Error using tlstsq

Submitted by Marieke on

Hi,

I repeatedly encounter an error, irrespective of measure, binary or weighted measures, and irrespective whether graph metrics are normalized with random networks. I attach the workspace I would prefer to use as well as the error code. :)

I would be very grateful for any help! :)

I am using MATLAB R2019a, we are calculating with 211 subjects. Here, I only selected 11 ROIs and 11 thresholds.

Thank you very much!

Marieke

 

 

Johann Kruschwitz

Mon, 08/26/2019 - 09:37

Dear Marieke,

the tlstsq error means that there are most likely NaNs in the data (either in the input data or generated somewhere along the analysis stream) that disamble the GLM to be fit. The setting in the GUI seem to be okay so far. However, it looks as if you specified your costum thresholds in the "Network Threshold" panel. Could it be that something went wrong with the implementation (e.g. more then malf of the field is empty) that causes NaNs when trhesholding the matrices? You could try with the original settings to exclude this error and then proceed looking in your input data.

Best,

Johann

Dear Johann,

thank you so much for your helpful reply!

I have one questions regarding your answer:

This error does no longer appear for all measures although for all measure some participants have NaNs as a result. This is puzzling as our input data has no NaNs. We further checked for fragmentation and only used thresholds that would lead to fully connected networks. Therefore, I would not see a problem in using costumized thresholds?

All in all, we are still wondering why so many single networks lead to NaNs as result for the measures and why for some measures the error occurs and for others it does not.

Your insight would be greatly appreciated! Thank you so much!

All my best,

Marieke

 

 

Johann Kruschwitz

Mon, 11/25/2019 - 14:49

Dear Marieke,
can you ccheck if the calculated graph metrics contain NaNs (i.e., in the interim results folder/GraphVars/ some .mat file) that are entered in the GLM? When GraphVar throws the tlstsq error the graph variables should already be calculated and you could track this. If everything is okay at this step, then it can be a fitting issue of the GLM itself (unlikely). In the latter case you could contact Lea, who programmed the GLM (for this write me an email and I will add Leas new adress in Cc).
Best,

Johann

Forums