Eugene,
Thank you for the prompt reply.
My apologies on this, but this isn't making any sense to me. Apparently I've
gained substantial DTI processing skills in the last couple of years, but
have ignored fMRI data processing.
So would you say that what I was doing originally was incorrect or maybe not
optimal?
In regards to your response, what do you mean by the 3 column format? In my
current design I have 3 EVs: a dual task condition (answer equations,
remember letters), letter response (identify letters presented), and a
single task (answer equations).
The contrast that I am interested in is: dual task condition - single task.
The RTs and response accuracy that I am looking at correspond solely to the
dual task condition. So would the RT and response accuracy EVs have the same
setup parameters as the dual task condition?
Then in the contrast window (this is where things are really not making
sense), the RT EV would have what value? The average RT? For the response
accuracy EV I would set it up as 1 to correlate the activity with correct
responses, and then in another contrast as -1 to correlate with incorrect
responses? I don't understand how FSL would know at what time points the
correct and incorrect responses are occurring.
Lastly, the activation maps gained from this would yield what exactly? The
regions correlated with correct responses, the ones with incorrect
responses, and how RT correlates with region activation? Also, I am
incorrect in saying correlate? Is this actually what is explained by these
EVs (regression)?
Again, I apologize for my ignorance on this topic.
-Carlos
On Fri, 22 May 2009 00:26:18 +0100, Eugene Duff <[log in to unmask]> wrote:
>Hi Carlos,
>Sorry, that link was for second level analyses. To define a first-level
>regressor modelling responses that show correllation with with reaction
>time, use the three column format, and have the amplitude of the responses
>vary according to the RT. Then orthogonalise this regressor with respect to
>the regressor that models a steady response to the task. You can use a
>regressor with +1 and -1 response amplitudes to model the effect of response
>errors.
>Eugene
>
|