Hi,
We would recommend using two EVs, each built with the 3-column format.
The first EV models the mean over all tasks, and therefore contains the value "1" in the third column for all entries.
The second EV models the changes about the mean, and you use the same 3-column file except that you change the third column so that each value is equal to the demeaned value of interest (i.e. the value minus the mean of all of these values).
This then allows the GLM to model the mean over the tasks separately from the relationship between the values and BOLD signal (in the same way that a correlation calculation removes the mean and only looks at the changes about the mean).
All the best,
Mark
On 5 Mar 2014, at 10:53, Riccardo Navarra <[log in to unmask]> wrote:
> Hi... I was looking for an answer to my question... In my case, I have a value of performance for every task periods (8 tasks, 8 values).
> What is the best way to include in my feat run?
> 1) add a regressor built with n-th value repeated for all TR of n-th task period and 0 for rest period.
> 2) change the weight of task period with normalized performance.
> 3) or...
> Thanks,
>
> Riccardo
|