On Thursday, 28 November 2019 16:51:15 Jurgen Bosch wrote:
> Think of completeness with an analogy to turkey.
> Say you happen to find a one-legged turkey (incomplete by conventional
> standard) you could still stuff it and put it in the oven and enjoy 93% of
> the turkey. The 7% missing, who cares? Other than I like both legs of the
> turkey :-)
Don't forget that it helps to evaluate turkey quality
if you can taste matched samples from both sides.
The cc1/2 "cooking comparison" test cannot be conducted
properly if one leg is missing. The turkey is still
just as good but the yumminess is on a subjective scale.
(turkey sampling is still hours away here)
Ethan
>
> Happy Thanksgiving everyone
>
> Jürgen
>
> P.S. back to my wine & ducks to be roasted. @BR, mit Rotkraut &
> Kartoffelknödel
>
> > On Nov 28, 2019, at 4:38 PM, Bernhard Rupp <[email protected]> wrote:
> >
> > Sorry for being late on this thread -
> >
> > but the completeness myth is one of these conventional wisdoms I am
> > seriously questioning and completeness
> > as a global statistic is almost uninformative, short of telling you 'fewer
> > than all recordable reflections up to the reported
> > (likely isotropic) resolution limit given whatever (likely isotropic)
> > cutoff was applied'. Sounds not very clear to me.
> >
> > Kay mentioned already that any information is better than no information,
> > with the caveat that you cannot expect
> > map quality (being an upper limit for model quality - not going into
> > precision vs accuracy issue here) corresponding to
> > the highest resolution reported, which is in reality frequently anisotropic
> > (but not reported or reflected adequately
> > in the PDB reports).
> > We posted some remarks to this effect recently, pointing out that highly
> > incomplete and anisotropic data can still
> > yield limited but useful information as long as your claim remains
> > correspondingly modest. Section 3.4 in
> > http://journals.iucr.org/d/issues/2019/12/00/di5032/index.html
> >
> > Having said that, while random incompleteness is not problematic,
> > systematic reciprocal space incompleteness leads
> > to corresponding systematic real space effects on the map, the simplest
> > being anisotropic data reflecting anisotropic
> > reciprocal map resolution. This is different for example when wedges are
> > missing or absence of serial extinctions makes
> > space group determination more challenging (although we are almost in the
> > age where 'record 360 deg of data and
> > try every SG' works). James Holton has video examples for incompleteness
> > effects and some images are also in my book.
> > https://bl831.als.lbl.gov/~jamesh/movies/
> >
> > Cheers & Happy Thanksgiving, BR
> >
> > PS: A systemic rant regarding data quality representation can be found here
> > https://www.cell.com/structure/fulltext/S0969-2126(18)30138-2
> >
> > ------------------------------------------------------
> > Bernhard Rupp
> > http://www.hofkristallamt.org/
> > [email protected]
> > ------------------------------------------------------
> > All models are wrong
> > but some are useful.
> > ------------------------------------------------------
> >
> > -----Original Message-----
> > From: CCP4 bulletin board <[email protected]> On Behalf Of Kay
> > Diederichs
> > Sent: Monday, November 25, 2019 08:07
> > To: [email protected]
> > Subject: Re: [ccp4bb] Xray-dataset usable despite low completeness ?
> >
> > Dear Matthias,
> >
> > Of course, high completeness is better than low completeness.
> > But as long as your low resolution is pretty much complete, there is no
> > such thing as "too low completeness" at high resolution. Each reflection
> > adds information to the map, and serves as a restraint in refinement.
> >
> > best,
> > Kay
> >
> >
> > On Mon, 25 Nov 2019 14:11:52 +0100, Matthias Oebbeke
> > <[email protected]> wrote:
> >
> >> Dear ccp4 Bulletin Board,
> >>
> >> I collected a dataset at a synchrotron beamline and got the statistics
> >> (CORRECT.LP) after processing (using xds) shown in the attached
> >> pdf-file.
> >>
> >> Do you think this dataset is usable, due to its low completeness? As
> >> you can see in the attached file the completeness is just 50% in the
> >> highest resolution shell, whereas the I over Sigma is above 2 and also
> >> the CC 1/2 (80%) and the R factors (36.8%) have reasonable values.
> >> Furthermore the overall statistic are good regarding R factor, CC 1/2
> >> and I over Sigma. The only problem seems to be the completeness. If I
> >> would set the cut-off at a lower resolution to get good completeness, I
> >> would throw away nearly half of my reflections.
> >>
> >> I'm happy to here your opinion. In Addition to that: The space group is
> >> orthorhombic and the dataset was collected over 120° using fine slicing
> >> (0.1°).
> >>
> >>
> >> Best regards,
> >>
> >> Matthias Oebbeke
> >>
> >>
> >
> > ########################################################################
> >
> > To unsubscribe from the CCP4BB list, click the following link:
> > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
> >
> > ########################################################################
> >
> > To unsubscribe from the CCP4BB list, click the following link:
> > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
>
> ########################################################################
>
> To unsubscribe from the CCP4BB list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
--
entia non sunt multiplicanda praeter necessitatem
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
