Hi Anderson
Thanks a lot, that was really helpful.
I have assumed compound symmetry within blocks, as the three sessions
are sampled regularly.
I believe that I should perform the contrast C18 (interaction) as an
F-test instead of a t-test, in order to see the difference between the
longitudinal effects of the two groups.
I have tried to do this be defining a design_C1617.con with two
t-contrast (Linear increase of G1 and Linear Increase of G2), and
furthermore a designC18.fts with one contrast to see the difference
between the two t-contrast (interaction). Is this correctly defined?
I now run the following:
./palm -i Sorted_ICA_4d.nii -d design.mat -t design_C1617.con -f
designC18.fts -eb design.grp ?o Output -n 5000 -within -whole -c 2.36
-savep-1
Here I use both the option -whole and -within, because this is needed
to test the interaction, but for the two t-contrast defined in
design_C1617.con should it only be -within? And if so how do I define
that?
I hope this was clear.
Kind Regards
Mette
On 2015-02-28 10:18, Anderson M. Winkler wrote:
> Hi Mette,
>
> Theres a tool now available that can shuffle the blocks as a whole
> and
> also within block. Please see here:
> https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/PALM [2]
>
> Read the documentation to see how to install it and use it. Please,
> feel free to ask if anything isnt clear.
>
> For contrasts testing within-subject effects, use "-within". For
> contrasts testing between-subject effects, use "-whole". For
> interactions, use "-within" together with "-whole".
>
> Note that the observations must be exchangeable within block for the
> "-within" (this means compound symmetry within block, which often
> isnt
> the case when the timepoints are too far apart), and the blocks as a
> whole must be exchangeable for the "-whole" (that is, the subjects
> must be independent, which normally is the case). This is described
> in
> the randomise paper [3].
>
> Youll need to rename the "design.mtx" to "design.mat", otherwise the
> file wont be read.
>
> PALM can also do MANOVA (assessed with permutations) if that could be
> useful for your design, as compound symmetry isnt assumed.
>
> All the best,
>
> Anderson
>
> On 27 February 2015 at 14:04, mettemn <[log in to unmask] [4]> wrote:
>
>> Hi experts
>>
>> I am new to FSL, and I want to do a permutation test using
>> randomise.
>>
>> I have a rs-FMRI dataset consisting of 42 patients, 22 of which
>> have received a drug and 20 who have received placebo. Each patient
>> have been scanned in three session: at baseline (week 0), at week 24
>> and week 48. So I have 2 groups each with 3 sessions.
>> Independent component analysis (ICA) has been performed on all 126
>> scans (42x3) volumes, and I now wish to perform a permutation test
>> on one of the components of interest (ICA output). These have been
>> concatenated in to a 4D file with ICA-map volumes for all my 126
>> scans ICA4D.nii.
>>
>> I have followed your guidelines in of your email correspondence
>> from July 2014 where you were asked questions about a similar study
>> design (2 groups each with 3 session). (see link)
>> https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;ab32c4cd.1407 [1]
>>
>> According to your guidelines I have created the design.mtx,
>> design.con and design.grp in the same way. I would like to use
>> contrast C2, C16, C17 and C18. (For C2 I use the option
>> -permuteBlocks to permute blocks as a whole.)
>> (i have attached the files design.mtx, design.grp and the .con for
>> the different contrasts to this email)
>>
>> I run the following code:
>>
>> randomise -i Sorted_ICA_4D.nii -o output -d design.mtx -t
>> design.con -e design.grp -n 5000 -c 0.001
>>
>> I have a question concerning contrast C18. You say that "For C18,
>> itd probably be a good idea to shuffle within-block in addition to
>> shuffle the blocks as a whole, but this option isnt yet available".
>> Is this an option that is available now (I'm using FSL version
>> 5.0.2.2), and how should this be implemented.
>> If not do you have an idea to how I can shuffle within blocks in
>> another way?
>>
>> Kind regards
>>
>> Mette Nielsen
>>
>> Master student at the Technical University of Denmark and Danish
>> Research Center for Magnetic Resonance
>
>
>
> Links:
> ------
> [1] https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;ab32c4cd.1407
> [2] https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/PALM
> [3]
> http://www.sciencedirect.com/science/article/pii/S1053811914000913
> [4] mailto:[log in to unmask]
%! VEST-Waveform File
/NumWaves 46
/NumContrasts 2
/Matrix
1 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 1 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
%! VEST-Waveform File
/NumWaves 2
/NumContrasts 1
/Matrix
1 -1
|