...

The permutations would lead to the same sequence of miller indices, but in the context of an unmerged reflection array (i.e. miller.array.sort_permutation), then the data associated with those miller indices would not necessarily be in the same order. As the dataset is then split into two half-datasets, this difference in sort order leads to a different value of the calculate correlation coefficient between those two half datasets:

I guess you need to extend it to also use the reflection intensity in the comparator if you really need reproducible sets :).

Interesting question! The simplest solution is definitively that advocated by Nick, stable sort. But that is arbitrary as far as the intensities are concerned to keep the order of the indices, especially since that indices order is most surely quite arbitrary in the first place. Sorting first on the indices and then on the intensities: is that any less arbitrary to compute the correlation?

Hi, Thanks all for the responses, I think the simplest solution would be to add an option stable=False to flex.sort_permutation that can optionally use std::stable_sort in place of std::sort. Luc: split_unmerged splits an unmerged dataset into two random half datasets, in order to calculate the correlation coefficient between the two half datasets: https://github.com/cctbx/cctbx_project/blob/master/cctbx/miller/merge_equiva... See also Karplus, P. A., & Diederichs, K. (2012). Linking crystallographic model and data quality. Science, 336(6084), 1030-1033: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3457925/ Cheers, Richard Dr Richard Gildea Data Analysis Scientist Tel: +441235 77 8078 Diamond Light Source Ltd. Diamond House Harwell Science & Innovation Campus Didcot Oxfordshire OX11 0DE ________________________________ From: [email protected] [[email protected]] on behalf of Luc Bourhis [[email protected]] Sent: 18 November 2016 16:45 To: cctbx mailing list Subject: Re: [cctbxbb] Unstable platform-dependent sort order for flex.sort_permutation On 18 Nov 2016, at 17:10, [email protected]mailto:[email protected] wrote: a different value of the calculate correlation coefficient between those two half datasets: What does split_unmerged do again? -- This e-mail and any attachments may contain confidential, copyright and or privileged material, and are for the use of the intended addressee only. If you are not the intended addressee or an authorised recipient of the addressee please notify us of receipt by returning the e-mail and do not use, copy, retain, distribute or disclose the information in or attached to the e-mail. Any opinions expressed within this e-mail are those of the individual and not necessarily of Diamond Light Source Ltd. Diamond Light Source Ltd. cannot guarantee that this e-mail or any attachments are free from viruses and we cannot accept liability for any damage which you may sustain as a result of software viruses which may be transmitted in or with the message. Diamond Light Source Limited (company no. 4375679). Registered in England and Wales with its registered office at Diamond House, Harwell Science and Innovation Campus, Didcot, Oxfordshire, OX11 0DE, United Kingdom

Hi Luc, Yes, the random numbers generated by the mersenne twister were the same, however the input tmp_array was in a different sort order, which meant that the output half datasets were platform-dependent: https://github.com/cctbx/cctbx_project/blob/master/cctbx/miller/__init__.py#... I have just committed the necessary changes to add the parameter stable(=False) to flex.sort_permutation(). miller.array.sort_permutation sets stable=True when calling flex.sort_permutation(). This looks to have made the CC1/2 calculations platform-independent. Cheers, Richard Dr Richard Gildea Data Analysis Scientist Tel: +441235 77 8078 Diamond Light Source Ltd. Diamond House Harwell Science & Innovation Campus Didcot Oxfordshire OX11 0DE ________________________________ From: [email protected] [[email protected]] on behalf of Luc Bourhis [[email protected]] Sent: 18 November 2016 17:06 To: cctbx mailing list Subject: Re: [cctbxbb] Unstable platform-dependent sort order for flex.sort_permutation On 18 Nov 2016, at 17:55, [email protected]mailto:[email protected] wrote: Luc: split_unmerged splits an unmerged dataset into two random half datasets, in order to calculate the correlation coefficient between the two half datasets: oh, so you mean that using the same random seed on MacOS and on Linux you did not get the same CC then? -- This e-mail and any attachments may contain confidential, copyright and or privileged material, and are for the use of the intended addressee only. If you are not the intended addressee or an authorised recipient of the addressee please notify us of receipt by returning the e-mail and do not use, copy, retain, distribute or disclose the information in or attached to the e-mail. Any opinions expressed within this e-mail are those of the individual and not necessarily of Diamond Light Source Ltd. Diamond Light Source Ltd. cannot guarantee that this e-mail or any attachments are free from viruses and we cannot accept liability for any damage which you may sustain as a result of software viruses which may be transmitted in or with the message. Diamond Light Source Limited (company no. 4375679). Registered in England and Wales with its registered office at Diamond House, Harwell Science and Innovation Campus, Didcot, Oxfordshire, OX11 0DE, United Kingdom

You might be interested in an alternative method of calculating CC(1/2) from variances, rather than from explicit half-sets, described tersely in this paper 1. Assmann G, Brehm W, Diederichs K. Identification of rogue datasets in serial crystallography. Journal of Applied Crystallography. 2016 Jun;49(3):1021–8.

On 18 Nov 2016, at 18:41, Luc Bourhis wrote:

...

I have just committed the necessary changes to add the parameter stable(=False) to flex.sort_permutation(). miller.array.sort_permutation sets stable=True when calling flex.sort_permutation(). This looks to have made the CC1/2 calculations platform-independent.

Thank you, that’s generally useful an addition. Sorry for not getting the context of your question in the first place!

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb