Phil Evans wrote:
> *** For details on how to be removed from this list visit the ***
> *** CCP4 home page http://www.ccp4.ac.uk ***
>
>
> Since this seems to be causing endless confusion, here is the
> definition used in Scala for <I/sigmaI> which is what I report to the
> PDB. This is the table column labelled Mn(I)/sd and printed in the
> summary at the end (in the latest version only)
>
>
> After scaling, for each unique reflection h we have several
> observations of the intensity Ih and an estimated "corrected" error
> estimate sd(I)
>
>>From these we calculate a weighted mean <Ih> and an error estimate of
> the mean sd(<Ih>), and a ratio for that unique reflection
> <Ih>/sd(<Ih>)
>
> What is printed as Mn(I)/sd is the mean value of that ratio for all
> reflections (possibly in a resolution bin) ie
>
> <<Ih>/sd(<Ih>)>
>
> This is an estimate of the average signal to noise. Its value does as
> has been frequently pointed out depend on the sd estimates being
> correct, which is always a doubtful proposition, but that's another
> story . . .
>
>
>
> Note the the source code of Scala is in the CCP4 distribution and
> anyone may look in it (look in subroutines ad5sts and prdres)
>
> Best wishes
> Phil Evans
>
>
>
> Edward A. Berry writes:
> >
> >
> > Edwin Pozharski wrote:
> > >
> > > I just want to point out that what is requested upon uploading data to
> the
> > > Protein DataBank is
> > >
> > > NET I OVER AVERAGE SIGMA I
> > >
> > > To me it sounds pretty much like <I>/<sigmaI>.
> > >
> > Yes, ADIT asks for:
> > Net I over average sigma(I) 19
> > And for the last resolution shell there is a different definition:
> > Mean I over sigma(I) (observed) 2.71
> >
> > But when the PDB file is produced, both values are presented as:
> > REMARK 200 <I/SIGMA(I)> FOR THE DATA SET : 19.0000
> >
> > ...|.
> > REMARK 200 <I/SIGMA(I)> FOR SHELL : 2.710
> >
> >
> > As to whether I/Sigma refers to unique reflections or measurements,
> > there is also the question, before or after adding partials?
> > And the cutoff criterion by which we decide which measurements are
> > classified as observations, which I understand should be -3 for
> > scalepack users: is that a cutoff on the raw (partial) measurement
> > or on the full measurement generated by summing partials? I had the
> > impression that the default -3 sigma cutoff in scalepack was for the
> > raw measurements, although reading the current manual I can't find
> > anything to justify that.
> >
> > As users of proprietary (closed-source) software, we depend on the
> > authors for definition of the output values.
> >
> > As a biologist I thought I could just collect data, run the programs,
> > and deposit my structure. Now I'm getting all confused!
> >
> > Ed
> >
> > > Ed.
> > >
> > >
> > >>Anthony Duff wrote:
> > >>
> > >>
> > >>>
> > >>>When I asked about "I on sig I" as to whether one should report: (1)
> > >>><I/SIGI>; (2) <I>/<SIGI>; or (3) I/SIGI, the responses were that it is
> > >>><I/SIGI> that should be reported, although it seems that "I/SIGI"
> > >>>cannot be reasonably interpreted as anything other than <I/SIGI>.
> > >>>
> > >>>Bart Hazes put it most clearly (noting that Jim Pflugrath is uncertain
> > >>>that reporting I/SIGI has much merit all):
> > >>>
> > >>>
> > >>>>- Well because of the first commandment. "Thou shalt report <I/SigI>"
> > >>>>
> > >>>>- Without it a Table 1 wouldn't be a Table 1 would it?
> > >>>
> > >>>
> > >>>As Fred. Vellieux said
> > >>>
> > >>>
> > >>>>You can compute a column containing I/SIGI using SFTOOLS,
> > >>>>then compute its average value.
> > >>>
> > >>>in detail...
> > >>>
> > >>> sftools read mymtzfile.mtz complete # calculate
> > >>> completeness in 20 bins calc col IoSI = col IMEAN col SIGIMEAN /
> > >>> # polish mathematics. # creates IoSI =
> > >>> IMEAN/SIGIMEAN plot col IoSI versus resol # gives average
> > >>> (IoSI) in 20 bins checkhkl # read average (IoSI)
> > >>>
> > >>>
> > >>>
> > >>>Having said all that, I note that colleagues using scalepack are
> > >>>liable to report:
> > >>><I>/<SIGI>, calculating it themselves using "average I" and and
> > >>>"average error" from the last table of the scalepack log file,
> > >>>or
> > >>><I/SIGI>, but not for unique reflections, but for all reflections,
> > >>>taken from the last line of the table "Summary of reflection
> > >>>intensities and R-factors by batch number". A quick investigation has
> > >>>revealed that I/SIGI for all reflections can be very much greater than
> > >>>I/SIGI for unique reflections.
> > >>>
> > >>>It seems to me that it is impossible to obtain the correct <I/SIGI>
> > >>>from a scalepack log file.
> > >>>Is this correct?
> > >>>
> > >>>I'm wondering if there is any consistency in the value to be found in
> > >>>the headers of pdb files following "REMARK 200 <I/SIGMA(I)> FOR THE
> > >>>DATA SET"
> > >>>
> > >>>Anthony
> > >>>
> >
>
