Hi all,
It is good to keep in mind that
2*[ <|Iplus - Imin|> / < Imean > ]
\approx
[ < (|Iplus - Imin|) / Imean > ]
i.e.
<deltaI>/<I> is NOT equal to <deltaI/I>
In fact, <deltaI/I> might be closer to what one is interested in from an
experimental point of view, as that talks about 'average signal' rather
than the 'signal average'.
Also, when reporting these ratios, it is important to stipulate what is
reported. Some flavours are:
< |dI| >/<I>
< |dI|/I >
< |dF| > <F>
< |dF|/F >
< dF^2 >^1/2 / <F^2>^1/2
.
.
. etc etc etc
They are all related of course, but all have slightly different values.
In the end, what matters is data quality anyway. Having good Bijvoet
ratios on paper doesn't help much if you don't have good data.
Some thoughts on these issues can be found in Acta Cryst. (2005). D61,
1437–1448 and references therein. The simulation tool described in this
paper is available on request.
HTH
Peter
--
Peter Zwart
Beamline Scientist,
Physical Biosciences Division
Berkeley Center for Structural Biology
Building 6, Room 2134
Tel: 510-486-4214, Fax: 510-486-5664
http://bcsb.lbl.gov
Lawrence Berkeley Laboratory
1 Cyclotron Road
BLDG 6R2100
Berkeley, CA 94720, USA.
--
