Dear Ana,
I would also go with five sessions within a single-subject model, as you would have to manually adjust/build the drift regressors (invisible in the design matrix) when concatenating. By default they would account for drifts across the two parts of that run, but usually you want them to account only for the drift within each of those.
Note that when going with five sessions, you can still weight them according to your needs, e.g. instead of something like [1/5 -1/5 ... 1/5 -1/5 ... 1/5 -1/5 ... 1/5 -1/5 ... 1/5 -1/5 ...], which treats the five sessions the same way, you could also go with [1/8 -1/8 ... 1/8 -1/8 ... 1/4 -1/4 ... 1/4 -1/4 ... 1/4 -1/4 ... ], in which case the two parts of the splitted runs would only be weighted "half" but the combination of the two be treated just like the other three "proper" runs. You could also go with something like [1/16 -1/16 ... 3/16 -3/16 ... 1/4 -1/4 ... 1/4 -1/4 ... 1/4 -1/4 ... ] in case you want to account for the second part of the splitted run being three times as long as the first part. However, in case one of the parts is much longer I would just drop the shorter part. You should just state this somewhere in the methods section then, "due to technical failure in one subject run no. ... consisted of only ... volumes instead of the otherwise ... volumes (corresponding to ... %), resulting in ... % of the originally planned ... trials".
Best
Helmut
|