Hi,
On 2 Apr 2009, at 11:05, Mahinda Yogarajah wrote:
> Dear Experts,
>
> I am a little new to design contrast, and matrix design with respect
> to TBSS
> and wonder therefore if I could run some questions past you all.
>
> I am trying to carry out a longitudinal study looking at the effect of
> surgery on white matter in the brain. So I basically have patients
> with DTI
> scans before and after surgery (resection of the same part of the
> brain).
> My questions relating to statistical analysis are (assuming that I
> have 3
> patients for the sake of simplicity):
>
> 1) My first instinct is to use a paired t-test to analyse the data -
> however
> might there be different cross-subject variance in the different
> groups as
> one has had an extensive surgical operation and the other has not ?
> In
> which case should I be using a 2 sample unpaired t-test
>
> 2) In going for the paired t-test model I have created the following
> design
> matrix
> (Input 1 - 3 = preop patients, 4-6 = postop patients in same order)
>
> Group EV1 EV2 EV3 EV4
> Input 1 1 1 1 0 0
> Input 2 1 1 0 1 0
> Input 3 1 1 0 0 1
> Input 4 1 -1 1 0 0
> Input 5 1 -1 0 1 0
> Input 6 1 -1 0 0 1
>
> Would this be correct ? Is it necessary in this case to include
> parameter
> estimates 2 to 4 in order to model each subjects mean effect because
> it
> might confound results ?
>
> My design contrast was (looking for increases or decreases in FA after
> procedure):
>
> Title EV1 EV2 EV3 EV4
> A-B 1 0 0 0
> B-A -1 0 0 0
>
> Woud this be correct ?
This is all correct - you do indeed want to use this paired t-test
design for this scenario to get maximum estimation efficiency.
>
> 3) If I then want to carry out a correlation analysis to see how the
> location of changes in FA before and after surgery correlate with
> changes in
> some psychology score - how would I do this ? In understand the
> changes
> would have to be demeaned but how would I enter this data into the
> design
> matrix/contrast ?
> Just to be more clear with respect to the correlational side of
> things - I
> want to contrast the correlation of FA with psyhcology between the two
> groups (ie compare how it correlates before surgery and how it
> correlates
> after surgery).
This would be a different model - you could for example just use an
unpaired t-test ie two EVs for the two group means, and then demeaned
psychology scores into two other EVs which you can then contrast.
Cheers.
>
> Thanks.
>
> Mahinda
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|