anisotropy woes

[Axel T. Brunger]

>   I've just started refining something against some disgustingly
> anisotropic data.  The manual shows how to apply an overall anisotropic
> B to Fobs, which is fine in theory, but in practice sigma is ignored in
> refinement, so I would rather not scale up the reflections in the bad
> direction where F/sigma is very low to be the equals of the strong
> reflections in the good direction.  I tried to get around this by
> refining an anisotropic overall B directly before refining the model,
> but it appears that only the trace of the matrix is saved, so as soon
> as Fcalc is updated, all benefit is lost.  Does anyone have a better
> idea?  Should I just alter Fobs and add some sigma-based weighting
> scheme?  Or should I just close my eyes, alter Fobs, and proceed as
> usual?
>
> Thanks in advance,
> Phoebe Rice
> Phoebe at vger.niddk.nih.gov
>

That's right, the trace is of the matrix is applied to the Fcalc array.
The off-diagonal elements are applied to Fobs.  You'll have to use
this modified Fobs in subsequent refinement jobs.  A sigma-based
weighting scheme should allow you to down-weight the reflections
with low F/sigma.

  Axel Brunger