I am setting up first-level GLMs for a task where error-related activity is one key contrast of interest. I am wondering if there is a way to handle imbalances in the number of trials between conditions in a contrast, i.e. error vs. correct trials, if a participant only makes errors on 10-20% of trials, for example. Would this negatively impact the estimation of the contrast if they are not weighted in someway by their variance?
It seemed like 3dMEMA might be one way to handle this at the group level, including the beta and t maps for the error vs. baseline and correct vs. baseline as within-subject repeated measures, but is there any other way to account for this at the individual-subject contrast level?
3dMEMA is a reasonable approach to handling potentially different reliability of the effect estimate. However, you would have to obtain the contrast between the two conditions of error and correct, and then use the contrast as input for 3dMEMA.