Dear Experts,
After reading the controversy about using froi and localization task
(Friston et al. vs. Saxe et al. Neuroimage, 2006) I would understand if the
cluster/roi homogeneity can be evaluated as follow. I've a 2x2 within
subjects design A (a1, a2) x B (b1, b2) so the design matrix has four
regressor a1b1, a1b2, a2b1, a2b2; I model with canonical hrf.
At first level I run four t-contrasts as ([1 0 0 0]; [0 1 0 0]; [0 0 1 0];
[0 0 0 1]) and in corrispondence to each I get the first eigenvariate, I
type xY.v after I export these weights in SPSS:
ROWS are Subjects: sub1, ..., Subn
COLUMNS are: weight_1_a1b1 ... weight_n_a1b1; ... ; weight_1_a2b2 ...
weight_n_a2b2.
In this way I can evaluate the weight variability for each subject in each
condition (take a row I've mean and standard deviation for a1b1, a1b2, a2b1,
a2b2) and the weight variability across subjects (mean and standard
deviation in coloumn). If the weight variability is low (the weights of the
voxels forming this cluster are the same) I can consider the first
eigenvariate as a cluster mean or in other words this functional roi is
homogenous. Is this correct?
Another question: when I extract the first eigenvariate shall I adjust for
the effects of interest or it better I choose don't adjust?
I've also submitted this mail to marsbar list so I apologize in case of
repetition.
Thank you.
--
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.1.405 / Virus Database: 268.12.9/457 - Release Date: 26/09/2006
|