Dear Nikos, as Satoru already mentioned his NS-toolbox reports the results in RESELs while my VBM5-toolbox reports the corrected cluster-size k. I have decided to report the corrected the cluster-size k to minimize the changes for defining and reporting the cluster-statistic in SPM. Thus, the k-values will differ, but the p-values for both approaches will be very similar. The small difference in the p-values might be caused by the slightly different computation of the local smoothness (Satoru is using fmristat from Keith Worsley). The local distortion due to non-stationarity might also affect the AAL results (mainly the percentage of voxels belonging to a VOI). However, this will only affect the results if local smoothness deviates inside a cluster. This might be true, but the deviation of local smoothness will be much larger between the different clusters and the deviation inside a cluster might be negligible. Best, Christian -- ____________________________________________________________________________ Christian Gaser, Ph.D. Assistant Professor of Computational Neuroscience Department of Psychiatry Friedrich-Schiller-University of Jena Jahnstrasse 3, D-07743 Jena, Germany Tel: ++49-3641-934752 Fax: ++49-3641-934755 e-mail: [log in to unmask] http://dbm.neuro.uni-jena.de On Thu, 7 Feb 2008 14:38:05 -0500, Satoru Hayasaka <[log in to unmask]> wrote: > > >Dear Nikos, > > > >sorry for bothering you with the following question: I would like to use >non-stationary RFT for my VBM analysis but I got confused when I compare >the results obtained by Dr. Hayasaka's NS toolbox with the results >obtained by the normal spm_results_ui or Christian Gaser's >non-stationarity correction implemented in the VBM5 toolbox. > >In a two-sample t test I found one significant cluster by using a >primary threshold of p <0.01 and a FWE-corrected extent threshold of >0.05: > >1) Normal stationary RFT shows me a corrected p of 0.021 (k=6066) for >this cluster. > >2) The NS toolbox calculates a corrected p value of 0.010 (k=6066, >resels=5.316). It seems as if k refers to the stationary RFT results...? > > > >In this case, k refers to the number of voxels. In the NS toolbox the >corrected p-values are calculated based on RESELs rather than the number >of voxels, but k is displayed for reporting purpose. > > > >3) Christian Gaser's toolbox produces a corrected p value of 0.009 >(k=7019). Why is the p value so much lower compared to the NS >toolbox...? > > > >The difference between the p-values is small (p=0.010 vs. p=0.009) but I >see that the number of voxels (k) is different as well. If the primary >threshold is the same in both analyses, I suspect there may be some >differences between the statistic images from the NS toolbox and the >VBM5 toolbox. Has the data been processed the same or differently in >both analyses (template creation, normalization, segmentation, >smoothing, etc)? > > > >Do you if there is any way to input the adjusted cluster sizes to the >cluster labeling routine of AAL? > >The gin_cluster_plabels script of the AAL toolbox is similar to >spm_list.m. I highlighted the line where I would make corrections but I >am not sure...I don't know how to pass the non-uniformity adjusted XYZ >to gin_det_plabels. > > > >I am not familiar with AAL, but the NS toolbox does not alter XYZ. In >other words, it doesn't create the non-uniformity adjusted XYZ. > > > >Hope this helps, > >-Satoru > > > >Satoru Hayasaka PhD ---------- >Assistant Professor, Public Health Sciences & Radiology >Wake Forest University School of Medicine >(ph) +1-336-716-8504 / (fax) +1-336-716-0798 >(email) shayasak _at_ wfubmc _dot_ edu > >