Dear Donald,
My model included ...
- one psychological regressor (with an FOE file that contains two conditions, contrasted in the third column by using the weights +1 and -1), convolved with a double-gamma HRF,
- one non-convolved physiological regressor (the time course of an ROI in the right amygdala - the ROI stems from a previous GLM group-level result and was transformed to each subjects functional space with FLIRT),
- the psychophysiological interaction (created with the interaction-function and psych-regressor with "centre" and phys-regressor with "mean"),
- two regressors for a second task (i.e., a few control trials in two conditions) that were not directly of interest for ppi,
- a regressor for missed trials
- standard and extended motion regressors (but no [other] preprocessing steps were done [except registration] as I used each subject's filtered_func_data from the previous GLM as input on the first level of the new ppi model)
The odd finding for the physiological regressor was visible in the thresh_zstat when using cluster-based correction, but of course also in the zstat-map. My ROI appeared to be correlated with pretty much the whole brain.
Best, Stefan
|