Hi Gang,
The experiment I am analyzing was an event-related 2 x 3 x 2 x 2 design.
The factors are Spatial Shifts (Switch, NoSwitch) x Distractor Salience (No,
Low, High) x Visual Field (LVF, RVF) x Task Difficulty (Easy, Hard) x Subject.
I used GroupAna to run a 5 factor ANOVA, with Subject as a random factor.
What I found was that when I compared the t maps generated during
the ANOVA to t maps I created with 3dttest, my t and p values were
different. To compare the two programs, I only looked at one map,
the first order contrast of Switch vs. NoSwitch. I will send you my
GroupAna script, my 3dttest script, and the maps generated by both
programs. If you set the threshold at p<.005, you should be able to
see the difference quite easily.
I suppose the way that I have set up the contrasts must not
be equivalent, but I am not sure where I have gone astray.
Incidentally, this ANOVA was run before the last change in the
program, but it is my understanding that the older version was too
lenient and so overestimated the p values. My problem is the exact
opposite. The p values from GroupAna are much lower than from 3dttest.
Thanks,
Stacia