hi anderson
my intended rationale for the second design is to evaluate the longitudinal differences at the subject level (and thus absorb the session factor)....and later use the resulting output in a 3rd level analysis where I can evaluate group differences (of the longitudinal effect) or the interaction between different covariates and group factor....
my real main question of interest is in fact:
do longitudinal differences (connectivity differences between session 2 and session 1) differ between groups? I'm evaluating the effect of a children training program
such 2nd level analysis should be almost equivalent to subtract the copes of the 1st level analysis of each session, isn't it ? (I actually tried to subtract the cope images from 1st level analysis, but without success, so i moved to the complex approach)
the other approaches instead confuse me a lot. I read dozens of posts on 2groups x 2 sessions model,
mining the list i found these two models you posted.
https://dl.dropboxusercontent.com/u/2785709/outbox/mailinglist/design_ekaterina.ods ( https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;3cee56eb.1510)
specifically evaluating between and within differences
you wrote:
" [....] One is for the within-subject effects (e.g., the effect of time, such as FU>BL, named "design1"), and the other is for between-subject effects (such as the effect of group, as PAT>CON, named "design2") ".
they look like main effects, while the interaction is generally more interesting.
on the other hand...in the contrasts labels, for the first design you wrote: g1>g2 t1>t2, what does it really mean? voxel where t1-t2 is higher in g1 wrt g2 ???
the other design, since manage the 2 sessions by subtracting them,
1 0 1 0 0
1 0 -1 0 0
1 0 0 1 0
1 0 0 -1 0
seems to provide the same result, voxel where the session differences is higher in one group wrt the other...
so C2 design1 == C1 design2 ???
another source of confusion is ...how these 2 models relate with the ANOVA: 2-groups, 2-levels per subject (2-way Mixed Effect ANOVA) described in the GLM Wiki. is the latter a preliminary analysis that allows testing the other two?
thanks in advance, sorry for this long post....hoping it may help other people in the future...
Alberto
|