 |
 |
Hi Michael > First level Analysis: > I have the following questions: > (1) Is there task related activity? I am setting this up as shown in the attached diagram > with a '1' for the task and all zeros for the rest of the parameters. Yes, that's right. > (2) Does the addition of the custom regressor (R) change my activation? That's a more > tricky question and I am not quite sure about it. Should I set that one up as a '1 1 > followed by zeros' - contrast or a '0 1 followed by zeros' - contrast? Ignoring motion regressors, you have 2 columns per session: the first is the main task, and the second the additional regressor R. I would think of it like this: A contrast of 1 0 (as you noted above) gives you mean activity related to your task across all levels of your custom regressor. A contrast of 0 1 gives you activity related to your regressor that deviates from the mean: if you perform this contrast and don't see anything, it would suggest that this custom regressor is not resulting in any extra neural activity. (This is assuming you mean-center your custom regressor, which is the default and makes sense in this case.) A contrast of 1 1 is probably not very meaningful, because it essentially takes the average of the main task and your custom regressor. If you get a significant result you don't know which of the two it comes from; and if you don't get one, it is hard to know why. > Alternatively, should I run 2 models, one including the custom regressor 'R' and one > without it and I am getting two .mat files in my 1st level analysis, how do I then set up > my model comparison in SPM? I understand that comparing nested models would be an F- > test, but how do I enter two .mat files in the contrast manager? I don't see how this would give you anything additional; if you include your regressor in your original model, the simple 0 1 contrast should tell you if it is contributing. > (3) If I figured out the proper way at the subject (1st) level, how do I use this information > at the second level. Presumably, this depends on the approach at the first level. If your questions on the group level are the same as those for the single subject level (as you explain above), and you have one group of subjects, you can just enter each of those contrasts in a 1 sample t-test. So, one t-test to look for a group effect of task, and a second one to look for a group effect of your custom regressor. Good luck! Hope this helps. Jonathan