I am doing the 1st level design specification and estimation and I have a
lot of confusion regarding this and further steps. During each visit, the
subject performs five tasks, all of them different. After preprocessing, I
ran the 1st level analysis (specify and estimate) on all the tasks together.
I’ve attached the “Specify 1st-Level” design .mat file. 

First off, is doing the analysis on each subject for each visit, thereby
grouping all the tasks performed by the subject during that visit, into one
analysis process the correct way to do things? Or should I analyze the each
individual’s tasks separately? Or should I group all similar tasks from all
individuals under the same condition into one analysis? Or does it not
matter? At this point, I am inclined to believe it doesn’t matter, but just
wanted to verify with the knowledgeable. Along the same line, am I correct
in assuming that all the corresponding hdr/img “beta_*” files for different
subjects/visits will have the same image dimensions, orientation and voxel
size, due to the preprocessing step? In other words, will the corresponding
beta files all be in register? I assume because of the normalize step. 

After I ran the estimate of the SPM.mat file using Classical modeling, there
were a total of 40 different “beta_*.hdr/img” files, plus a “con_004”, a
“mask”, a “ResMS” and “RPV” hdr/img files generated. I am confused as to
what each one of the beta images stand for (I checked for just one Task and
that process generates 8 files, and that totals 40 beta files for 5 tasks).
I looked at the manual, and it doesn’t give me a lot of details. I am also
currently looking at Friston et al.’s “Classical and Bayesian Inference in
Neuroimaging: Theory” paper of 2002. Will that help? 

Finally, when I do the 2nd level specification, what do I do? How do I group
the individual scan statistics together to get the group activation map and
the statistics? I also want to compare between conditions and groups. My
goal is to do the functional connectivity eventually. I need to extract the
statistics and create a group-activation map first. I have pre-defined ROI’s
that I want to extract the time-series from, and then I am planning on using
the ImCalc module to extract the TS of the activated voxels within the ROI.
Thanks.