Dear FSL experts,
I see on the FIX website that at least 10 subjects should be used to create a custom training set. As that seems to be a minimum recommendation, do you have a more general recommendation that we should aim for (e.g., n=30? n=60?)?
Also, is there a recommended sample size when doing a LOO validation/accuracy analysis using "fix -C" (for the validation set).
Thanks in advance for your time and help!
Best regards,
Ryan
|