Dear all,
We have modeled a simple subtraction contrast [Task Rest] with both SPM99
and SPM 2, and have noted differences in the results provided by the two
versions of SPM depending on whether the Rest condition is explicitly
modeled (Analysis 1) or not (Analysis 2).
SPM99
The contrast [1 -1] in Analysis 1 gives practically the same results as the
contrast [1 0] in Analysis 2.
In a specific voxel:
Analysis 1: beta task beta rest = 5.9614 (T=24.96, df=551)
Analysis 2: betatask = 6.036 (T=25.37, df=554)
SPM2
The contrast [1 0] in Analysis 2 does not give similar results to the
contrast [1 -1] in Analysis 1.
In the same voxel:
Analysis 1: beta task beta rest = 9.6652 (T=21.44, df=1208)
Analysis 2: betatask = 6.6919 (T=16.07, df=1212)
Why the difference?
In addition, if both task and rest conditions have been explicitly modeled
how is it possible for SPM2 to return a reasonable-looking result for a
contrast of the type [1 0]? If everything has been explicitly modeled (it
is a blocked design), then surely there is nothing left with which to
compare the Task condition to?
What is it being compared to?
thank you in advance
some SPMers from Marseille
|