 |
 |
Hello Allison, There is no way we could pass the masked contrast up: at higher levels, complete data is needed for all subjects, but the masking would make many voxels disappear for various subjects. Consider an extreme case in which a subject activates an area that has nothing to do with the overall areas of the other subjects. If masking was passed up, the intersection would be empty, and nothing could be found. If you need contrast masking at higher levels, consider reassembling the model and levels in a different way, so that masking can be done at the highest. All the best, Anderson On 25 June 2015 at 20:57, Allison Jack <[log in to unmask]> wrote: > Hi all, > > I understand that the topic of contrast masking has been covered a good > bit already on the listserv, but I am still struggling with my particular > analytic problem. > > I have a group of subjects who engaged in an experimental task with two > conditions, and we are comparing the two conditions to each other, C1 > C2 > and C2 > C1. > > They were each scanned twice, once under a treatment condition (TX) and > once under a placebo condition (PLC). During each scan they were exposed to > both experimental stimulus conditions, C1 & C2. > > I am interested in the differences in brain activity to the experimental > contrasts for TX > PLC. Obviously, I want to know each subject's individual > TX > PLC difference and then get the overall mean of this difference for > the whole group. > > I originally thought that the way to do this would be: > > Level 1: Estimate C1, C2, C1 > C2, and C2 > C1 for each scan > Level 2: Carry up C1 > C2 and C2 > C1 copes and combine the TX and PLC > scans for an individual (fixed effects). Contrasts at this level are TX, > PLC, and then TX > PLC and PLC > TX, where what I'm interested in is TX > > PLC, and I contrast mask it with where the TX contrast is above threshold. > My presumption here is that what I would be getting would then be, for a > particular individual, where brain activity to C1 > C2 (or C2 > C1) is > above threshold during the treatment condition, and also significantly > greater than under PLC. > Level 3: Mixed effects analysis to get mean of the TX > PLC experimental > contrasts across all individuals. > > But, I read something on here that the contrast masking doesn't get > carried up into higher level analyses. So...? Help? > > Thanks, > Allison >