Hello everyone,
I would like to ask two follow-up questions to Brad Buchsbaum's questions
on sessions:
(1) What is the recommended way of doing a single subject analysis if
there are only a few sessions (say 2 or 3)? I have found that for a 2-run
experiment with very robust activations in each run, a 2-level analysis
fails to yield any significant activations. This makes sense since there
are only 2 observations at the 2nd level. So should we resort to
concatenation of runs in a case like this? If so, presumably it would be a
good idea to include covariates of no interest to account for differences
in baselines/trends across the different sessions?
(2) In the example 3-level analysis on the web, the 3rd level analysis is
based on the 2nd level COPEs. Does this mean that the between-session
variance (modelled as the same for all subjects) is ignored at the 3rd
level?
Thanks very much for any help on these questions...
best regards,
Stephen Wilson
|