Hi Chelsea,

I think there is a little confusion here: The distribution of TFCE isn't symmetric around zero, and further, for negative t- or z-stats, it is zero. So, when you see results in the [0, 0.05] interval (of the 1-p image), it doesn't mean these results are opposite, or are in the opposite tail. It just means these p-values aren't significant (i.e., close to 1). This is even more the case for the corrected map, where the distribution of p-values under the null is skewed towards 0 (in the 1-p, that is, towards 1, meaning not significant).

Hope this clarifies.

All the best,

Anderson


On 24 July 2017 at 15:43, Chelsea Stillman <[log in to unmask]> wrote:
Hi Anderson, Thank you for the response.  But do you have any explanation for why the effects from the unthresholded maps seem to be in the opposite direction from TFCE? In other words, when I open both the unthresholded maps (and set z=2.3) and the TFCE maps at 0-.05, they are overlapping.  

Thanks again,
Chelsea


Chelsea Stillman, Ph.D.
NIH/NIMH T32 Postdoctoral Fellow
Department of Psychiatry 
University of Pittsburgh Medical School
Commercial Translation Architect, sciVelo 



On Sat, Jul 22, 2017 at 2:50 PM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Chelsea,

All p-value maps produced by randomise are 1-p, so you'd need to threshold at 0.95.

All the best,

Anderson


On 21 July 2017 at 12:52, Chelsea Stillman <[log in to unmask]> wrote:
Hello, 

We are running randomise on VBM data and want to make sure we are properly viewing the results: When we open the unthresholded zstat image and manually provide a minimum height threshold of z = 2.3, we see a fair number of clusters. Our question arises from the observation that the clusters that appear in the tfce corrected image thresholded at min of p = 0 and max of p = .05 overlap with those in the uncorrected z images. However, the website directions indicate that we should be viewing the tfce images at a 1-p threshold of .95 - 1. We have started viewing the images with a 0-.05 threshold because we noted that the values in the clusters appeared to be p values (not 1-p values as indicated on the website). Can anyone confirm whether it is okay to threshold the tfce output image in this way?

Thanks in advance for any responses!
chelsea

Chelsea Stillman, Ph.D.
NIH/NIMH T32 Postdoctoral Fellow
Department of Psychiatry 
University of Pittsburgh Medical School
Commercial Translation Architect, sciVelo