Compute misc R-factors in scripting

[briantoby]

As you know, the fits for LaB6 are never as good as one might like, but
we run out of parameters in a pseudo-Voigt function. We get Rwp, and
Rexpected for the whole pattern, and Rwp-background. For this phase,
the Rwp and Rexpected are dominated by the huge number of points in the
pattern which are only background. Is it possible to generate an
Rexpected-background? We can argue about whether this is a meaningful
or valid number, but it would in this case be helpful as a measure of
the fit to the peaks, and also as a measure of the signal/noise of the
peaks only. I don’t know that RF or RF2 are quite right for this
purpose, as they don’t consider mis-fits to the shape, only to the
integrated area.

[briantoby]

I think I can see the arguments for why this would be useful for you, but I would also respond that IMO GSAS-II already confuses novices by presenting too many different R-factors (several of which I find rather questionable). Also, R-expected is only meaningful from a statistical perspective and background subtraction breaks any statistical meaningfulness. Perhaps a way to provide this would be to code it in GSAS-II scriptable rather than put it into the GUI.

I am wondering if it might make more sense to compute Rwp, etc. for only points that are, say 2 sigma, from a peak position. That sort of partition of data is valid statistically. Would that serve a similar role for you?