Hi Mark and Steve,
thanks for the clarification! Almost there. However I would like to make
sure of one more thing. Since the "Interaction" button at first level
analysis, according to the manual, merely creates an EV where the paradigm
file 0:s and 1:s for EV1 (stimulus A OFF/ON) are multiplied with those for
EV2 (stimulus B), then the outcome is a paradigm file for the AB stimuli (ie
where A and B are presented simultaneously)? Then, the contrast where the
EV3 (for "Interaction") is set as "1" and everything else as "0" is NOT the
factorial ANOVA interaction we are looking for, rather, it is the stimulus
AB vs REST comparison?
And consequently, to calculate the 2x2 factorial ANOVA main effects and
interaction, we need to additionally set up 3 F-tests.
The final setup, then, should look like:
3 contrasts and their interpretations:
OC1 [1 0 0] ; for A vs. REST
OC2 [0 1 0] ; for B vs. REST
OC3 [0 0 1] ; for AB vs. REST (multiplied A and B)
with 3 F-tests and their interpretations:
F F F
1 2 3
OC1 [X - -] ; A main effect
OC2 [- X -] ; B main effect
OC3 [- - X] ; Interaction (finally!)
Err. Right, guys?
One more small thing, if the above is correct, is there any cost in
calculating the negative interaction OC3 [0 0 -1] in the same analysis, and
if not, how should we set the F-tests then (add a fourth F-test, and
continue pushing the buttons towards the lower right corner?). We
experimented with that somewhat and it seemed that we had better run two
totally separate analyses, one with OC3 [0 0 1] and another with OC3 [0 0
-1], and use just 3 F-tests. If we tried to run both in the same analysis
(negative as OC4) then the fourth F-test was not generated.
Thanks again,
Tommi
|