Dear FSL experts,
I'm performing simple correlations between VBM images and behavioural scores with a canonical randomise command line.
Results look fine to me except unpermuted raw t-stats. In fact the values of this statistic is ridiculously high (even 50 or more in absolute value) for a large amount of voxels.
I know that this output should not be used for inference as it is not based on the null distribution created from my sample.
However, as i find such high values, I have realised that I probably do not understand what raw t-tstat images means in correlation analysis and how they are calculated (and how to decide a threshold to interpret them)
I hope this is not a trivial question for the forum.
Thanks,
Marco
|