Dear Chun-Yo,
if your four event types are cells of a 2x2 factorial design it is
straightforward. Have a single vector defining the onsets of any trial
('all trial types') with this subject to parametric modulations by a)
the first factor b) the second factor c) the interaction between factors
and d) the RTs. Contrasts vectors for the parametric modulator terms
will indicate the main effects of each factor or the interaction ( [0 1
0 0 0 ] = first factor expressed by a) above , eg attention >
no-attention) after having removed the influences of common RT
effects on estimations of a) to c). The opposite contrast [0 -1 0 0 0 ]
would identify voxels for which the first factor had opposite effects,
eg no-attention > attention.
James
> But let's say if I have 4 event types (conditions): type 1,2,3,4.
> How should I model them then?
>
> And after modeling this 4 even types, how do I make the contrasts to see
> "which two" are different?
>
> Can someone kindly shed me some lights or point me to relevant references or
> examples?
>
|