Dear Pei Ling,
> cannot explain why but after fixing the model, I no longer get the bizarre activations
As fMRI is not quantified you have different average signal intensities on different runs, which can be much larger than any change in intensity due to brain activations. That's why one goes with separate constant terms for separate sessions/runs. If you don't do so, then a major part of the variance might be unexplained in your model and findings can be weird, like those you've encountered yourself.
> I'm thinking activation refers to over-additive activation and de-activation refers to under-additive activation. What is additive activation?
Additive just means the activations as seen in task 1 and task 2 separately add up for the combined task, this can refer to positive activations, negative activations/deactivations, or a mixture of positive and negative activations. E.g. if single task 1 results in an estimate of +1, task 2 results in an estimate of -1, and task 3 in 0, then the combined task is (or seems to be) additive. Over-additive means that the estimates for the combined condition exceed the sum of those of the single tasks, under-additive that those of the combined are smaller than the sum. If you were to encounter activations and deactivations / positive and negative estimates it might be useful to make sure what you're refering to though.
For the other question I have to look at the meaning of the contrasts more closely still.
Stay tuned
Helmut
|