help > p-value priority for connectome ring
Aug 3, 2020  11:08 AM | Paul Summers
p-value priority for connectome ring
Hi Alfonso,

     regarding the p-values in second level analysis.

  on the basic 2nd level page, if I choose my combination of covariates, condition and a seed ROI, for roi-to-roi analysis with set a pvalue, the table in the lower center shows the destination ROIs with p-values below the criteria bright and shades out the rest. If I choose p-FDRcorrected <0.05 the this is the over-riding threshold applied to the table.

   In results explorer, using alternate multivariate parametric stats (apologies if I've remembered the name incorrectly) I get a connectome ring and table. Expanding the table (twice) to see both the source and destination ROI names, the list almost matches the list mentioned above, but not quite... One needs to open the settings (last item in the list at top), to access the p-value criteria controls. Here there are two options for making p-value choices.
an upper list:
p-uncorrected,
p-FDR corrected

and a lower list with longer names like: 
p-uncorrected
MVSA p-FDR corrected ...

setting p-FDR corrected (i.e. same name as in the basic second level page) in the upper list to the same threshold as in the basice page, and using the default (I think p-uncorrected) in the to the lower list, doesn't always result in the same regions being listed / included in the connectome ring as on the basic 2nd level page.

similarly, setting p-uncorrected in the upper list to the an arbitrary value above the maximum uncorrected p-value in the basic 2nd level table, and setting the MVSA p-FDR corrected option to <0.05 in the to the lower list, also doesn't result in the same regions being listed / included in the connectome ring as on the basic 2nd level page.

To achieve the same selection of connections, one needs set the p-uncorrected criteria in the upper list to a value between the last of the highlighted and the first of the greyed out values for the uncorrected p-value in the basic 2nd level table, AND set the MVSA p-FDR corrected option to 0.05 (or whatever the criteria was on the basic 2nd level page). (I didn't check, maybe setting p-FDR corrected in the upper and MVSA p-FDR corrected in the lower list to the same value would work without needing to figure out the intermediate value from the preview table).

This suggests that the whereas the basic 2nd level roi2roi page applies only the threshold for the specified p-value type, the connectome ring is prepared with an OR operation that results in the display connections that satisfy thresholds for either the p-value types thresholds.

Accepting that there may be a need for the two p-value lists in Results Explorer, the above consequence is nonetheless, somewhat counterintuitive. By default I think people expect one criteria to be applied.

Is it possible to prioritize just one of the p-value lists so that the OR operation on the p-values is avoided?

related subquestions:
Could the default view match whatever is do in preparing the basic preview in the basic 2nd level page?
(e.g. passing the threshold set, and making active any Seed ROI(s) chosen in the basic 2nd level page and using whichever stats seed is used for the preview?)

Does the pFDR corrected in the basic 2nd level page correspond to pFDR corrected in the upper list or MVSA pFCR corrected in the lower list of the Results Explorer?

Hopefully that made sense.

Paul S