Hi Glenn,
the newest version of tksurfer has FDR thresholding built in. You should do the smoothing *before* generating the p-values, which mris_glm supports. I think you can grab the latest snapshot with the FDR stuff in it, or if not, Kevin can build you one.
cheers, Bruce
On Wed, 26 Jan 2005, Glenn Lawyer wrote:
Hi,
We are experimenting with statistical methods for analyzing cortical thickness differences in a case/control study. After doing the cortical reconstructions and running mris_glm, we will have the significance (p-value) of the t-ratio of the contrast between case and control. What we want to do is first apply some spatial smoothing to the p-values, and use a false discovery rate controlling method on the resulting smoothed p-values to find an appropriate significance level.
We can run mris_glm, save the output in bfloat format, and load into Matlab using 'fread'; no problems.
What we are lacking is a description of the topology of the vertices, and the mapping from the vector of floats to this topology. I assume a triangulation implies a 6-connect structure. Correct? Assume the vertices are numbered 1:163842 (as in the output of mris_glm --sigt), which vertices are neighbors (what are the six neighbors of vertex 100) ?? Are distances between neighbors (nearly) constant, or do they vary???
Very much appreciative of your reply,
+glenn