 |
 |
Dear SPM experts, I need some advice on statistical analyses regarding an fMRI dataset that is currently in revision. I have a 2x2 experimental design. Let's call them A and B. The conditions I have are then A1B1, A1B2, A2B1, A2B2. In addition, there is a parametric modulator (efffect of delay on factors A and B). I am not interested in the effect of the parametric modulator, but BOLD activity linked with responding to the parametric modulator (detecting vs. no t detecting a delay). I therefore divided the 4 factors into 8 regressors which also involved whether a delay was detected or not (A1B1_detected, A1B1_undetected, A1B2_detected, A1B2_undetected, A2B1_detected, A2B1_undetected, A2B2_detected, A2B2_undetected). I carried two separate analyses (first and second level): in the first, I looked at the effects concerning factors A and B. In the second, I added the detection factor and looked at the effect of detection as well. The reason why I conducted two separate analyses was because in the second analysis, I had to exclude data from 2 participants due to insufficient number of detected/undetected trials within a session. Here, the reviewers were concerned that delay detection is confounded by the physical delay, and that two analyses were performed separately. For the first problem, it was suggested that I conduct partial regression analysis in which I model physical delay and delay detection, and look whether delay detection explains any variance in BOLD once physical delay is accounted for. Regarding missing data, they suggested me to conduct a linear mixed model analysis. I have no experience in LMEs and other neuroimaging analysis packages, and have limited time. As far as I understood, LME cannot be performed in SPM. My questions are: - How can I perform partial regression using SPM? - Is it possible to adopt an approach similar to LMEs using SPM? If not, which program would be easiest to learn and implement given limited time? - Is there any way I can add all these factors in one analysis (e.g. parametric modulation as covariate)? Thank you very much in advance. Best Ezgi