- The anisotropy of the
*I*/σ(*I*) distribution; this is determined by taking the local average of*I*/σ(*I*), weighted inversely by reciprocal distance from the reflection whose mean*I*/σ(*I*) is to be determined.No assumptions whatsoever are made about the functional form of the anisotropy of the

*I*/σ(*I*) distribution: it is likely to be totally arbitrary (and not even necessarily ellipsoidal), since it will depend on the reflection redundancy which in practice will vary arbitrarily throughout reciprocal space, depending on the data collection strategy. - The anisotropy of the
*I*/<*I*> distribution, where <*I*> is the expected value of*I*based on the idealized isotropic*I*distribution. This anisotropy is determined by optimization of either the likelihood function proposed by Popov & Bourenkov (2003) or one obtained by marginalising out the true intensity from the French & Wilson (1978) posterior likelihood function. The important difference is that the P & B function does not account for experimental errors whereas the F & W one does.In contrast to the distribution of

*I*/σ(*I*) it is assumed that the anisotropy of the*I*/<*I*> distribution may be represented by a smoothly varying ellipsoidal function of the same form as the standard anisotropic temperature factor (but constrained to the point-group symmetry in the case of merged data). An optimal collection strategy will be one in which the σ(*I*) are varied in such a way as to remove as much of the anisotropy of*I*/σ(*I*) as possible. However any anisotropy of the*I*/<*I*> distribution will always remain since it is a property of the crystal and does not depend on the collection strategy.

If the distribution of *I* is anisotropic then the standard
uncertainty of *I*, *i.e.* σ(*I*), will normally
also have an anisotropic distribution since under default data
collection conditions σ(*I*) is positively correlated with
*I*. Hence in this situation the anisotropy of
*I*/σ(*I*), which determines the anisotropic diffraction
cut-off, will be different (in fact less) than the anisotropy of
*I*/<*I*> which is used to determine the anisotropy
correction of *I* and/or <*I*>.

STARANISO also perform Bayesian estimation of structure amplitudes by
the method of French & Wilson [1978], but using the anisotropic prior in
place of the traditional isotropic prior originally suggested by F &
W. STARANISO incorporates subroutines from the Netlib repository,
in place of the approximate look-up tables used in TRUNCATE, to compute
high-accuracy
parabolic cylinder functions (scaled to avoid numerical
under/overflow issues: Gil *et al.* [2006]) and thereby obtain all the required
moments.

Intensities used in averaging are corrected for the symmetry
enhancement factor *ε*. However only reflections in
lines containing systematic absences due to screw axes (or zones in the
case of glide planes) are corrected for the symmetry factor; any absent
reflections present in the input file are ignored and are not included
in the averages. Reflections lying on pure rotation axes or mirror
planes are not corrected when taking averages (for the explanation of
why it's not correct to do so see Wilson, 1987). Note that the CCP4
library routines for calculating *ε* do not distinguish
properly between these cases.

French, S. & Wilson, K.S. (1978) "On the treatment of negative
intensity observations". *Acta Cryst.* **A**34, 517-525.
See also: "Bayesian
treatment of negative intensity measurements in crystallography".

Gil, A., Segura, J. & Temme, N.M. (2006) "Algorithm 850: Real
parabolic cylinder functions *U*(*a,x*),
*V*(*a,x*)." *ACM Transactions on Mathematical Software
(TOMS)*. **32**, 102-12. See also: "Computing
the real parabolic cylinder functions *U*(*a,x*),
*V*(*a,x*)".

Morris, R.J., Blanc, E. & Bricogne, G. (2003) "On the interpretation
and use of <|*E*|^{2}>(*d**) profiles."
*Acta Cryst.* **D**60, 227-40.

Popov, A.N. & Bourenkov, G.P. (2003) "Choice of data-collection
parameters based on statistical modelling". *Acta Cryst.*
**D**59, 1145-53.

Wilson, A.J.C. (1987) "Treatment of enhanced zones and rows in
normalizing intensities". *Acta Cryst.* **A**43, 250-2.

echo |staraniso HKLIN in.mtz HKLOUT out.mtz >run.log

echo DISP=t,LABF=F_new |staraniso HKLIN in.mtz HKLOUT out.mtz >run.log

echo LABF=F_new 'F(+)_new' 'F(-)_new' DANO_new ISYM_new \ |staraniso HKLIN in.mtz HKLOUT out.mtz >run.log

staraniso HKLIN in.mtz HKLOUT out.mtz <<EOF >run.log DISP=t LABF=F_new F(+)_new F(-)_new DANO_new ISYM_new EOF