Hi,
It AFNI Howto 2 and 5, multiple runs are concatenated while in Howto 1, they are averaged.
Yet another option when we have multiple runs is to do 3dDeconvolve run by run.
Is there a general rule as to when each of the above methods should be used?
My opinion is that when you concatenate or take average, you get more reliable coefficients from level I analysis, at the cost of losing (n-1) observations if you have n runs.
In other words, the issue is whether to have less but cleaner observations or to have more but noisier observations.
Am I right?
Thanks,
George