I just finished analyzing my data using the fixel-based analysis framework. Unfortunately, I didn’t get statistically significant results as I expected (using the FWE corrected p-value).
I suspect one of my control subjects might be an outlier that’s throwing off the statistical analysis. I was just wondering if there’s a way I can test whether that’s true not without having to exclude the subject and starting the FBA pipeline from scratch.
Thanks for the help and for the very well-documented pipeline!
Ultimately the data files used in the new fixel directory format are just one-dimensional lists of numbers, which happen to be stored using an image file format. If you use the mrdump command, you can obtain these values in text format, which you can then analyse using any software of your liking.
While this will give you a value per fixel per subject, you will not have access to the spatial & orientation information of the fixel representation.
The mrcalc command operates on fixel data files just as it does image files of higher dimensionality. So you can derive any kind of measure you want on a per-fixel basis, and visualise the results within mrview's fixel tool.
For advice over and above that, you would need to give a more detailed description of the way in which you think this subject may be an “outlier”. E.g. May their FD be decreased throughout the entire brain? Could there be a registration problem such that their FD values are zero for a large portion of the brain? There’s a large number of possibilities here in terms of what could be derived from the data, whereas an explanation of what convinced you that this particular subject may be an outlier may give a clearer picture as to how this could potentially be visualised / quantified.