I have a within-subject variable that I would like to explore - task
difficulty for an encoding event; three levels represent 'easy', 'medium'
and 'hard' difficulty levels. I have explored this variable using t-
contrasts at the single subject level and then selected each subject's con
image for the second level and defined a 2nd level t:contrast [1]. This
has only allowed me to look at two levels at a time eg medium > easy.
I would like to use an F contrast to look at all three levels to
investigate the trend in activation data for particular voxels over
increases in difficulty level. I have defined an F:contrast 1 1 1 (for
easy, medium, hard) at the single subject level & then selected various
voxels and obtained the beta values for the three regressors. This has
provided useful information about linear increases in regional activation
but only at a single subject level.
How can I do this in a 2nd level analysis? Do I specify the 2nd level
using ess or SPM_F images? I tried using ess images and the SPM glass
brain looked wrong.
Is there any way of obtaining beta values (as above) averaged over the
group?
Does anyone have any other suggestions for the F contrast which may test
for linear trend or quadratic trend?
Many thanks for any suggestions
Nicky
|