Re: [ccp4bb]: R-sym I/sigI etc...

[Bernard Santarsiero <bds@uic.edu>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

Unless something is seriously wrong, like the R-factors are high even in
the lower resolution shells, the R-factor is dominated at high resolution
by the denominator, i.e., the F's are weak so the R's are high.  That's
why the chi^2 are a better measure of precision.  

I never independently judge the reliability of a data set by just looking
at the R's.  I do see R's that are very high, greater than 50%, in the
higher resolution shells, and don't worry about it, so long as the chi^2
are fairly uniform throughout the data set.

I also don't agree that the chi^2's should be equal to one.  That's
assuming zero systematic bias in the data sets, and that's rarely the
case.  I typically know more about the beamline and detector that I
collected data at, and inflate the error scale to a reasonable value based
upon experience.   If I wasn't so lazy, I'd collect a great data set on
something that freezes and diffracts well, and determine a reasonable
value for the error scale directly.  (It would be great if the beamline
scientists did this.)  This is analogous to the "ignorance factor" or
"instability constant" that many of us used with diffractometer data to
inflate estimated standard errors in the data (using redundant data).  

Bernie Santarsiero

On Thu, 1 Aug 2002, Roberto Steiner wrote:

> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
> 
> Maybe naive but if you have refined your structure and in your outermost shell 
> you have R factors lower that 59% I would say there is a non random 
> contribution in that shell and therefore no need to cut.
> 
> Roberto 
> 
> > Hello All,
> >
> > I thought this was a settled issue.  That R-syms and mean I/sigI's are a
> > non -ussue and that the best stratety is to include all information.
> >
> > I sent a crystallization manuscript with
> > R-sym in final shell 60.7 Redundancy 8.1 I/sigI 0.9 9all in the final
> > shell).
> >
> > Acta D refused to publish it (a crystallization manuscript) unless I cut
> > my resolution or my redundancy  to get better R-sym at the outermost shell
> > or cut I/sig I at 1.5..
> >
> > The catch is as you collect data to this redundancy the data does get
> > weaker and I/sigI gets poorer. I personally think this fine as we have
> > solved the structure refined etc.. etc.. and everything looks fine.
> >
> > So, do I give up to referees request and publish with lower resolution
> > cutoff / lower redundancy cutoff etc.. of do I fight or do I just not
> > publish the crystallization paper (which is what I am now inclined to
> > do)..
> >
> > I thought I would poll the web before I decide....
> >
> > Thanks.
> >
> > Rams.
> 
> 
> 
> -- 
> Roberto A. Steiner
> York Structural Biology Lab, Department of Chemistry
> University of York, Heslington, York, YO10 5DD, U.K.
> Tel:  (+44) 1904 434531
> Fax: (+44) 1904 410519
> 
>