SVD error

[Smita Menon]

Hello,

I have RAW 2.0.3 installed on a MAC running catalina vs 10.15.6

When I try to run SVD on my SEC-SAXS data i get this error "Initial SVD matrix contained nans or infinities. SVD could not be carried out"

Sorry i tried to search but couldnot find any solution from the forum. Sorry if i missed it somehow.

Thanks for your help.

best,

smita

[Jesse Hopkins]

Hi Smita,

The most likely explanation is that there's something off about your data. You're getting an error that's basically saying there's a problem with the series data you're attempting to do the SVD on. If you send me the data I can take a look.

You might try doing the SVD without error normalization. If you have zero values for the uncertainty of some of your data, that normalization would be causing a problem. To do this, uncheck the 'Normalize by uncertainty' checkbox just above the plot on the left side of the SVD window.

All the best.

- Jesse

----

Jesse Hopkins, PhD

Beamline Scientist

BioCAT, Sector 18

Advanced Photon Source

--
You received this message because you are subscribed to the Google Groups "BioXTAS RAW" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bioxtas_raw...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/bioxtas_raw/407936a3-3ed7-4549-9df5-1651128c700dn%40googlegroups.com.

[Smita Menon]

Hey Jesse,

Hey Jesse,

Sorry couldnot figure out your gmail to send the data privately so i sent it to your  j..@iit.edu email address. But the email bounced back. Is there a secure way that i can share my project details/data with you privately?

As you suggested unchecking the 'Normalize by uncertainty' checkbox just above the plot let it calculate SVD. but the EFA function still doesnt work.

Thank you.

regards,

smita

[Jesse Hopkins]

I ended up directly emailing with Smita to resolve the issue. Just for the record in the mailing list, the issue was that there were q points at the start/end of the data (depending on which profiles were used) with zero uncertainties, which were then throwing off the normalization by uncertainty for EFA/SVD. This was diagnosed simply by opening relevant files in a text editor and inspecting the scattering profile values.

This can be fixed using the 'Adjust scale, offset, q range' tool (described in step 42 of this tutorial: https://bioxtas-raw.readthedocs.io/en/latest/tutorial/s1_sec.html) to truncate the q range appropriately.

All the best.

- Jesse

----

Jesse Hopkins, PhD

Beamline Scientist

BioCAT, Sector 18

Advanced Photon Source

To view this discussion on the web visit https://groups.google.com/d/msgid/bioxtas_raw/e21cc29d-4f14-43c9-b760-4368ab7370d0n%40googlegroups.com.