Hello Matthias
>Dear SPM-Experts,
>
>I would appreciate any help concerning the following question:
>In an event-realted design with 3 different types of stimuli and a parallel
>measurement of the reaction times, I aquired data from 14 subjects. I am now
>interested to see in which areas the correlation between the BOLD-response
>and the reaction time is modulated by the item type. I suppose this is the
>interaction of the factor item type and the parametric regressor reaction
>time.
>
An ANOVA should do the job
>However, I am unsure about how to specify the design matrix on the
>first level and how to combine the data of all subjects at the second
>level. In a first step, I entered the three item types [A B C] and the
>corresponding reaction times as a parametric regressor to my design matrix
>on the first level, yielding a matrix with 7 colums [A Art B Brt C Crt
>Const]. I am now unsure on how to proceed to test the above mentioned
>interaction.
>
>
You can play around with the regressors ABC and ArtBrtCrt separately,
e.g. contrasts [1 0 0 0 0 0] for A etc .. or if you have one item type
as control, for ex C then compute [A-C], [B-C] and [Art-Crt] and [BrtCrt] ..
Then use contrasts of this first level in an ANOVA for ABC or ArtBrtCrt
.. Another interesting test to do (at least I found that quite
interesting when looking at some data) is to compare the result you get
between the effect of a parametric regressor (one-sample t-test) and the
correlation between a regressor and the corresponding RT (simple
regression).
Best
- Cyril
|