 |
 |
Hi Anda, > Because I have only 1 condition/run do you think that is it a good idea to > concatenate all the runs in a single run after preprocessing is done? No, I wouldn't. It just introduces errors. I'd suggest analysing each of them separately at the first level and use hgher levels to combine them. > I would like to include the baseline as a condition ==> 2EV's , one for the > condition, one for the rest period. > > contrast (cond1-rest): 1 0 In fact, if you include the rest condition explicitly, this isn't a meaningful contrast. You would need to compute [1 -1] to compare your activation condition to rest. Having said that, it's probably best *not* to include rest *explicitly* in the model because you will make it rank deficient. Because you include all of the scans in your time series, not just the ones from the activation condition, you include rest in the model in either case. When it's implicit (ie not modeled by a single EV), you avoid the problem of rank deficiency and you can still have the same contrast: activation > rest but now its simply [1]. Rest > activation would be [-1]. > Regarding the behavioral measures (RT and Accuracy), how can I incorporate > them in my analysis? How can be formulated a hypothese with behavioral > measures? > Should I define an EV for each behavioral measures at the first level ? It depends on what question you want to ask. How do you expect that your behavioural measures affect the BOLD response? One thing which I commonly do is even for blocked designs such as this, I analyse it in an event-related fashion and exclude incorrect responses in the EV of interest. The incorrect trials can either be modeled separately or ignored. If there aren't many, ignoring them may be sufficient otherwise you can just model them and ignore the EV in your contrasts. The (potential) problem with modeling the errors is that if any subject made no errors, then they wouldn't have this EV and this would screw up higher level analyses (but see Steve's email suggestion for dealing with this in the last day or so). Hope this helps, Joe