On Mon, 6 Feb 2017 07:29:14 +0000, Haobo Zhang <[log in to unmask]> wrote:
>Dear Prof. Gaser,
>Your toolbox TFCE is very convenient to use in SPM. However, I am not clear on how to report the results after TFCE correction.
>1) After TFCE estimation, your toolbox also asks for the choice to correct multiple comparisons. In that case, should we report the result as p<0.001 uncorrected, TFCE-estimated, or p<0.05, FDR corrected, TFCE-estimated?
>However, I saw that some papers (including DTI, T1 and fMRI studies) only reported as p<0.01 (TFCE-corrected), but did not indicate multiple comparison way. Is this an acceptable way to report for publication?
Here, the same applies as in all other studies that are based on parametric statistics. The corrections for multiple comparisons using FWE or FDR are the preferred approaches. Uncorrected thresholds should be only reported in rare cases (you should have a clear and convincing anatomical hypothesis) and should be treated with care.
>2) after using your TFCE toolbox to show the results, in the whole brain p-value report (SPM-Graphics), there is a TFCE column, the values like 394.50, etc for each peak voxel. What is the meaning of that column? Shall we report that in the Tables?
These are the TFCE value that combine voxel height as well as cluster size. Please check the initial TFCE paper of Smith & Nichols about this topic:
>3) also in that report, no p-value is provided for each supra-threshold cluster. Should we understand that as each supra-threshold cluster's p < the value that we set for multiple comparison correction (e.g uncorrected p<0.001, FDR<0.05, etc)?
There are no p-values for cluster size or voxel height available because it's a combined single measure.
>School of Psychology and Sociology