteria of fit".
>How does TOPAS determine weights in the absence of errors
The default weighting is as follows:
weighting = 1 / Max(Yobs, 1);
cheers
alan
johnsoevans
Some of the Rietveld texts discuss why you can sometimes get unusual Chi^2 values (either much greater than one or less than one). Missing experimental uncertainties from the DAC data could be one cause. Was your file initially an .xye file?
You might look at Lynn McCusker et al's Rietveld guidelines paper: J. Appl. Cryst. 1999, 32, 36-50. It discusses some of the common issues with Rwp.
hanna
Thanks both of you! That was a good link; nice paper to refer to.
The files were given to me as a dat file (only x and y columns). My refinement looks visually alright, so I'm not worried about it as such, but it's useful to try to understand the various indicators of good fitting.