Hello Gwenaelle,
I do doubt age may have an effect on the FA value and I would like to
treat it as confound and get rid its effect before I resume to the group
comparison (control vs patient).
Can my design matrix answer the question (Which part of the brain show
differences in FA value when comparing control to patient). I would also
like to know if I need to demean the data by adding the option -D, since I
have already split the age factor to *demeaned age for control* and
*demeaned age for patient*, and padded the rest of the cells with zeros.
It would be even nicer if someone could explain to me the reason for
adding the -D command, and in what kind of situation one will need to do
that.
Thank you.
Yi-Shin
> Hi Yi-Shin,
>
> Tom will correct me if I'm wrong but it also seems that in your model,
> you're taking into account the effect of an interaction between diagnosis
> (control vs patient) and age. In other words, you're considering that the
> effect of age might be different on the FA between the two groups (which
> might be the case!).
>
> If you simply want to look at a linear effect of age no matter what is the
> group, then you need to put the age in one sole EV and remove the mean age
> of both controls + patients.
>
> Cheers,
> Gwenaelle
>
>
> --- En date de : Jeu 11.9.08, Yi-Shin Sheu <[log in to unmask]> a
> écrit :
>
>> De: Yi-Shin Sheu <[log in to unmask]>
>> Objet: [FSL] TBSS 2 sample Unpaired T-test & demean option
>> � [log in to unmask]
>> Date: Jeudi 11 Septembre 2008, 16h03
>> Dear FSL experts,
>>
>> I searched the FSL archive but it is still kind of
>> confusing to me about the
>> demean option.
>>
>> I have two groups (control vs patient) and I would like to
>> see the difference
>> between their FA value in TBSS analysis while controlling
>> for the nuisance
>> covarietes (age).
>>
>> Could anyone tell me if my design matrix is correct and if
>> I need to add -D
>> option in randomise.
>>
>> deisgn matrix:
>>
>> EV1 EV2 EV3 EV4
>> 1 0 -2 0
>> 1 0 -2 0
>> 1 0 4 0
>> 0 1 0 -1
>> 0 1 0 2
>> 0 1 0 -1
>>
>> EV3 is the demeaned age for control group which I padded
>> with zeros. EV4 is
>> hte demeaned age for patient group, which I also padded
>> with zeros.
>>
>> design contrast:
>>
>> 1 -1 (Control >Patient)
>> -1 1 (patient > control)
>>
>> Then I run: randomise -i all_FA_skeleton -o tbss -m
>> mean_FA_skeleton_mask -
>> d design.mat -t design.con -n 5000 -c 3 -D
>>
>> My question is, do I have to run randomise with the -D
>> option in this case?
>>
>> I appreciate any input you might have. Thank you so much.
>>
>> Yi-Shin
>
>
>
>
Sheu, Yi-Shin
Research Assistant
McLean Hospital
Developmental Biopsychiatry Research Program
Tel: 617-855-2942
Fax: 617-855-3712
email: [log in to unmask]
|