Hi SPM experts,
I'm trying to do for the first time a correlation analysis on fMRI data with some clinical parameters, testing separately the control (10 subjects) and pathological (20 subjects) groups.
The problem is that, while with the pathological group I found a clear positive or negative correlation within the ROIs tested (e.g.: p<0.1 r=0.45 for the positive and p<0.001 r=0.89 for the negative), in the control group I had controversial results: typically I observed a strong correlation in both sense (e.g.: p<0.005 r=0.93 for the positive and p<0.005 r=0.94 for the negative).
I explained this results as a missing correlation in the control group between the activations and the clinical parameters (assuming that the strong, double direction, fake correlation is created by the correction for the confounding parameters mixed with data with a great error): is it correct?
My other clue is that maybe I had a small number of subjects in the control group compared with the number of interesting and confounding covariates (my DOF in this part of analysis is 4) and so I cannot infer nothing on my control group as long as I'll recruit more healthy subject: is that so?
I hope to be enough clear in my explanation and I'll waiting for an answer with trepidation,
Chiara
|