At 05:25 PM 1/28/98 +0200, you wrote:
>Bill David and Devinder Sivia wrote:
>
>>Well this must sound like heresy to a Rietveld mailing list but ...
>>
>>if you have the correlated integrated intensities from a Pawley
>>refinement then you can either refine on these or use the Rietveld
>>method - it is really the same thing ...
>
>Oh dear, I think they really mean it Armel :-( I hope we are not going to
have
>this old argument with the statisticians that Sakata, Cooper et al. started
>25 years ago against the Rietveld method. To avoid that, and to convince us
>experimentalists, how about refining the same data with the two "equivalent"
>methods and demonstrating that it is "the same thing" in practice ie that
>you get the same estimates of the parameters and standard deviations.
There is a hidden assumption in the statements by Bill & Devinder, namely
that the correlation between the parameters that describe the shape of the
profile and those that describe the Bragg intensities can be safely
ignored. I'm not so sure this is true so that "equivalent" calculations
won't be. I subscribe to the idea that a full matrix Rietveld refinement is
the only "unbiased" method of extracting structural information from a
powder diffraction pattern. Anything else omits possible correlations in
the problem.
Bob Von Dreele