Dear SPM experts,
I'm conducting a study in which I'm applying a similar first-level fMRI analysis using different datasets. I would like to report the average of the parameter estimate across subjects, from different datasets, removing the influence from the dataset. This way, I'm running a second-level analysis in which I include the subject-specific contrast images and study-specific covariates (in "Factorial design specification"). Covariates A,B,C,... are vectors comprised of 0's and 1's, in which 1 indicates that the subject comes from dataset A,B,C,... respectively.
However, I've realized that the estimated beta/contrast range differ quite a lot across datasets. In my case, most datasets led to beta values that have a magnitude of 10e-2, but one dataset led to betas that can reach 10e0 - 10e1. Even with the covariates in the second-level analysis, I found out that the final result is quite driven by this one dataset with the high beta range...
What can I do to properly remove the influence of the dataset in my second-level analysis?
I can provide further details if necessary. I very much appreciate in advance any ideas you can provide.
Best wishes,
Gustavo
|