> From [log in to unmask] Wed Apr 12 19:50:08 2000
> Original-Received: from
> 204.68.24.73 by nw173 for [205.143.94.2] via
> web-mailer(M3.4.0.33) on Wed Apr 12 18:48:45 GMT 2000
> Date: 12 Apr 00 11:48:45 PDT
Dear Jessica,
> We're piloting a motor learning paradigm, in which subjects are scanned
> while performing a motor task A and a motor task B, then are trained
> for a week on A alone, then are scanned while doing both tasks again.
> What we are looking for ideally is 1) strength of signal--those voxels
> which show a different activation for A than B in the second scan, but
> not in the first scan; and 2) extent of activation--a larger or smaller
> area for A than B only in the second scan, but not in the first scan.
> We have three subjects, and will only run more if we think the results
> look promising, since this is a pilot study.
>
> Given the preliminary nature of the study, is it a fair assessment to
> compare cluster sizes for each subject individually for A before and
> after training, and for B the same way? Or is it better to put all
> three subjects in a single design matrix and look for conjunctions? Or
> is some other approach even better?
The conventional approach to this problem would be to create a SPM of
the effect of interest, namely the condition x session interaction.
This obtains from putting all your data into a single, session-
separable model and testing for (A1 - B1) > (A2 - B2) or (A1 -
B1) < (A2 - B2) interactions with the appropriate contrast. This
approach controls for non-specific time effects and should be
practice-specific.
The question about differential areal activation is implicitly answered
in the interaction SPM (i.e. if the area contracts the penumbra will
show a negative interaction and if it expands it will show a positive
one).
A more sensitive analysis obtains if you use a conjunction of the
interaction and main effects: i.e. test for a conjunction of the two
hypotheses: This motor responsive area (Hypothesis 1 = main effect of
A vs B) shows learning-dependent adaptation (Hypothesis 2 =
interaction).
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|