Dear FSL experts,
We conducted a fear-conditioning study in which participants saw four different stimuli. Two of these stimuli were followed by a shock (CS+ stimuli) and two were not followed by the shock (CS- stimuli). In addition, two of these four stimuli (one CS+ and one CS-) were followed by a startle probe. This is an event-related, 2x2 within-subject design and data (n=18) was obtained within a single run. We are interested in both the main effects of CS type, probe/no probe and the interaction CS x Probe. We created a model similar to Figure 1b in this paper:
http://www.mrc-cbu.cam.ac.uk/wp-content/uploads/2015/03/Henson_EN_15_ANOVA.pdf
EV1 EV2 EV3 EV4 EV5 EV6 EV7
CS+probe CS+noprobe CS-probe CS-noprobe subject1 subject2 subject3
subject1 1 0 0 0 1 0 0
subject2 1 0 0 0 0 1 0
subject3 1 0 0 0 0 0 1
subject1 0 1 0 0 1 0 0
subject2 0 1 0 0 0 1 0
subject3 0 1 0 0 0 0 1
subject1 0 0 1 0 1 0 0
subject2 0 0 1 0 0 1 0
subject3 0 0 1 0 0 0 1
subject1 0 0 0 1 1 0 0
subject2 0 0 0 1 0 1 0
subject3 0 0 0 1 0 0 1
Contrasts
Main CS 1 1 -1 -1
Main Probe 1 -1 1 -1
Int CSxProbe 1 -1 -1 1
Wherein EV1 is CS+ probe, EV2 is CS+ no probe, EV3 is CS- probe, and EV4 is CS- no probe.
As input we used 72 COPEs: 4 conditions (CS+ probe, CS+ no probe, CS- probe and CS- no probe) x 18 participants.
However, when we try to run this analysis we get the error that the design is rank deficient.
Then we tried to use the rotated version of this model (figure 7): http://www.fil.ion.ucl.ac.uk/~wpenny/publications/rik_anova.pdf
EV1 EV2 EV3 EV4 EV5 EV6 EV7
Main CS Main Probe CSxProbe Constant subject1 subject2 subject3
subject1 -1 -1 1 1 1 0 0
subject2 -1 -1 1 1 0 1 0
subject3 -1 -1 1 1 0 0 1
subject1 -1 1 -1 1 1 0 0
subject2 -1 1 -1 1 0 1 0
subject3 -1 1 -1 1 0 0 1
subject1 1 -1 -1 1 1 0 0
subject2 1 -1 -1 1 0 1 0
subject3 1 -1 -1 1 0 0 1
subject1 1 1 1 1 1 0 0
subject2 1 1 1 1 0 1 0
subject3 1 1 1 1 0 0 1
Contrasts
Main CS 1 0 0 0
Main Probe 0 1 0 0
Int CSxProbe 0 0 1 0
Wherein EV1 is main effect CS, EV2 is main effect probe, EV3 is interaction CS x Probe and EV4 is a constant.
As input we used 72 COPEs: 4 conditions (CS- no probe, CS- probe, CS+ no probe and CS+ probe) x 18 participants.
And we got the same error.
We would like to know why this model is rank deficient and how we can solve this problem. Or should we use a completely different analysis?
Any help is much appreciated!
Thank you very much in advance, kind regards, Michelle de Haan
|