Dear Andreas,
as far as I know, the threshold of 0.05 is completely fine here, also
for two-tailed situations. This leaves 0.025 at each tail of the
distribution (someone with more insight into the exact SPM statistics
implementation might correct me if I'm wrong). You get the complete
statistics for both contrasts, just the signs are inverted and SPM does
not display negative significant results even if they are there. You can
check that easily using image viewers like MRIcro with your t-maps.
Best regards,
Manfred
Am 08.01.2018 um 14:30 schrieb Andreas Jansen:
> Dear colleagues,
>
> I am interested in activation differences between two conditions
> (contrasts active > baseline, baseline > active). For that, I used an
> one-sample t-test and a threshold of p=0.001 at the voxel level,
> p<0.05 at the cluster level. Now someone has correctly pointed out
> that I use a typical one-tailed threshold for a two-tailed test. What
> threshold do I have to use? Do I just have to correct the cluster
> level threshold, i.e., p=0.001 at the voxel level, p<0.025 at the
> cluster level for each of the two contrasts?
>
> Best wishes Andreas
|