the reason they look the same is because *-they are the same. And as
you can see, they each show you both sides of the coin.
At the group level, your first contrast 1 -1 (which is the condition by
group interaction) is showing you two things: where group1 activated
more than group2 to the experimental condition, AND where group 2
activated more than group1 to the control condition. This isnt really
interpretable.
There are a few ways to deal with this. One simple thing you could do
is use the mask with other contrasts option. you first need to specify
an additional contrast at level 2: the contrast 1 0 would tell you where
group1 activated to condition 1 relative to condition 2 (regardless of
what group 2 is doing). This could be used as a mask for your 1 -1
interaction and the result would be a map of voxels that were active to
the experimental condition in group 1 and were also more active in
group1 than group 2. this should yield an interpretable finding.
i hope this helps,
SCJ
>>> Luke Stoeckel <[log in to unmask]> 2/12/2007 9:48 PM >>>
I am conducting a two-stage random effects analysis for a block-design
study
comparing 2 subject populations using SPM2. At the first level, I am
specifying a t contrast for each subject, experimental condition minus
control condition by entering "1 -1" (con1). I am also specifying a t
contrast for the control minus the experimental condition by entering
"-1 1"
(con2)...which essentially amounts to a two-tailed t-test...I think.
Then,
at the second level I am taking each of these contrasts and comparing
population 1 to population 2 using a two-sample t-test by entering "1
-1"
and "-1 1." However, my results look the same for the con1, pop1 vs.
pop2
comparison and the con2, pop2 vs. pop1 comparison. What am I missing?
|