the difference for the FDR between the Xjview and the DPABI

Submitted by daozhe on

Hi experts, 

       I do a statistic analysis for a task-based fmri with SPM.  When seeing the results, I used the software with the Xjview and DPABI Viewer. I found some discrepancy of results between the Xjview and DPABI. when using the Xjview, there is no cluster remained when the FDR was sett at p=0.05 and the enxtended  voxels were 10. However there are two clusters remained when the FDR(DPABI) is set the q value=0.05 and the mask is the group mask. 

  why for this phenomenon?  I want to know that is it  reliable for the FDR correction in DPABI?  I remembered that the Pro Yan said that the FDR in the REST was adopting from the Xjview and Cui. "this report is based on CUI Xu's xjview".    

  I also want to know the what is the FDR method of DPABI ?   If I want  to report the results based on the FDR using the DPABI in the paper, how should I to explain it?  Maybe, the Xjview FDR is voxel-wise FDR p value.

  Another quesetion: why the FDR on DPABI did not to provide the the thresholding T or Z value? I note that the GRF on DPABI have a  thresholding T value.

 

Best Wishes

YAN Chao-Gan

Wed, 04/04/2018 - 02:37

1. DPABI's FDR is not based on Xjview, but this one: https://warwick.ac.uk/fac/sci/statistics/staff/academic-research/nichols/software/fdr/FDR.m. A possibility is DPABI using two-tailed while xjview using one-tailed.

2. Please read more about FDR (e.g., https://warwick.ac.uk/fac/sci/statistics/staff/academic-research/nichols/software/fdr), then you will know why it doesn't need a voxel-wise threshold at first.

Predefined Types