 |
 |
Dear SPM experts, I'm dealing with some peculiarity with my results and wondering if anyone can help me with it. So, I have two sets of results and their activation pattern is more or less similar to each other when I look at it using p<0.001 (uncorrected), k=15. However, when I compare those two with FDR 0.05, majority of voxels in one set still survive, whearas nothing shows up in the other set. The intact one under FDR has threshold t-value 5.42, whereas the other one has Inf t-value. I guess this makes the huge difference, but I'm not really sure why Inf t-value was calculated. In fact, there are many voxels whose t-value is greater than 5.42 in the data set that did not survive under FDR and I feel like I'm losing a real activation due to the way how FDR is performed. Then, can I use uncorrected p value with a reasonable cluster size that matches the threshold t-value of FDR? Any advice is greatly appreciated. Thanks, YSL