I answer to my own e-mail :
I did the same analysis with the same data but after a fMRIprep preprocessing and it worked !
results are as Han & Park said : between results of the frequentist statistics (voxels and clusters FWE corrections)
but I still don't know how to test the null hypothesis since the spm_bms_test_null is asking for log* images that do not exist...
thank you
Isabelle
Dear "Bayesian" users,
Following the guidelines from Han et al. 2018, I tried to use Bayesian inferences at group level (35 patients) but I am surprised by results...
The one-sample t-test shows very intense and spread activations after FWE corrections in classical inferences.
Then, I estimated this SPM.mat with the "bayesian inference" option, threshold at d= 0.5 (using 0.005 since the default number was 0.01) and logBF = 3 and 0 voxel extent.
But there nomore suprathreshold voxels....
What can I check to understand what happens ?
Isabelle
|