And Kay Diedrichs suggested today that averaging out possible errors in the Lorentz correction is a potential benefit of collecting "real redundant" data.
(this would apply to both SR and lab sources). Colin -----Original Message----- From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Colin Nave Sent: 15 May 2013 22:17 To: ccp4bb Subject: Re: [ccp4bb] Fwd: Re: [ccp4bb] reference for "true multiplicity"? Times up! The 3rd case I had in mind was the presence of anisotropy of the anomalous scattering in the presence of a polarised beam. John Helliwell and Pete Dunten suggested this. Dave Waterman identified the general issue with polarisation. All three kept the fun going (as John put it) by just responding to myself. The Templetons and then Schiltz and Bricogne regard this as an opportunity rather than a source of errors. See also the paper by Fanchon and Hendrickson. This leads me to think that there are four ways of handling this sort of issue (or rather among the many ways there are at least four). 1. Pretend these errors don't exist and ignore them. This is maximising the probability of not getting a structure. 2 The pragmatic approach of recognising these errors occur and are difficult to correct. Then devise a data collection and processing strategy where they can be mitigated. This would typically involve a high "real redundancy". 3. Developing a model to handle the errors - for example a model for the absorption errors. This is presumably better in principle but suffers from the danger that the model is wrong. 4. Rename the "errors" as an opportunity - the approach taken by Gerard Bricogne and Marc Schiltz following the Templetons. The latter 3 are all fine and involve selecting an appropriate data collection strategy followed by appropriate data processing. For systematic "errors", it is worth considering, for each type, which of the 4 cases above are the most appropriate way of handling them. My present thoughts are Absorption errors - category 2 above. Detector non uniformity - category 3 Anisotropy in a polarised beam - category 4. As understanding evolves one would hope each systematic error would move at least to category 3. I think Felix Frolow is a supporter of category 3 and, by implication, category 4. Apologies for all the quotation marks round errors, redundancy etc. This of course relates to identifying potential transmission errors while communicating in the English (I guess it would also apply to Latin) language. Colin -----Original Message----- From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Colin Nave Sent: 15 May 2013 10:21 To: ccp4bb Subject: Re: [ccp4bb] Fwd: Re: [ccp4bb] reference for "true multiplicity"? Oh – I seemed to have diverted Frank’s thread. Fortunately most languages themselves are highly redundant, with following characters and words being quite predictable. The entropy and redundancy of English language was analysed by Shannon (with the help of his wife) and he obtained figures of about 1 bit per character rather than log base 2 (27), a redundancy of around 75%. I guess this redundancy helps us put things in context. However, in order to avoid future misunderstandings, I would like to suggest that further communications with CCP4BB be done in Latin which I believe has less ambiguity. I hope people will adopt this helpful suggestion. OK – perhaps not a good idea. More relevant to Frank’s question, I was referring to cases where, for a particular reflection, the path of x-rays through the crystal was altered to average out systematic errors. What type of systematic errors would be mitigated by this? There is one potential addition to the list (absorption errors, detector calibration) I produced but it applies to synchrotron sources rather than the type of x-ray source used in the Acta D paper. I will let others have a think before suggesting it. Colin
