Why the standard error (PDF simulation) does not converge with different model but same experimental data and setting?

[Leon Wang]

Hi, Bachir,

I built several small models using PDFgui for the same experimental sample, and then applied pymatgen to generate a large super cell. Same settings and scripts were used under the same environment.

For model A, I got standard error = 281, however, for model B, the standard error does not change a lot after 12 hours fitting. The structure of model B looks fine. I discussed with Peter and we are unable to figure out what is going wrong. Any suggestion and comments would be highly appreciated.

Attached are the PDF figure, my input ".pdb" file and run.py scripts (credit to Peter).

Thanks,

Leon

[Robert Koch]

Hi Leon,

It's not clear from this what the issue could be, as you are discussing two models ("model A" and "model B"), but have only included the fit and input structure from a single model (A or B?).

Regards,

Rob

[Leon Wang]

Hi, Bachir,

Sorry to provide you confusing info. Attached are files for model B. Model A looks fine.

Best,

Leon

[Robert Koch]

Hi Leon,

It might be useful to include the files for Model A as well, for direct comparison.

Regards,

Rob

[Bachir Aoun]

Hi Leon,

Comparing standard error values and intensities is irrelevant for the goodness of the fit. Its value is tightly correlated to the experimental data number of points and intensity.
With this being said, the fig you shared is not enough to diagnose the problem. But i have a clue of what's happening. Please share all your files with me (in private if you feel more secure that way) so i can get you a better explanation of what is going on.

Thanks

[Bachir Aoun]

Hi Leon,

The system configuration is correct and all seems good but your limits are off. this was obvious in the fig you shared and i wanted to look at the code before giving you any answer.

your simulation box is way smaller than your experimental data limits. either get a bigger box or set your pdf limits to 25A

regards

[Leon Wang]

Hi, Bachir,

Thanks for your help.

Appreciatively,

Leon

[Bachir Aoun]

Leon, 

Yesterday I answered a bit quickly. You will encounter another problem which is the window function. 

when you will adjust the limits the window function you provide has a size of 17999 data point. This will create a problem when your limit size shrinks your experimental data smaller than then the window function array size.

please be mindful of that and adjust your window function to be smaller than the total experimental data size within the limits your specify.

In the newer fullrmc version i will take care of that and forcefully reset the window function to None but for now you have to do that manually.

thanks 

[Leon Wang]

Hi, Bachir,

Thanks for your reply. I checked my codes and the PairDistributionConstraint codes. I did not see any place to set the limits on. Actually, in the run.py, rmin and rmax have been set.

Any other reason?

Thanks,

Leon

[Bachir Aoun]

Leon,

Yes exactly, you must set the limits.

PDF_CONSTRAINT.set_limits((None,25))

[Robert Koch]

All,

I think the confusion is arising here:

In Leon's code he nominally sets the limits using 

rmin, rmax = 1.0, 20. 

PDF_CONSTRAINT = PairDistributionConstraint(experimentalData=grDataPath,   # pass grDataPath
                                                weighting="atomicNumber",      # X-ray data weighted by Z
                                                limits=(rmin, rmax),           # set data range used
                                                )

Bachir is suggesting one uses 

PDF_CONSTRAINT.set_limits((None,25))

Is there a difference?

Rob

[Leon Wang]

Hi, Rob,

Thanks for your correction.

Best,

Leon

[Leon Wang]

Hi, Bachir,

Thanks.

Leon

[Bachir Aoun]

Technically no. I don't have access to my computer to check that but if there is a difference than it must be fixed ...

Regardless,  use what i gave you and don't forget to reset your window function otherwise you will encounter size errors when computing the constraint.

thanks 

[Bachir Aoun]

Just had a chance to look at the code.

If you set the limits in the initialization the way you did it Leon it won't work. This is on me i will fix it in the next distribution.

Do as I told you please. first set the limits using set_limits method then set your window function

thanks