Dear SPM experter,
I met a problem about orthogonized regressors during parametric analysis.
Our task is to compare a parametric modulation effect on Task A with that of Task B. We obtained Task A and B data of the same subject on two different days. We have two parametric regressors, just say, Regressor a and Regressor b. We hypothized that the modulation effect of Regressor a could be found in Task A but not in Task B, but the modulation effect of Regressor b could be found in Task B but not in Task A. Due to the fact that Regressor a and b are not effectively orthogonalized, we performed our analysis as follows:
We first considered the modulation effect of Regressor a. In the first level, we designed a GLM in which there is only one parametric regressor, i.e., Regressor a, for Task A and Task B separately. So we got the contrast image (e.g., [0…1…]) of the modulation effect during Task A and that during Task B for each subject. And then a paired test was performed to compare the differences of contrast images between the two tasks in the group level. In the same way, we compared the modulation effect of Regressor b on the two tasks.
Our question is “ Is this a valid approach?”.
Regards,
Catherine
|