Dear Megha,
> I have a task with three conditions A, B, C. The trials in each of the
> conditions are varied on a parameter which is a feature of the trial
> stimulus. I used two models to assess activation differences across
> three categories. The first model looked at each condition, A, B, C
> and i compared across conditions (A-B, A-C, B-C...etc). In the next
> model, I additionally used a parametric modulator for each of the
> conditions A, Apm, B, Bpm, C, Cpm and used the second column as the
> regressor for each condition. However, now when i compare across
> conditions I do not see any differences. My question is, would it be
> correct to infer that the difference I see across the conditions A, B,
> and C is due to the parameter in question and not any other feature of
> the stimulus?
You say that in the case in which you added the parametric modulators,
you used the second column as the regressor for each condition. Does
this mean that if your conditions were
A Apm B Bpm C Cpm
you did a contrast looking at, for example, for A-B:
[0 1 0 -1 0 0]
I.e., Apm - Bpm? Or did I misunderstand what you meant about using
the second column? If this is the case, it's not surprising that you
got different results, as you are testing something different. It
could well be that conditions A and B differ (contrast [1 0 -1 0 0
0]), but that the effect of the parametric modulator does not differ
by condition (contrast [0 1 0 -1 0 0]).
If you want to say that the difference across condition cannot be
attributed to the parametric modulator, you would need a slightly
different model. As it stands now the parametric modulator tells you
how that factor affects activity within each condition, but not about
how it may contribute across the whole experiment. To look at this, I
think you would want a single column that codes for every event
(assuming you have the same number of events in each A,B,C condition),
a single parametric modulator, and then additional columns that code
for condition A, B, C. So:
[any_event pm A B C]
In this case, because the parametric modulator goes across all
trials/conditions, I think you would be safer in concluding that an
A>B difference exists apart from anything that could be explained by
the parametric modulator.
Hope this helps!
Best regards,
Jonathan
--
Dr. Jonathan Peelle
Department of Neurology
University of Pennsylvania
3 West Gates
3400 Spruce Street
Philadelphia, PA 19104
USA
http://jonathanpeelle.net/
|