 |
 |
Dear Victor, > 1. During trial specification step SPM asks me about > > vector of onsets - so I enter my vector... > variable duration - I answer "YES" > durations (scans) - 34 34 34 34 > > and I do so for each condition type > > However later, when I define my trials as epochs (and choose "box-car", > "convolved with hrf"), I am asked to enter "epoch length (scans)" for > each trial type. Should I enter again the same number (34)? If so, why > does SPM ask me to enter the same information twice? Do I do anything > wrong ? You are right, you entered the same information twice. The reason for that is, that you did NOT have variable durations, because each epoch of the same trial has 34 scans. So, just say 'No' in the question of variable durations. > 2. In our study there are long epochs (34 3sec scans each, belonging to > 4 different active conditions) separated by 6 scans intervals (gaze > fixation + instruction picture viewing). I am not interested in these > 6-scans intervals. Should I include them in the model anyway? What could > be the best way to treat them? If you have no explicit rest condition, which can be modelled implicit, you do not need to specify this condition, because it is something like your baseline, if all 6scans intervals are identical. In the case, that there are happening different things or you have also a not explicit specified rest condition, it will increase the error of the parameter estimation, because you introduce additional 'noise', which is not specified within your model. So it is really important, that you model all known effects within your design, not only the effects of interest, because, every effect, which is not explained by your design matrix can decrease the significance of your main effect by increasing the unexplained variance. So, be careful. > 3. I am going to use scaling to remove global effects. Yet I know that > it is not always good thing to do. My question is: should I use global > scaling anyway if the result of the individual data processing are > supposed to be used later in a multisubject analysis ? In my experience, if you have a single subject, single session analysis, it can be useful, to skip the scaling procedure. If you want to analyse several sessions (same or different subject), I prefer the scaling. If you want to do an random effects analysis, based on the individual statistical results, you should not mix the individual procedures of the single subject level. Good luck, Karsten --------------------------------- Dipl. Phys. Karsten Specht Medizin Center Bonn Spessartstrasse 9 53119 Bonn Germany Phone: ++49-(0)228/90 81-178 Fax: ++49-(0)228/90 81-190 E-Mail: [log in to unmask] WWW: http://www.mcbonn.de/Praxis/praxis15/fmri1.htm