Thank you Anderson and Rosalia for your responses!
~meisan
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Rosalia Dacosta Aguayo
Sent: Thursday, August 04, 2016 1:25 AM
To: [log in to unmask]<mailto:[log in to unmask]>
Subject: Re: [FSL] Two group unpaired t-test design matrix using glm_gui
Hi Anderson, thank you for your answer. I was worried by the fact she was entering 1 and 2 values in the column of controls as well as in the column of patients...I guess that something has changed in the last release of FSL...the contrasts are fine...she can covariate for age and attention....but when she is asking for the effect of age and attention....it would be more interesting to see those effects for every group separately and give a look for intercepts...for example.
I am sorry...I have to give another look into this.
Best wishes,
Rosalía
El 4 ago. 2016 9:42 AM, "Anderson M. Winkler" <[log in to unmask]<mailto:[log in to unmask]>> escribió:
Hi Rosalia,
The design as posted by Meisan is correct. It is identical to the one you suggest. The order in which the rows are entered doesn't matter.
All the best,
Anderson
On 4 August 2016 at 01:05, Rosalia Dacosta Aguayo <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Hi Meisan,
The design is not fine. For example..imagine you have four controls and four patients, that is a total of 8 subjects.
The first column that appears in the Glm and says "Group"...do not change anything here.
I understand that EV1 is for Group A and EV2 is for group B. You have not differenciate them properly...it should be as follows:
EV1 control EV2 patients EV3 EV4
1 0
1 0
1 0
1 0
0 1
0 1
0 1
0 1
You can demean age and attention by hand or simply add -D when you run randomise.
As you say you are interested in differences between groups adding age and attention as covariates...the contrasts should be:
C1 1 -1 0 0
C2 -1 1 0 0
But, in your design it seems you are interested in the effect/association of age and attention for every group...in this case, you should split age in 2 EV one column for age in the control group and another column for age in the patient group instead of adding them together in the same EV...the same would be for attention....This is because you are interested in inspecting the effect of age and attention in every group separately.
I would recommend reading the following in order you can understand the process better:
http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM
Here you will find examples related to what you are asking.
I also recommend Jeanette Mumford lectures You will find a lot of classes in youtube.
Hope this helps.
Rosalía
El 3 ago. 2016 10:52 PM, "Meisan Brown-Lum" <[log in to unmask]<mailto:[log in to unmask]>> escribió:
>
> Hi, I am confused by the instructions about setting up the design matrix in the glm_gui. Can someone please confirm whether my design matrix is correct. I am running a two group unpaired ttest to see if the mean FA (and MD) differ. I would also like to add age and attention measures as covariates (they have been demeaned). Below is a sample design matrix.
>
> Input
> Group
> EV1
> EV2
> EV3
> EV4
> Control
> Patient
> Age
> Attention
> 1
> 1
> 1
> 0
> -1.0
> -16.4
> 2
> 2
> 0
> 1
> 4.0
> -25.4
> 3
> 2
> 0
> 1
> 1.0
> 25.4
> 4
> 1
> 1
> 0
> -4.0
> 16.4
>
>
> Here is the contrast file:
>
> Title
> EV1
> EV2
> EV3
> EV4
> C1
> Control>Patient
> 1
> -1
> 0
> 0
> C2
> Patient>Control
> -1
> 1
> 0
> 0
> C3
> pos age effect
> 0
> 0
> 1
> 0
> C4
> neg age effect
> 0
> 0
> -1
> 0
> C5
> pos attention effect
> 0
> 0
> 0
> 1
> C6
> neg attention effect
> 0
> 0
> 0
> -1
>
> Here is the script i used to run the stats analysis:
>
> randomise -i all_FA_skeletonised.nii.gz -o tbss -m mean_FA_skeleton_mask -d design.mat -t design.con -n 5000 —T2
>
> Thank you in advance for your help!
> ~mei
