On 11 Apr 2019, at 14:04, FERNANDA MEIRELES <[log in to unmask]> wrote:
Dear Mr Matthew,
Let me see if I understood:
First of all: I did not use the -R option when I runned randomise but I have this data: tbss_FA_IMC_long_dir_tstat1.nii.gz (beyond the datas like: tbss_FA_IMC_long_dir_tfce corrp_tstat1.nii.gz); so with this I can extract the t-values, is that right?
Then, I can use the formula: r = t/sqrt(t^2+df), putting the df like: the number of subjects (n= 101) - 1 ? But, made a regression in GLM, I should discount one value for using -D and another one value for the number of EVs (EVs = 2)
So: my df will be = 101 - (1 + 2) = 98???
<image.png>
<image.png>
Is that correct?
Respectfully yours.
Fernanda Meireles
Hello Fernanda,If you run randomise with the the -R option, you will get the “raw” t-stats ( the original statistic image ) for use with the formula in my previous email. You could also try something like Fischer’s transformation if you want to use the ( 1 - p ) values instead.
Note that fslstats -R is returning the _maximum_ value ( here 1 - p ) for the input, it might also be worth considering the -M ( mean of non-zero voxels ) instead as that might be a more indicative figure.
Kind RegardsMatthew
--------------------------------
Dr Matthew Webster
FMRIB Centre
John Radcliffe Hospital
University of Oxford
On 8 Apr 2019, at 14:29, FERNANDA MEIRELES <[log in to unmask]> wrote:
Thanks, Mathew....
Do you mind if I put an example of what I've done so far?
- randomise_parallel -i all_FA.nii.gz -o TBSS_FA_cingulum_left -m cing_left_JHU_1mm_mask.nii.gz -d design.mat -t design.con -n 5000 -D --T2
Then, I used:
- fslstats TBSS_FA_cingulum_left_tfce_corrp_tstat1.nii.gz -R
- fslstats TBSS_FA_cingulum_left_tfce_corrp_tstat2.nii.gz -R
So, as results I had something like:tstat1: 0,0000 0,348213tstat2: 0,0000 0,575439
And for my p-value, I made:
= 1 - tstat1 = 1 - 0,348213 = p value (FA_ positive contrast - correlation)
0,651787
= 1 - tstat2 = 1- 0,575439 = p value (FA_negative contrast - correlation)
0,424561
Now, to find the -r value, I need to:convert the “raw” t-values output for your contrast to a correlation using r = t/sqrt(t^2+df)? - as you wrote before.
Is everything right?
Respectfully,
Fernanda
Hello,
You can convert the “raw” t-values output for your contrast to a correlation using r = t/sqrt(t^2+df) where df is the degrees of freedom ( note the implicit demeaning introduced with the -D option will remove an extra degree of freedom ).
Kind Regards
Matthew
--------------------------------
Dr Matthew Webster
FMRIB Centre
John Radcliffe Hospital
University of Oxford
> On 8 Apr 2019, at 13:48, FERNANDA MEIRELES <[log in to unmask]> wrote:
>
> Thank you, Anderson for your reply but the thing that I really want to know is:
>
> HOW DO I FIND THE CORRELATION VALUE (-r) USING FSL?
>
> By the way, ... yes, I used DEMEAN on randomise.
>
> Respectfully,
>
> Fernanda
>
> On Sun, Apr 7, 2019 at 4:52 PM Anderson M. Winkler <[log in to unmask]> wrote:
> Hi Fernanda,
>
> Design is fine. Just follow the randomise manual. Since your design doesn't have an intercept, make sure you use the option -D, to mean-center both data and design.
>
> All the best,
>
> Anderson
>
>
> On Wed, 3 Apr 2019 at 15:41, Fernanda Meireles Ferreira <[log in to unmask]> wrote:
> Dear FSL's expert,
>
> I would like to know about the correlation value (-r).
> We did an analysis with TBSS correlating FA with a score (numerical variables). The design matrix was done in FSL itself in GLM setup with 2 contrasts: postive (+1) and negative (-1) as you can see in the image below. (the GLM bellow was with AGE as a variable of no interest).
>
> Please, can anyone help me?
>
> Thank you,
>
> Fernanda
>
> <image.png>
>
>
>
> <image.png>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> Fernanda G. Meireles Ferreira
> Medical Physicist
> IDOR - D'Or Institute for Research and Education
> Rua Diniz Cordeiro, 30 – 3º andar
> Botafogo - Rio de Janeiro - RJ - Brazil
> CEP: 22281-100
> Phone: +55 21 3883-6000 / +55 21 99998-7892 / +55 21 7829-1558
> Site: www.idor.org
>
>
> The D'Or Institute is a not for profit research and educational organization partially supported by independent grants from the Rede D'Or São Luiz health organization.
>
> To unsubscribe from the FSL list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
>
>
> To unsubscribe from the FSL list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
>
>
> To unsubscribe from the FSL list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
>
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
--
Fernanda G. Meireles Ferreira
Medical PhysicistIDOR - D'Or Institute for Research and Education
Rua Diniz Cordeiro, 30 – 3º andarBotafogo - Rio de Janeiro - RJ - BrazilCEP: 22281-100Phone: +55 21 3883-6000 / +55 21 99998-7892Site: www.idor.org
The D'Or Institute is a not for profit research and educational organization partially supported by independent grants from the Rede D'Or São Luiz health organization.
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1