Dear FSL experts,
We recently used FSL's TBSS tool to examine white matter microstructure in depressed adolescents vs. healthy controls. We performed a whole-brain TBSS and a ROI-based TBSS, while testing for lower as well as higher fractional anisotropy values in patients compared to healthy controls. The results of the analyses were reported in a manuscript, and the manuscript was submitted to a journal. One of the reviewers critiqued the statistical thresholding of the TBSS results (p<0.05, TFCE corrected) saying: Tests of contrasts in the linear model to look for A > B and A < B simply represent a two-tailed test, and thus if the direction of effects are to be tested separately, statistical thresholds corrected for multiple comparisons should be halved (i.e. p<0.025).
I was slightly surprised by this comment, as TFCE correction for multiple comparisons is already a rather stringent method, hence lowering the thresholds from 0.05 to 0.025 could be over-conservative. Furthermore, I rarely read imaging papers in which this procedure is followed!
Given your expertise on statistical modeling and permutation testing, I am interested to hear your opinion on this issue.
Many thanks for considering my inquiry.
Kind regards,
Moji Aghajani
PhD candidate
Leiden University Medical Center
The Netherlands
|