Hi everyone,
We set-up the following experiment to test for a session-effect in our study (our participants completed 2 identical sessions):
EV1 EV2 EV3 EV4 EV5 EV6
Input 1 1 1 0 0 0 4
Input 2 1 0 1 0 0 2
Input 2 1 0 0 1 0 -2
Input 4 1 0 0 0 1 -4
Input 5 -1 1 0 0 0 4
Input 6 -1 0 1 0 0 2
Input 7 -1 0 0 1 0 -2
Input 8 -1 0 0 0 1 -4
EV1 = Session 1 or session 2.
EV2 = Participant 1
EV3 = Participant 2
EV 4 = Participant 3
EV 5 = Participant 4
EV 6= Covariate
However, when we try to run the model we get an error ('overflow error'). We now found out that the reason for this error is the covariate. Our covariate (a physical trait) did not differ between sessions. Thus, for each participant, we put two identical values for the covariate in our model (1 for session 1 and 1 for session 2). This seems to cause the problem (we also get the following error message for our model in feat 'at least one EV is (close to) a linear combination of the others.'). Is it possible to test for the session-effect of our experiment without removing the covariate from our model?
Best,
Esther
