Hi Jeanna,
I'm not really sure of what you want to do ...
> I want to define my contrasts using time derivatives, but I want to use a t-test instead
> of an f-test. The only documentation I've seen on this topic (regarding use of time
> derivatives) seems to only explain the entering the contrasts in an f-test.
> Is it possible to use a t-test, and how would these contrasts be entered? I have a
> one session, within-subject, event-related design with five conditions. If I
> want to make the contrast N1 > N2 with time derivatives, could I enter 1 1 0 -1
> -1 0 ... for a t-test? Or is an f-test my only option, and then how would I
> specify direction, while also using time derivatives?
Is it possible to use t tests with time derivatives ? I'm not sure ...
--- I will give you an explanation .. (it's also a good way for me to
test if I'm right :-) but please, dear spm experts, correct me if I'm
wrong :-( )
The positive estimate of the temporal derivative corresponds to an
earlier response than the classical hrf whereas the negative estimate
corresponds to a later response. At the individual level, the estimation
of the time derivate (t contrast +1) would capture both types of event
... so you cannot distinguish positive and negative loads using a
t-test a the second level (if it's correspond to your goal ?).
To compare conditions 1 & 2 it would be better to compute the difference
between C1 & C2 at the first level for the hrf, the 1st der. and the 2nd
der. (t contrasts 1 -1 for each function). Next at the second level,
select ANOVA without cste, 3gp (hrf, 1st, 2nd) and look at your effects:
F (100, 010, 001) for the diff between C1 and C2, and T(100) for the
hrf, T(010) for the 1st der and T(001) for the 2nd der .. select a
cluster and plot the F contrast ...
Hope it helps you
Best,
cyril
|