 |
 |
Dear Stan & Darren, I'm afraid I don't have any definitive solution to contribute, but I think there may be two separate problems here: 1. Generating corrected p values for some contrast in a small volume defined by a 'localiser' in a separate scanning session within a single subject 2. Generating some group statistic for activation in such a 'localiser' defined area, where the 'localised' area may be spatially variable across subjects. The first problem is addressed by the SPM99b Small Volume Correction button as Darren suggests, using an image defined by the localiser. But I believe that it is the second problem that is at the heart of Nancy K style analyses, and that would seem more difficult. My understanding of the argument is that because the position of the 'ffa' is highly variable across subjects, then it may be most powerful to define the area within individual subjects, extract signal and then average in a spatial-location invariant way. In other words, any averaging is across functionally homologous (rather than anatomically homologous) areas. This seems orthogonal to the SPM philosophy, which is voxel based! Moreover it seems that any advantage this method might have is highly dependent on how variable the spatial location of some functionally defined area might be across subjects. It would be presumably possible to effect a simple analysis of this type by extracting signal from individual subjects, then effecting a sort of one-voxel random-fx analysis on the raw timecourses within MATLAB. But this doesn't seem either elegant or intrinsically more powerful than a fixed effects conventional SPM analysis. Any comments?! best wishes, Geraint --------------------------------------------- Dr. Geraint Rees Wellcome Advanced Fellow, California Institute of Technology, Pasadena, CA 91125 voice (626) 396-2880 fax (626) 796-8876 web http://www.klab.caltech.edu/~geraint --------------------------------------------- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%