Non-linear least squares fitting of the theoretical curves to the data
was performed using the LMDER subroutine from the MINPACK library (Garbow
et al., 19xx). Starting values of the parameters were estimated from experience
and tens of different starting values were tried to reduce the possibility
of getting caught in local minima. The basic parameters in the fit were
, L, 
and the amplitude for each peak in the simultaneous fit to all peaks with
the constraint in Eq. 10. Because of small (less than 1%) mechanical hysteresis
in the 
arm
at CHESS that we carefully documented, some of the peak position values
do not index
perfectly and were therefore made parameters. Also, for some of our earlier
data the out of plane resolution 
was not measured carefully and was allowed to be a parameter, although
its range was consistent within xx% with our estimates from slit widths.
Uncertainties in parameter values were estimated from the inverse curvature
matrix for the residual sum of squares.