Correlational Tractography Scatter Plot Issue

[Tate Overbey]

Hi Frank,

I am having an issue with the strength and direction of correlations being extracted from data output from correlational tractograhy analyses. I followed the instructions on the website and created a scatter plot of bundle statistics vs an independent variable in excel. The correlations revealed in these plots are always weak and sometimes in the wrong direction. For example, change-in-QA stats loaded from neg.corr.tt.gz show a positive correlation when plotted against the independent variable used in the correlational tractography analysis. I have tried looking at the linear relationship between variables, parametric correlation coefficients, and spearmen coefficients and the result are all the same, occasional positive correlations are being displayed for a relationships that according to DSI Studios should be negative (and visa versa). 

I have tried messing with the correlational tractography parameters using varying t-thresholds (2.5 - 4.0), length thresholds (25 - 35) and thread counts (4 - 64) and maintain an FDR < 0.05 but the correlations remain relatively unchanged, weak and sometimes in the wrong direction. What can I do to correct this issue?

Thanks,

Tate

[Fang-Cheng Yeh]

The most likely cause for wrong directions in QA happens when using DTI data.

Did you use multishell or DTI acquisition?

I would suggest you extract DTI_FA as the connectome database metric to see if it works better.

I can also help run the analysis on my side to check if there is other quality issue if you won't mind sending me the data.

Best regards,

Frank

--
You received this message because you are subscribed to the Google Groups "DSI Studio" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dsi-studio+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dsi-studio/25ba2a89-93b7-4949-a611-ed32c7ea1e22n%40googlegroups.com.

[Daniela Cossio]

I've also been having this issue with some of my output and most of the results are very weak or in the opposite direction. We collected our data via DSI acquisition and I have issues with QA, RD, MD, and AD metrics. I thought it was an issue with the way I was plotting in r studio ( which still may be the case). 

Next week I'm going back through my protocol to see if maybe I made an error somewhere that's leading to this issue. Otherwise I'll keep an eye on this thread in case there are other suggestions

-Daniela 

[Frank Yeh]

The wrong direction issue happens when the effect size is small, or if the locations of findings are very different across subjects.

If the study is a case-control study, then using differential tractography on each subject would help understand the cause.

Frank