Interpreting FBA outputs

Dear experts,
I’ve been having some difficulty interpreting FBA outputs, especifically abs_effect.mif and fwe_pvalue.mif. I am working on two sets of data ( patients x controls ), investigating group differences with the Multi-tissue CSD pipeline. Mrcalc command generated percentage_effect.mif files for each contrast and Fixel2tsf command generated fdc_fwe_pvalue.tsf and fdc_abs_effect.tsf files in order to display results with streamlines.
In order to visualize the results, I’m opening a track file with 200k streamlines in the tractography plot tool. On the scalar file option, I’m opening fd_percentage_effect.tsf, with the results displayed below:

Whenever I run abs_effect.tsf file in scalar file options, the results are quite similar from the image above:

The regions displayed in different color ( mainly the splenium of the corpus callosum ) are in accordance with our hypothesis, but when I proceed to threshold the image with fwe_pvalue.tsf (0.95 value in the the value box on the left hand side), the results are completely gone and no streamlines are generated.
I hope someone can help me understand the meaning of the output generated with FBA ( abs_effect.mif, fwe_pvalue.mif), mrcalc (percentage_effect.mif) and fixel2tsf (fdc_fwe_pvalue.tsf and fdc_abs_effect_size.tsf) and how to use these files to make sense of my results.

Thank you!

Hi Thiago,

The regions displayed in different color ( mainly the splenium of the corpus callosum ) are in accordance with our hypothesis, but when I proceed to threshold the image with fwe_pvalue.tsf (0.95 value in the the value box on the left hand side), the results are completely gone and no streamlines are generated.

This would probably suggest that while there’s some sort of effect there, it’s not large enough to reach statistical significance; the absence of a significant effect is not necessarily indicative of a problem.

I would suggest looking more closely at the native fixel data output by the statistical inference command, rather than performing further processing and then feeling puzzled by that data.

I hope someone can help me understand the meaning of the output generated with FBA …

  • The “absolute effect” is the vector dot product between the GLM beta coefficients and your contrast vector. So its interpretation is predicated on how you have constructed your design matrix, and what your effect of interest is. The “standardised effect” is Cohen’s d, and is generally more interpretable.

  • “fwe_pvalue” is simply (1-p), where p is the familywise error corrected p-value.

  • The percentage effect is explained in the relevant documentation page.

Rob

1 Like