Dear Olivier,

Dear Olivier,

Yes, use&report the stats in the results explorer window (e.g. p-FDR based on cluster size or cluster mass of each individual cluster) when reporting results from voxel-level analyses. Those stats are valid and properly corrected for multiple comparisons across the entire brain or anaysis mask (cluster level uncorrected p-values represent the likelihood of observing a cluster as large or larger than the one obtained in your results under the null hypothesis, and p-FDR/p-FWE corrected p-values apply an additional multiple comparison correction to take into account multiple clusters across the entire brain).

The results in the REX results table are meant to be used only in the context of post-hoc analyses, and they are only valid as confirmatory tests when the cluster or ROI definitions are obtained independently from the data being analyzed, which is definitely not the case when using REX to display the effects within a cluster found significant in the voxel-level analyses of the same data. In general the REX results take the average functional connectivity values across all voxels within a cluster/ROI and perform a second-level GLM analysis on that data (that is why the resulting stats are only valid when the ROI definition is independent from the contrast evaluated in the GLM analysis), so they can be used for example to analyze associations between functional connectivity and symptom severity within a cluster obtained from a patients vs. controls between-group comparisons.

Hope this helps

Alfonso

Originally posted by Olivier Brown:

Dear CONN experts,

 

Simple question: The size p-FDR values from results explorer differ from the p-FDR values in the plot effects REX results table. I imagine this is because the size p-FDR is for the cluster after cluster and voxel thresholds.

Should I avoid reporting the F value from the REX results table from plot effects then considering the corresponding p-FDR is different from that of the cluster size p-FDR from results explorer? How do I interpret the p-FDR value in the results explorer vs those in the REX results under plot effects?

Thanks!