Basically, I am doing the following steps in separate scripts to conduct a beta-series context dependent correlation analysis:

1) 3dDecon with IM option to get individual regressors for each condition

2) Extract betas from conditions of interest to create a beta file using 3dbucket
-This creates a separate file for betas of interest for each trial (e.g., ConditionA_betas.nii)

3) Extract and de-mean betas from a seed region using 3dmaskave and 1d_tool.py
-So for example, I created an FFA seed that contains the trial-by-trial betas for Condition A (e.g., Seed_ConditionaA_betas.1D)

4) Use 3drefit -TR 2 on "ConditionA_betas.nii" from Step 2, to make it seem as though it is a 3d+time file (although it is not technically).

5) 3dfim + -input ConditionA_betas.nii -polort 2 -ideal_file Seed_ConditionaA_betas.1D -out Correlation -bucket ConditionA_BetaSeries.nii
-Supposed to create a correlation map of the Seed region with the rest of the brain corresponding to ConditionA

6) I would then Z-score this and take it to a group level analysis (e.g., 3dtest++).

Looking at the individual subject outputs from the above step seem somewhat reasonable (i.e., strong correlation surrounding voxels from seed region). Although the group-level is not making sense: within "ConditionA", the seed robustly correlates positively with just about every voxel in the brain...

-Danny