Dear SPMers,
using a repeated measures ANOVA (partitioned errors) I found an interaction
group x time (2groups of subjects / 2measuring time points: t1 & t2 ). The
task was an n-back task with 0-back (control) and 2-back (active) in a
blocked design.
Main effects were modeled on the first level; the main effect of TIME was
modeled like follows:
0-back t1 / 2-back t1 / 0-back t2 / 2-back t2
-1 -1 1 1
The interaction effect was found by entering the first level contrasts for TIME
in a 2-sample t-test using t-contrast 1 -1 (result: one cluster, max. t-value at
x/y/z).
To further explore which specific changes over time might have caused the
interaction effect would it be appropriate to:
1. Model the first level contrasts for:
0-back at t1 (1 0 0 0)
2-back at t1 (0 1 0 0)
0-back at t2 (0 0 1 0)
2-back at t2 (0 0 0 1)
2. Plot the values at the max. t-value of the interaction effect (x/y/z)
for each of the 4 newly created contrasts (see 1) & read out values for each
participant
3. Enter these values in e.g. SPSS and analyze if the values for 0-back
t1 vs. 0-back t2 or the values for 2-back t1 vs. 2-back t2 in the two groups
changed over time and therefore caused the interaction effect?
Thanks (btw. also for the very helpful answers to my numerous previous
questions) &
kind regards
Ben
Phd student / University of Cologne
|