Hi again,
I think that the problem is this:
I'm doing a memory experiment. It was an event-related design. In the experiment participants could answer 'R' 'K' or 'N' to 3 stimulus types. However, 4 of the options are generally not reported (i.e., 1 response per run), so they end up not really being modeled....
When I remove these 4 stim types from the analysis, and keep only th 5 stim types of interest in, there is not collinearity problems.
So what I was wondering - does removing these stim types from the analysis create havoc for my analysis? Or is it okay to take this type of step?
Thanks,
Erin