 |
 |
Hi Curren > The design: There are 4 runs. In each run, there are 3 blocks where subjects > use button press to respond to question 1, question 2, or passive viewing > when viewing photos (the events). Within each block, there are 4 second > events (viewing 4 different catagories of photos) alternating with 10 second > fixation. Just a quick point of clarification...you say you are setting up a "mixed block design experiment". Although it's true that you are grouping events by the question subjects respond to, you still may wish to analyze the experiment in a purely event-related manner. I.e. if you think the type of question affects the visual processing, it may only affect the first 1 sec of each presentation, not the whole 4 seconds, and not the fixation periods. > For model specification, I have a few questions: > > 1) Should the events be conditions with duration=0 or sort of 'mini' blocks with > duration=4sec.? This depends on how long you think the neural activity of interest is. For example, if you are interested in early visual processing, and you think your subjects are looking at each photo for all 4 seconds, then I would think you should put in the full duration of 4 seconds. However, it might be you are interested in some other process that you think takes less than 4 seconds (e.g., visual object identification), in which case, it makes more sense to put in a shorter time. > 2) Can the response the participant gave (its a 1-4 'not at all' to 'very much' > rating) be modeled as a regressor? Yes—you can model this as a separate regressor or as a parametric modulator of another condition (i.e. picture category). If you think the subject response has some effect above and beyond the picture category—and particularly if this might be different for different conditions—the latter seems appropriate. > For the contrasts: > > 1) Assuming it is modeled as described above and I want to see a certain > contrast for a certain block (say picture1 GT picture2 in block 3), how would > that be specified with weights or would I have needed to have each block as > its own 'Subject/Session' rather than having each run be a 'Subject/Session'? This depends on how you set up your model. For instance, if I understand you correctly, for one scanning session, you have 12 cells in your design: 3 types of questions that are being asked x 4 picture categories One way to do this would be have each of those 12 conditions modeled as a separate condition, giving you 12 columns in your design matrix for this session (ignoring any other regressors or parametric modulators). You have 4 sessions (runs), giving you 4 x 12 = 48 columns in your design matrix, which will be for all scans acquired for this subject. You can then work out any comparison across columns you like fairly easily, so long as you've kept track of which column is which. Hope this helps! Good luck. Jonathan