Hi,
I would like to implement a second-level conjunction analysis to test the hypothesis that any contrast of my set of contrasts is significantly different from 0. In this case the contrasts in the first-level analysis are pretty simple, and I am simply estimating if a subset of K beta coefficients in the GLM are different from zero (i.e. C1 = [1 0 0 ....], C2 = [0 1 0 ...] , CK = [0 0 ... 0 1 0 .... 0]).
Most of the discussion to do the second level conjunction analysis I'd like to do refer to this previous post, however it is still unclear to me how to proceed in my case: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind04&L=SPM&P=R187333
How to perform the 2nd-level analysis in SPM? I know that variable 1 will be the group of C1 contrast files for all subjects, variable 2 the group of C2 contrast files for all subjects, etc. but I am not sure which is the best way to implement it via the GUI. I am not interested at this point at other approaches as I would need subsequently the SPM.mat file associated with this analysis (more precisely to use it in NeuroPower for power analysis calculations.
I would be interested in both the "global null hypothesis" to start with (any contrast is different from 0) but I would like to then consider the more conservative "conjunction null hypothesis" to test if all contrasts are different from 0.
Any help on creating the SPM.mat file including both the first-level analysis (simply estimating individual GLM beta coefficients) and the second-level analysis (conjunction analysis to test if any is significant different from zero) would be greatly appreciated.
Michael
|