If you want even more confusion on the labeling -- take a look at the PDB to mmCIF correspondence mappings for conversion between PDB and mmCIF format.
In the PDB file format under REMARK 200 http://www.wwpdb.org/documentation/format32/remarks1.html#REMARK%20200 there is a line written as REMARK 200 <I/SIGMA(I)> FOR THE DATA SET : and REMARK 200 <I/SIGMA(I)> FOR SHELL : <I/SIGMA(I)> is not defined, but I always read it as the mean of [I/sigma(I)] and not the mean of I / mean of sigma(I). However, using the pdb to mmCIF correspondence guide. http://mmcif.pdb.org/dictionaries/pdb-correspondence/pdb2mmcif.html#REMARK200 This <I/SIGMA(I)> FOR THE DATA SET is linked to the pdb mmCIF dictionary token _reflns.pdbx_netI_over_av_sigmaI http://mmcif.pdb.org/dictionaries/mmcif_pdbx.dic/Items/_reflns.pdbx_netI_over_av_sigmaI.html this is defined as "The ratio of the average intensity to the average uncertainty, /." which sounds like <I>/<sigma(I)> and not <I/sigma(I)>. Likewise, the REMARK 200 <I/SIGMA(I)> FOR SHELL : shell value is linked to the mmCIF token _reflns_shell.meanI_over_sigI_obs using the PDB exchange dictionary give the definition "The ratio of the mean of the intensities of the reflections classified as 'observed' (see _reflns.observed_criterion) in this shell to the mean of the standard uncertainties of the intensities of the 'observed' reflections in this shell." There is a separate pdb mmCIF dictionary token _reflns.pdbx_netI_over_sigmaI http://mmcif.pdb.org/dictionaries/mmcif_pdbx.dic/Items/_reflns.pdbx_netI_over_av_sigmaI.html Which is defined as "The mean of the ratio of the intensities to their standard uncertainties," or < I/SIGMA(I) > So I have never understood why the PDB to mmCIF correspondence maps <I/SIGMA(I)> FOR THE DATA SET to _reflns.pdbx_netI_over_av_sigmaI and not to _reflns.pdbx_netI_over_sigmaI. Or if the PDB file format is supposed to have the mean of <I> / mean <sigI> then why is it written as <I/SIGMA(I)> in the header and not as <I>/<SIGMA(I)>? I never got a satisfactory answer when I asked the deposition staff. I haven't checked the latest version of pdb_extract, but in one of the previous versions, depending on which scaling program you used it would extract either mean (I) / mean (sigmaI) or mean (I/sigmaI) and assign it to the same _reflns.pdbx_netI_over_av_sigmaI token. Regards, Mitch (P.S. There are other strange mappings in the conversion between PDB and mmCIF formats but that is for another day...) -----Original Message----- From: CCP4 bulletin board [mailto:[email protected]] On Behalf Of Anastassis Perrakis Sent: Monday, March 30, 2009 11:40 AM To: [email protected] Subject: Re: [ccp4bb] I/sigma continued On 30 Mar 2009, at 20:30, James Holton wrote: > Frank von Delft wrote: >>> So, what statistic do we want to look at? That depends on what you >>> are trying to do with the data. There is no way for Phil to know >>> this, so it is good that he prints out lots of different >>> statistics. That said, when talking about the data quality >>> requirements for structure solution by MAD/SAD, I suggest looking at >>> I/sigma(I) where: >>> I - merged intensity (proportional to photons) assigned to a >>> reciprocal lattice point (hkl index) >> Does ANY program print this out...? > SCALA calls this "Mn(I/sd)". Sounds like d*TREK calls it "I/sig avg". That is my understanding as well. > > With HKL you compute it "by hand" from the average I and average > "error". hmmm ... from "error" or from "stat."? Should chi^2 be 1 first? > Not sure about XDS... Confusingly, XDS calls that I/SIGMA from what I understand (which as I said before is NOT what SCALA calls I/sigma) Since we only use XDS and (mostly) SCALA in the lab, that is very confusing. I am pretty sure btw that I have myself -wrongly- quoted I/sigma as being <I/sigma(I)> in at least 3-4 papers. And I can bet I am not the only one that did so. <I>/sigma<I> and <I/sigma(I)> are in my view more deterministic labels and will get safer on their way to "Table 1". Tassos > > > -James Holton > MAD Scientist
