Hi Gabriel,

Mark and I had a brief exchange on your question.

You can do voxelwise SIENA. Just cd into the siena dir and run
siena_flow2std as indicated.
Your usage sounds right.
>I am not sure if the above is OK.
Sounds ok.
>Now my question is how I have to design my matrix for cross-subject
>statistics?
What is your question? If your null hypothesis is that there is no
difference in local edge flow between time points across groups, than this
is a two sample unpaired t-test.
You can use randomise or palm for statistical inference.

>my two groups are the same subjects
I don¹t quite get that.
14 subjects with two time points will give you 14 siena flow images in
standard space, so your design can¹t exceed 14 subjects.

Also, please have a read of
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1606&L=fsl&P=R122322&1=fs
l&9=A&J=on&d=No+Match%3BMatch%3BMatches&z=4

Voxelwise SIENA preceded VBM and asks a different question than VBM or
cortical thickness analyses.

If you look at all studies that have been published using the method,
including our own, it is probably good to look where ventricular volume
has changed across subjects.
But it is unlikely to be very sensitive in the detection of outer brain
atrophy changes. 
And it does, per se, not reflect gray matter change.

So depending on what you want to assess, voxelwise SIENA may or may not
give you what you are after.

Cheers,
Andreas


Am 10.07.17, 13:32 schrieb "FSL - FMRIB's Software Library on behalf of
Mark Jenkinson" unter <[log in to unmask] on behalf of
[log in to unmask]>:

>Dear Gabriel,
>
>Siena is a tool for calculating global atrophy and so the result of
>running it for a pair of images from one subject is a single number - the
>percentage brain volume change (PBVC).  Hence it is not necessary to
>create 4D images and use a GLM for the statistics.  You simply take the
>single numbers from each subject and run them through a standard
>statistical package (e.g. SPSS, R, etc).
>
>If you want to look at voxelwise changes then you need to use VBM instead.
>
>All the best,
>	Mark
>
>
>
>> On 6 Jul 2017, at 18:51, FSLGRG <[log in to unmask]> wrote:
>> 
>> Hi FSLuser
>> 
>> I am starting with the SIENA analysis.
>> 
>> I have a group of subjects scanned at two different points in the time
>>(n = 14, n = 14), (my two groups are the same subjects)
>> 
>> Following the steps, I ran
>> 
>> Siena <subj_time1> <subj_time2>
>> 
>> And then, in order to carry out voxelwise SIENA statistics
>> 
>> I changed me to the output directory and I ran
>> 
>> Siena_flow2std A B
>> 
>> In this case, are A and B the images created by siena command (A.nni.gz
>>and B.nii.gz), is not it?
>> after this command, I just got 14 images resulted.
>> 
>> Now, I merged the output images A_to_B_flow_to_std in only one 4D file,
>>therefore I had only one image with 14-time points.
>> 
>> I am not sure if the above is OK.
>> 
>> Now my question is how I have to design my matrix for cross-subject
>>statistics?, 
>> I was following the FEAT style as FSL says, but I am not sure if my
>>design.mat is OK. Because I have one 4D image with 14 subjects and the
>>FEAT indicate that I have to make a design with 14 (first point) and
>>other 14 (second point). I use a matrix like this but obviusly taht did
>>run. 
>> 
>> Any suggestion about to make the matrix for my analysis?
>> 
>> Gabriel