> Is there a program which can take as input the atomic positions and
> lattice parameters and the associated errors and give as an output an
> estimation of the bond lengths, angles and associated errors? I think
> that doing this would give an overestimation of the errors (for example it
> neglects any correlation in atomic positions) even if such a program
> exists.
This may be a pedantic point, but standard uncertainties (aka esd's)
really should be estimated using the covariance matrix. If you estimate
uncertainties without using the correlation information in the matrix,
your results can be significantly higher or lower (or unchanged)
depending from the best estimates, depending on the sign and magnitude
of the correlation coefficients.
It has been a couple of decades now, but I seem to recall modifying
ORFEE for exactly this purpose -- reading atomic parameters and the
covariance matrix from a refinement program to get distances and angles
with S.U. I suspect Larry Finger's program BONDAN and many others could
be adapted as well.
Brian
