[Mne_analysis] Doubt: <(Group study) Spatio-temporal clustering test on Source space Data >

[Diptyajit Das]

*Dear all, *


I am currently analyzing MEG data recorded from 3 different groups of
subjects. Each group of subject has 4 experimental conditions. (2 congruent
and 2 in-congruent conditions.)
For example:
2 congruent conditions are : LLst and RRst
2 in-congruent conditions are : LRst and RLst


I am interested in finding differences in (i.e., LLst Vs LRst) (congruent
Vs in-congruent) conditions within a *specific group of subjects*. p.s: two
conditions are independent from each other but belongs to same group of
subjects. Now I want to use spatio-temporal clustering test as a
statistical to find out the difference between these conditions.


*Possible Solution:*
Do a *'One Sample spatio-temporal clustering test*' where test statistic is *a
paired t -test (Mean difference test)* (paired contrast between LLst Vs
LRst ).


*Drawback: *When the brain signals are recoded for each condition, the
synchronization of the signals during recording may differ for two test
conditions. Therefore, time sample and Voxel/vertices sample may shift in
case of two test conditions. Therefore, paired contrast between LLst Vs
LRst may not be ideal or sufficient enough to draw a conclusion that those
two test conditions differ from each other.


*Doubts I have:*
Is there somehow, I can utilize *'Two samples spatio-temporal clustering*'
to find a difference between ( LLst Vs LRst ) test conditions from the *same
group of subjects*?


Note: Please correct me if I am wrong, the test statistic used in *'Two
samples spatio-temporal clustering'* is a F-test (i.e., Variance test). The
main assumption of a F- statistic is, the variance testing should be done
on the data coming from *two different groups of subjects*. This assumption
is violated when I want to do a test ( LLst Vs LRst ) from a same group of
subjects although my two test conditions are independent from each other.

So is it wise to use *Two samples spatio-temporal clustering* in this case?


I have doubt on my assumption that Two samples spatio-temporal clustering
test may not be appropriate to handle this kind of problem (LLst Vs LRst:
two conditions from same group of subjects). I feel *Two samples
spatio-temporal clustering* are specific test that allows to find answers
when I want to test two conditions from *two different groups of subjects* [for
example: LLst (group 1) Vs LLst (group2)].


As the test is delicate, your help or suggestion will be really helpful.
Looking forward to a positive reply.


Many thanks and kind regards!

Diptyajit Das
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20170825/a2da7e73/attachment.html