Re: Full analysis: prolate and oblateRe: Full analysis: prolate and oblateHi Vitaly, I've now fixed this bug in relax. It would still be useful to have a bug report in case someone else encounters the same problem. As I said in the commit message (https://mail.gna.org/public/relax-commits/2011-01/msg00114.html), this bug was recently introduced due to changes for BMRB support. It affects relax 1.3.6, as well as the 1.3 line from revision r11678 to r12174. As this bug is quite big affecting all users of model-free analysis, I will try to release a new version of relax very soon. If you have a repository version, you can type 'svn up' to receive the bug fix. Cheers, Edward On 10 January 2011 11:00, Edward d'Auvergne <edward@xxxxxxxxxxxxx> wrote:
Hi, I can confirm a recently introduced bug in relax whereby the prolate tensor is actually oblate! I will try to solve this as soon as possible and release a new relax version. Strangely this was not caught by the relax test suite. But if you look in the prolate results file, you may see that the diffusion tensor is labelled as 'oblate'. Would you be able to submit a bug report for this? This will help other relax users who encounter the same problem. The submission form is located at https://gna.org/bugs/?func=additem&group=relax. Cheers, Edward On 6 January 2011 19:15, V.V. <vvostri@xxxxxxxxx> wrote:Ok, looks like "diff_min.py" is already setup to do everything I wanted. :-) The problem with the prolate/oblate uncertainty still exists though. I have a feeling that the constraint for Da=>0 or <=0 is removed by the statement fix('diff', fixed=False). As a workaround at the moment I use the following in the "diff_min.py": fix('diff', fixed=False) cdp.diff_tensor.spheroid_type = 'prolate' This seems to do the trick. Vitaly On Wed, Jan 5, 2011 at 16:14, V.V. <vvostri@xxxxxxxxx> wrote:Hello, I am trying to optimize the diffusion tensor outside of "full_analysis.py", to avoid this prolate/oblate issue. I understand that the first round should be models tm0-tm9, followed by model selection, while the latter ones should alternate between the tensor optimization and model-free parameters fitting. I am not entirely sure how to start though. Should I: 1. Take the results from 'local_tm/aic', fix spins, initialize the diffusion tensor and then run grid search + minimization, then fix the tensor and optimize model-free parameters? or 2. Fix spins, initialize the diffusion tensor and perform m0-m9 runs followed by model selection, and after that fix the tensor and optimize MF? This seems to be what "diff_min.py" does. Thank you, Vitaly On Sat, Dec 18, 2010 at 23:25, V.V. <vvostri@xxxxxxxxx> wrote:Hello, I have performed a "full_analysis" run, and the Chi2 values for prolate and oblate models were reported to be identical. I have checked the contents of the /prolate/round_**/opt, and the results file appears to contain the data for the oblate model: ================ <diff_tensor desc="Diffusion tensor" type="spheroid"> <Da ieee_754_byte_array="[144, 224, 168, 43, 17, 72, 82, 193]"> -4792388.6821824461 </Da> <fixed> False </fixed> <phi ieee_754_byte_array="[127, 197, 226, 164, 9, 205, 0, 64]"> 2.1001160508828316 </phi> <spheroid_type> 'oblate' </spheroid_type> <theta ieee_754_byte_array="[157, 42, 47, 69, 3, 109, 252, 63]"> 1.7766144468793421 </theta> <tm ieee_754_byte_array="[131, 104, 243, 90, 176, 172, 60, 62]"> 6.6763176508168525e-09 </tm> </diff_tensor> ================ So, does it mean that the system is converging to the oblate model in the prolate pipe? I thought the Da sign is constrained during the search. Thank you, Vitaly_______________________________________________ relax (http://nmr-relax.com) This is the relax-users mailing list relax-users@xxxxxxx To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-users