Hi Steve,
Thanks for replying. Sorry but i didnt understand you completely..Perhaps i
probably asked the wrong question. i'm trying to compare 2 stimulus
paradigms.In both paradigms, there are 21 subjects with 4 runs, each. the 4
runs are split into 2 runs followed by 30 mins break and then 2 runs.What
i'm trying to achieve is to see if stimulus paradigm 1 has a better/worse
Inter-Run Variability than paradigm2. I figured One of the methods that this
can be achieved with is by using the method that you proposed in your
HBM2005 paper 'Variability in fMRI'. But i'm trying to figure out if the
approach i'm using is valid as well..i.e i found %BOLD change in different
rois for each subject for each run. for eg
roi : Amygdala Paradigm1
%BOLD change run1 run2 run3 run4
Subj1 0.5 0.2 0.4 0.1
Subj2 | | | |
|
|
Subj21 0.6 0.2 0.5 0.3
and similar stuff for Paradigm2.
So i'm treating Run1 as group1 ,run2 as group2 and so on..
then when i run one factor ANOVA using SPSS on the 4 groups to find if the
group means for any of the runs are significantly different from any other,
one of the outputs i get are the 'within group' and 'between group
variances'. In my previous mail what i probably should have asked if its
valid to use the 'between group variances' as a parameter to assess which
paradigm has less inter-run variability.
i.e Paradigm1 Paradigm2
Roi1 x1 y1
ROi2 x2 y2
'
'
Roi n xn yn
x,y being between group variances
Also,another unrelated issue had popped up: when i run featquery on 1st
level analysis using for eg Amygdala roi to get %BOLD change..what value
should i consider ..max ,min or mean?
in certain rois there is min -0.35 and max +0.45..and the mean is almost 0.
in such a case wht should i be reporting.
and some %BOLD changes seem to be around an impossible 3-4% which arent
even activated voxels.
Any ideas as to why this is? all i can think of is overlapping functional
boundaries within an imperfectly registered roi, for the former and
correlated motion for the latter.
What i did so far is to make a mask of the group level zstat and use that
to mask the rois that i used in 1st level featquery so that i'm evaluating
only those voxels that are activated in group analysis. However it doesnt
solve the problem completely ? Any suggestions?
Sorry for the long mail! figured 1 long mail is better than 10 small ones :)
cheers,
vish
|