Hello,
We're trying to enter face ratings into our model as covariates in spm5 and
from reviewing previous posts on this topic, I want to make sure that I
understand how to model this.
In brief, our study looks at impression formation during encoding of
face-behavior pairs into memory. We want to remove "noise" from the face
itself (we now have ratings of attractiveness, trustworthiness, &
distinctiveness) in order to get a better model of the processes manipulated
with our stimuli/task (in a somewhat unwieldy 2x2x2 design).
I gather that we should:
- collapse across conditions to model all of the onsets for faces as 1 column
- then use parametric modulations to enter the ratings for each of the rated
face dimensions (attract, trust, distinct)
- then create "categorical" columns to model each of our diff conditions of
interest. To account for our 8 conditions, we will have 7 columns, with a
"1" denoting the trials falling into a given condition.
Something like the following (if formatting holds...):
onset Rate1 Rate2 Rate3 Cond1 Cond2 Cond3 (up to Cond 7)
1 3.2 2.3 1.5 1 0 0
3 5.1 2.7 5.7 0 1 0
4 4.3 3.9 4.4 0 0 1
To look at contrasts of interest and treat ratings as of "no interest", we
would enter 1 & -1 in the last 7 columns.
Is this all we need to know to get started with this?? Is there a simpler
way to use the "covariate" option in spm5 (unlike motion parameters, we
don't have values for baseline periods)?
Bonus question:
How would we look at the ratings as a covariate of interest? Presumably we
could do main effects with a "1" in any of the ratings columns, but how
would we look at interactions with task?
Thanks for any guidance!!
Angela Gutchess
|