Dear Meera,
> I have a somewhat (basic) question regarding 2nd level analyses. I have been conducting my analyses such that at first level I'll specify my contrasts (i.e., task vs baseline), and then at 2nd level I run a t test entering "1" when defining my contrast. I heard once that it is better to run only main effects at 1st level, and specify your actual contrasts at 2nd level (though I can't remember the rationale for this).
>
> I'm just wondering if anyone has any insight into this (e.g., is the activation more robust in one case versus another), and why it would make a difference when you specify your contrast.
If you specify the contrasts appropriately, there shouldn't be any
difference. See, for example:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=spm;4d2421a6.1109
to see this in regard to a paired-samples t-test vs. a one-sample
t-test on difference images.
The significance of the result depends on the effect size (how big a
difference is), the error, and the degrees of freedom. If you specify
a 2nd-level contrast incorrectly—for example, not accounting for
subjects in the design matrix—you can artificially (inappropriately)
inflate your degrees of freedom, and thus get a "more significant"
result.
There has been quite a bit of discussion on the list about 2nd-level
ANOVAs and appropriate error partitioning, if you're interested. But
certainly for common cases of comparing 2 conditions using a t-test,
it's fairly straightforward, and doesn't matter one way or the other.
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/
|