 |
 |
Dear SPMers, I am doing a simple block design task fMRI experiment. I found that after I run the 1st level model estimation. There are lots of NaN values in the regions I am interested (visual cortex). Those NaN voxels have 0 values in spmT map. I tried to use threshold mask out certain regions in the poor susceptibility where is far from my interested area. Then I found less NaN values in the Beta maps. I am confused that I thought beta values is calculated from the least square multiple regression analysis, inv((X)*Y which is independently voxel by voxel. And, inv(X) is pseudo inverse, so each voxel should has a real value no matter what. Does the NaN values in beta has anything to with FDR or FWE random field p value correction? But I thought that is calculated later after defining the contrast… Thanks, Susan Postdoc at JHU