Dear all,
I have a anomalous dataset, processed in HKL2000. Scalepack outputs a
file containing the separately merged sets of the Friedel pairs I- and
I+ and their standard deviations sigI+ and sigI-. Scalepack does not
output the averaged intensities (Imean) and the standard deviations
(sigIMean).
The CCP4 program truncate that I use to convert the intensities to
amplitudes requires Imean, I- and I+ and the respective standard
deviations in its input file.
I have now found at least three different methods to generate the
averaged intensities from the Friedel pairs:
1) scalepack2mtz
uses standard deviations for the weights:
weights w = 1/sigI
Imean = (w+*I+ + w-*I- ) / (w+ + w-)
sigImean = 1 / (w+ + w-)
2) Method described in Biomolecular crystallography by Bernhard Rupp, p.
332/333
to average symmetry equivalent reflections
uses variances for the weights:
weight w = 1/sigI^2
Imean = (w+*I+ + w-*I- ) / (w+ + w-)
sigImean = 1 / sqrt(w+ + w-)
3) Method used in cctbx
function miller.set.average_bijvoet_mates() that calls generic
merge.merge_equivalent_obs():
same as methods 2, except that
sigImean is the larger of either
a) sigImean = 1 / sqrt(w+ + w-)
or
b) sigImean = sqrt( wvariance )
where wvariance =
(w+ + w-) / [ (w+ + w-)^2 - (w+^2 + w-^2) ] *
[ w+*(F+ - Imean)^2 + w-*(F- - Imean)^2 ]
What are the advantages and disadvantages of each method? Should method
1 be used at all?
Some example from my dataset:
Reflection (1, 1, 0), space group P3 2 1
I+: 23841.50 sigI+: 634.01 I-: 9628.57, sigI-: 264.75
Method 1: Imean=13815.32, sigImean=186.76
Method 2: Imean=11738.95, sigIMean=244.31
Method 3: Imean=11738.95, sigIMean=7106.47
Thanks a lot!
Cheers,
Markus
--
Markus Meier, Ph.D.
University of Manitoba
Department of Chemistry
144 Dysart Road
Winnipeg, MB, R3T 2N2, Canada
Phone: +1 204 474 7172
Fax: +1 204 474 7608
E-mail: [log in to unmask]
|