I have a question regarding different models (same onsets, different durations) during a fmri-analysis.
Is it possible to change condition durations in the model in order to minimize possible correlations between regressors caused by paradigm limitations?
During fmri-sessions we are showing pictures with emotional content and the participants first should simply look at that picture for some seconds and then they were instructed by a cue to react to that picture in some seconds following. Pictures were shown with three different contents A, B, C and there were two reactions, the participants were asked for: 1,2.
We are interested in effects of looking and also in effects of reacting and therefore we modeled "Looking" and "Reacting" phase in different regressors, with a total of 9 regressors:
3 * "Looking": A, B. C and 6 * "Reacting" A1, A2, B1, B2, C1, C2
Because there are some limitations in the paradigm: the regressors are correlated to each other (Reacting A1 or A2 is always following Looking A)and there is no rest between Reacting onset and the end of the initial Looking phase I would like to use two different models, one is optimized for analyzing effects of initial Looking and another model, which is optimized for analyzing the Reacting phase.
For example in order to minimize the binding of variance by the "Reacting"-regressor for analyzing initial Looking effects, I would minimize the duration of the "Reacting"-regressors in the model. And to optimize the model for "Reacting" phase I set the duration of initial "Looking"-phase to 0.
So we do not testing with different models for different hypotheses of behaviour but instead we would like to minimize interactions between regressors, to compensate paradigm limitations.
Assuming we could find effects only with this different models, but is this strategy feasible? Can I use for each effect of interest a different model?
kind regards and many thanks for any help