Hi,
You are right that it passes the (arbitrary) z-threshold (which doesn't mean anything statistically) but fails to pass the cluster (corrected) p-threshold, where the latter is the only statistically meaningful result.
If you want to report non-significant findings, such as trends, then you can either report voxel-wise zstats or lower the cluster p-threshold until you see the area appear in the cluster table, and then report its p-value. I don't think there is any standard reporting procedure for non-significant results really.
All the best,
Mark
On 15 Feb 2013, at 19:44, Dana Wagshal <[log in to unmask]>
wrote:
> Hi Mark,
>
> My problem is that when I ran my analysis with a z threshold of 2.3 and a cluster P threshold of 0.05, I noticed that a specific area that I'm interested in didn't pass this threshold. But when I looked into the cope3.feat/stats/zstat1.nii.gz image in fslview the peak voxel in this region had a z value of 2.651. So if I'm interpreting this correctly, it passed the z threshold value but not the cluster P threshold value, correct?
>
> In my paper I want to say that this region was a trend. How would I go about reporting a p value for this region if it's not in the cluster table that contains the corrected p-values? And would I report the uncorrected P value or a FWE corrected P value?
|