Group analysis with permutation tests


I am planning on running a group analysis to determine whether there is greater BOLD activation in group A vs group B for different trial types. I plan to fit a GLM to get beta weights for each subject for the following contrasts: 1. trial 1 vs rest, 2. trial 2 vs rest, 3. trial 1 vs trial 2. Then I plan to test the group differences with 3dttest++ to see whether the main hypothesis that there are greater beta values for contrast 2 in group A compared to group B. My questions are as follows:

  1. Is any thresholding applied to the first level analysis (besides potentially brain masking) or are all Beta values from an individual subject input into the second level analysis?
  2. Is this thresholding procedure the same for most software packages or unique to afni?
  3. Finally are the group labels permuted with the -clustsim option or something else (I saw previous posts mention the contrast weights were flipped but I wasn’t sure what this means or if this was the current method)?

Thanks for the help!