I have a question about how to set up parametric estimation or contrast.
The design I have on hand had 3 levels of a subjective ratings (S1/S2/S3)
of the stimuli, each with a non-overlapping range (e.g. S1: 100-199, S2:
200-299, S3: 300-399). Currently I am only setting up the EVs for each
level of the ratings (one EV for each of S1, S2, and S3) and the contrasts
between any two of them (e.g. S1 vs. S2).
Now I want to see whether brain activation is parametrically affected by
the level of ratings (e.g. linear). There are two ways I can think of to
set up the analysis.
Model 1:
Set up 1 EV for the ratings of each stimulus event (time point) and
estimate the brain activation for this EV.
Problem: It is too restrictive and not flexible in terms of contrast setup.
Brain activation closely following the pattern of the ratings could be very
limited. Also interpretation of the negative activation is questionable
since fixation period will take the value "0" for the rating.
Model 2:
Set up 3 EVs for the levels of ratings (binary 0/1 for off/on) and set up
contrasts using the mean ratings for each level as weights.
Problem: How? Say, to test the hypothesis that brain activation is
modulated linearly with the mean ratings (e.g. 162, 248, 356, for S1, S2,
S3 respectively). Any F test needed?
EV1(S1) EV2(S2) EV3(S3)
C1 ??? ??? ???
C2 ??? ??? ???
C3 ??? ??? ???
Thanks very much for any suggestion.
|