 |
 |
Dear Anderson
Thank you so much – this makes sense and works J
Just to be absolutely sure, the order of the eb file is the same as the order in both the 4D input file and the design matrix, i.e. each row 1 indicating subject 1 and so on.
Using the contrast you suggested, the option – cmcx is required. Is this a problem?
C1: [1 -1 0 0 ...], for Con > HR
C2: [1 0 -1 0 ...], for Con > Aff
C3: [0 1 -1 0 ...], for HR > Aff
C4: [-1 1 0 0 ...], for Con < HR
C5: [-1 0 1 0 ...], for Con < Aff
C6: [0 -1 1 0 ...], for HR < Aff
Best wishes
Iselin
Fra: FSL - FMRIB's Software Library [mailto:[log in to unmask]] På vegne af Anderson M. Winkler
Sendt: 4. januar 2017 12:33
Til: [log in to unmask]
Emne: Re: [FSL] SV: [FSL] SV: [FSL] fMRI analysis of non-independent groups
Hi Iselin,
Unfortunately, sibships of size 2 cannot be permuted with sibships of size 1. However, sibships of size 1 can be permuted with each other. Likewise, sibships of size 2 can be permuted with each other, and further, the two subjects within sibship can be permuted.
Say you have 7 sibships of size 1 (that is, singletons -- subjects that don't have any sibling in the sample), and 4 sibships of size 2 (that is, pairs), thus with 15 subjects total. Then the multi-level EB file would be like this:
-1 1 1 1
-1 1 2 1
-1 1 3 1
-1 1 4 1
-1 1 5 1
-1 1 6 1
-1 1 7 1
-1 2 1 1
-1 2 1 2
-1 2 2 1
-1 2 2 2
-1 2 3 1
-1 2 3 2
-1 2 4 1
-1 2 4 2
Hope this helps!
All the best,
Anderson
On 4 January 2017 at 11:09, Iselin Meluken <[log in to unmask]> wrote:
Hello again,
Thank you for all your help!
I have been struggling to design a multi-level eb file for my sample of twins.
The user guide states that between block shuffling requires same block size. Just to recap: I have a sample of MZ twins, where either both in the twin pair have been scanned, or just one from the twin pair have been scanned. This giving an unequal block size of 2 and 1 respectively.
I have copied a part of the file I have made below. I believe this file allows only within block permutations. Is this interpretation correct? Is there a way to code the file to allow for permutations between blocks of size 2 with blocks of size 1? (Column two represent twin numbers, and column 3 indication individual id numbers – the file is sorted by twin-number).
-1
92
65
-1
99
2
-1
99
4
-1
108
97
-1
110
91
-1
110
92
-1
121
88
-1
128
80
-1
128
85
-1
129
22
-1
131
72
Fra: FSL - FMRIB's Software Library [mailto:[log in to unmask]] På vegne af Anderson M. Winkler
Sendt: 6. december 2016 11:23
Til: [log in to unmask]
Emne: Re: [FSL] SV: [FSL] fMRI analysis of non-independent groups
Hi Iselin,
This is fine. However, to accommodate sibships of different sizes (e.g. with sizes 1 or 2) you'll need to use multi-level exchangeability blocks then. Please see some information here: http://fsl.fmrib.ox.ac.u
k/fsl/fslwiki/PALM/Exchangeabi lityBlocks
All the best,
Anderson
On 6 December 2016 at 08:36, Iselin Meluken <[log in to unmask]> wrote:
Dear Anderson
In your original answer to my question concerning permutation of a sample of twins you wrote:
I assume all sibships are complete and have size (cardinality) = 2
In our sample this is not the case, of 39 twinships are incomplete (only one twin in a twin-pair have been scanned). Is this a problem in group level inference using PALM?
Thanks in advance!
Best Iselin
Fra: FSL - FMRIB's Software Library [mailto:[log in to unmask]] På vegne af Anderson M. Winkler
Sendt: 6. oktober 2016 09:43
Til: [log in to unmask]
Emne: Re: [FSL] fMRI analysis of non-independent groups
Hi Iselin,
This can be done using permutation tests, although not yet in randomise. It is available in PALM though. You'd proceed as this:
1) Assemble a design matrix as usual for a 3-group comparison:
EV1: Control group, coded as 0 if not control, or 1 if control.
EV2: High-risk group, coded as 0 if not high-risk, 1 if high-risk.
EV3: Affected group, coded as 0 if not affected, 1 if affected.
EV4, etc: Additional nuisance variables as needed, e.g., age, sex, etc.
2) Define the contrasts also as a 3-group comparison. For instance:
C1: [1 -1 0 0 ...], for Con > HR
C2: [1 0 -1 0 ...], for Con > Aff
C3: [0 1 -1 0 ...], for HR > Aff
C4: [-1 1 0 0 ...], for Con < HR
C5: [-1 0 1 0 ...], for Con < Aff
C6: [0 -1 1 0 ...], for HR < Aff
3) Define a file with the exchangeability blocks, one such block per sibship. I assume all sibships are complete and have size (cardinality) = 2. If the subjects are entered in the design in pairs, the EB file would be something as:
1
1
2
2
3
3
4
4
... etc
4) Run PALM with the desired options, making sure to use the options "-eb <EB file>", "-within" and "-whole". Something as this:
palm -i 4d_copes.nii.gz -d design.mat -t design.con -eb design.grp -within -whole -n 2000 -corrcon -logp -o myresults [other options]
The way as the permutations are created for these cases is described in this paper. Subjects will be permuted within sibship, then the sibships will be permuted as a whole.
Interestingly, this design isn't properly seeking either "within-pair" or "between-pair" effects. Yet, it will inform about group differences while respecting the family relationships.
Hope this helps!
All the best,
Anderson
On 5 October 2016 at 17:23, Iselin Meluken <[log in to unmask]> wrote:
Dear Experts
I want to compare fMRI activity between three groups consisting of twins.
I want to be able to account for both non-independence between and within groups and unequal group size. We have performed fMRI using four paradigms assessing different domains of emotional processing.
The individual twins are grouped according to personal and co-twin history of affective disorders:
1) Control Group: Both twins have no personal history of affective disorders. Both twins in a twin pair are in the same group. N=29.
2) High Risk group: The healthy co-twins of twins with affective disorder. Only one twin from a twin pair is eligible in this group. N=33.
3) Affected Group: Twins with a personal history of affective disorder. Either the affected twin form a pair discordant for affective disrders or both twins from a twin pair concordant for affective disorders. N=63.
Would randomise/permutation testing be the best method of comparing these groups or would you recommend another approach? Is it possible to account for non-independence in traditional FEAT group analysis?
Looking forward to you response!
Best Regards
Iselin Meluken
Denne e-mail indeholder fortrolig information. Hvis du ikke er den rette modtager af denne e-mail eller hvis du modtager den ved en fejltagelse, beder vi dig venligst informere afsender om fejlen ved at bruge svarfunktionen. Samtidig bedes du slette e-mailen med det samme uden at videresende eller kopiere den.
Denne e-mail indeholder fortrolig information. Hvis du ikke er den rette modtager af denne e-mail eller hvis du modtager den ved en fejltagelse, beder vi dig venligst informere afsender om fejlen ved at bruge svarfunktionen. Samtidig bedes du slette e-mailen med det samme uden at videresende eller kopiere den.
Denne e-mail indeholder fortrolig information. Hvis du ikke er den rette modtager af denne e-mail eller hvis du modtager den ved en fejltagelse, beder vi dig venligst informere afsender om fejlen ved at bruge svarfunktionen. Samtidig bedes du slette e-mailen med det samme uden at videresende eller kopiere den.