regressors for error and feedback events.

[jongmin-529]

Dear Dr.Kay,

Thank you very much for developing awesome tool!
I wanted to ask about the design matrix.

Let's say there is a hypothetical flanker task where there are 6 conditions (2 X 3; high vs. low loads X target color red vs. blue vs. grey). The feedback (O or X) is given on each trial. The interval between stimulus and feedback is jittered. The correct trial is coded as one of 6 conditions and the incorrect trials are coded as Error.

In this design, I'm thinking of two design matrix although my interest is the 6 conditions:

1. columns of the design matrix: 6 conditions + error + feedback
2. columns of the design matrix: 6 conditions

I think error and feedback can explain some variance in data but the second design matrix seems okay in GLMsingle because omitting error and feedback regressors can make blank time.

But, I'm not sure which one is optimal..

Thank you in advance!

[kendrickkay]

If I understand correctly, then I would say you should use #1. The idea is that you want to account for all possible response components in the data. If you omit the error and feedback (in #2), then it will (as you say) be effectively assumed to be "blank time". But in the actual data, there may be BOLD responses, so that could confuse the blank/baseline signal estimation.