Hi
These are raw ICA map values divided by the std. error of the residual noise. Depending on the amount of signal per map there ight be a global histogram shift, that's why in melodic mm the fitting includes fitting the mean and variance of the background noise Gaussian (rather than assume it's N(0,1) which is what you're doing when thresholding). Check out the histogram of these maps in e.g. fslview to see what's going on
hth
Christian
On 25 May 2012, at 14:22, Rebecca Steketee wrote:
> Dear FSL users,
>
> Can somebody please explain to us what values are held in the melodic_ic file? We thought the file contained (unthresholded) z-values, however when we threshold the melodic_ic file and look at the mean values, we see some of them exceed 4.5. These values seem rather high to us (when you consider the z-distribution).
>
> Any input is appreciated!
>
> Thanks in advance,
>
> Rebecca
|