This is a reply to a message posted by Deborah on 01-16-04 17:09:

===================
Hi,

In order to have a direct comparison between 2 tasks (A and B), I entered task A as the baseline and looked at the brick comparing task B to the baseline (which in fact was comparing it to task A).

This worked well for individual subjects but now I am not sure how to look at this comparison for the group. This is because for each subject I have just one file (a brick comparing task A to task B), but in order to use 3dANOVA I need at least 2 files per subject. It also doesn't look like one-sample ttest would solve my problem since I don't want to choose the "bval" but to use the "baseline" (in actual fact, task A's) values.

3dANOVA outputs a tstat for each condition in addition to outputing comparisons, which looks like what I need. In other words, I am looking for a way to have a mean and a statistic for each voxel, across all subjects, looking for significant differences is between task A and B (without taking any other "baseline" into consideration).

Thanks for your time.
Deborah
===================

Deborah,

Did you run 3dDeconvolve with task A as baseline in contrast to another task B for individual subjects? If this is the case, it is not appropriate. First, by setting one task as baseline it would be confused with the usual baseline (resting period); Second, the baseline fitting is just some polynomial (usually straight line), which is not a good fit for a regressor.

I suggest that you rerun 3dDeconvolve with both tasks as two separate regressors and leave the baseline for the resting period. You can also make glt contrast test between tasks A and B if you are interested.

In doing so, you would be able to run 3dAVOVA with two subbricks as input, and obtain mean and t statistic as well.

Gang