On 2 Sep 2018, at 03:38, chenhf_uestc <[log in to unmask]> wrote:
Hi, MatthewYes, my design matrix has three columns, and I just would like to demean the last column (age). So, I must manually demean age column rather than using "-D" option, right?If the "-D" option is to demean all the columns separately, when should I use this option? Could you give me an example?In the command randomise: -D demean data temporally before model fitting ( demean model as well if required ). By default, this option is just demean the data (the dependent variable)? Or demean the data and design matrix (model, independent variable) simultaneously? "demean model as well if required", I think, by default, it just demean the data; and if I want to demean the model at the meantime, I should add additional option to implement it, right?Best,Feng
发件人：Matthew Webster <[log in to unmask]>发送时间：2018-08-24 23:54主题：Re: [FSL] How to mean centering covariate in the design matrix in FSL收件人："FSL"<[log in to unmask]>抄送：Hello Feng,
1) “Temporally” means across subjects - since we use the 4th NIFTI dimension to refer to subjects. Demeaning the model refers to removing the column-wise mean from your design.
2) The two methods you have tried here are not equivalent, I suspect you just want to _only_ demean the third-column in your design, in which case the -D option is not going to be helpful here.
Hope this helps,Kind RegardsMatthew
--------------------------------
Dr Matthew Webster
FMRIB Centre
John Radcliffe Hospital
University of Oxford
On 23 Aug 2018, at 11:27, chenhf_uestc <[log in to unmask]> wrote:
Thanks, Matthew,Attached are the design matrix and contrast files that I used. Actually, I have two groups (6 subjects vs. 6 subjects) and there are four contrast: 1) 1 -1 0; 2) -1 1 0; 3) 1 0 0; 4) 0 1 0;Then I type: randomise -i all_FA_skeletonised.nii.gz -o tbss_D -m mean_FA_skeleton_mask.nii.gz -d design.mat -t design.con -n 5000 --T2 -x -D --uncorrp
The command window display:Critical Value for: tbss_D_vox_tstat2 is: 12.1341
Critical Value for: tbss_D_tfce_tstat2 is: 103504
924 permutations required for exhaustive test of t-test 3
Doing all 924 unique permutations
Starting permutation 1 (Unpermuted data)
Starting permutation 2
Warning: tfce has detected a large number of integral steps. This operation may require a great deal of time to complete.I wait for a long time and it's still Starting permutation 2.However, if I demean the covariate manually, and type: randomise -i all_FA_skeletonised.nii.gz -o tbss_manual -m mean_FA_skeleton_mask.nii.gz -d design_mean.mat -t design.con -n 5000 --T2 -x --uncorrp
There will output all the results of four contrasts and no warnings. Why?Another question is that randomise -D: demean data temporally before model fitting ( demean model as well if required ).What does temporally mean? Across subjects? What does demean model mean? Could you explain a little more?Best,Feng
<design.mat><design.con>发件人：Matthew Webster <[log in to unmask]>发送时间：2018-08-21 22:17主题：Re: [FSL] How to mean centering covariate in the design matrix in FSL收件人："FSL"<[log in to unmask]>抄送：Hello,Could you say why the -D option appears to not be mean-entering your covariate? For discrete covariates ( as with continuous ) the decision of whether or not to demean depends on the contrasts of those covariates and how their values will be interpreted.
Kind RegardsMatthew
--------------------------------
Dr Matthew Webster
FMRIB Centre
John Radcliffe Hospital
University of Oxford
On 17 Aug 2018, at 10:33, chenhf_uestc <[log in to unmask]> wrote:
Dear FSL expert,I have a question about the mean centering covariate for a group fMRI analysis in FSL (http://mumford.fmripower.org/mean_centering/).When I conduct group level analysis (randomise, two-sample t-tests), I want to mean centering covariate in the design matrix.I used to manually mean centering covariate (or try to modify the design.mat file: mean centering covariate in the MATLAB or EXCEL, copy the centering value, copy and replace in the original value in the design.mat). Here, I am wondering is there a convenient method to implement my goal? I have check the -D option in the randomise: -D demean data temporally before model fitting ( demean model as well if required ). However, it seems not to mean centering covariate.Another question is that if my covariate is discrete covariates, need I perform mean centering?Best,Feng
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