Hello-
We have run the python script on our individual subjects and would like to combine data from some conditions/regressors (i.e. correct +incorrect responses= all responses) to run group level analysis of both correct response and all responses. In lieu of re-running each individual analysis again with a combine regressor, were planning to use 3dcalc but realize that the number of events per condition varies significantly for some participants. The concern is that for conditions with a very small number of events (i.e. 1 or 2 events for incorrect responses compared to the more common 20 events for correct responses), the signal to noise ratio will be reduced, and thus disproportionately skew the combined regressor. Is it appropriate/possible to use 3dcalc and weight the datasets according to the number of events per condition? If so, would should the denominator be (see ?? below)?
3dcalc \
-a correct.tlrc
-b incorrect.tlrc
-prefix all_responses.tlrc
-expr " ((a*20)+ (b*2))/??
Thank you,
Elizabeth