Hi,
On 29 Jul 2009, at 22:28, Matthew Hoptman <[log in to unmask]> wrote:
> Hi,
> I've been playing with TFCE results. They appear to offer a more
> lenient
> correction than clusterwise corrections. But my question is how
> should I view
> the tfce_corrp_tstat?.nii.gz results. Are any visible clusters
> significant by tfce
> familywise correction, or do I have to threshold the images prior to
> interpreting
> the results?
Just as with other p-values output by randomise (I.e., 1-p), set the
lower value of the intensity display range in fslview to .95 to get p<.
05 (in thus case corrected). Anything that survives that is valid....
>
>
> In one of my analyses, the p-values range, for example, autoscale
> from 0
> to .2961. Does that mean that none of the clusters are significant
> in this
> analysis, or is something else going on?
>
If you mean that (1-p) has maxvalue .3 then yes this means nothing is
significant.....
Cheers, Steve
|