open-discussion
20326 Subscribers 
open-discussion > covariates design CONN
Showing 1-4 of 4 posts
Apr 22, 2015 01:04 PM | Bahri Mohamed Ali - Cyclotron Research Center, University of Liege
covariates design CONN
Hello,
I have 4 groups (3 of patients and 1 control group). I would like to add covariates. When I add disease duration, I add 0 for controls, of course. In spm if I enter all values for the 4 four groups in the same column spm assigns to each subject the mean of my values. In order to avoid that spm offers the possibility to split the covariates in groups (interaction and centering). Is there a possibility to do it with CONN?
Thanks in advance for your help!
I have 4 groups (3 of patients and 1 control group). I would like to add covariates. When I add disease duration, I add 0 for controls, of course. In spm if I enter all values for the 4 four groups in the same column spm assigns to each subject the mean of my values. In order to avoid that spm offers the possibility to split the covariates in groups (interaction and centering). Is there a possibility to do it with CONN?
Thanks in advance for your help!
Apr 24, 2015 09:04 PM | Alfonso Nieto-Castanon - Boston University
RE: covariates design CONN
Hi,
Yes, in CONN you would do that 'semi-manually'. If you have four groups (e.g. named 'PatientsA', 'PatientsB', 'PatientsC', and 'Controls') and a disease duration covariate (e.g. named 'Duration') with values for your patients and 0's for your controls, you could create the interaction terms manually doing the following (in Setup.Covariates.SecondLevel):
a) create a new variable named 'Duration_PatientsA' and enter in the values field the string "Duration.*PatientsA" (without the quotes). The new 'Duration_PatientsA' variable will contain the 'Duration' values for the PatientsA subjects and 0's for everyone else
b) repeat the step above for the PatientsB and PatientsC groups
If you want to then test the Duration by Group interaction with functional connectivity (i.e. if the association between connectivity and disease duration is different across your three patient groups) you can do so (in Second-level results tab) by selecting the 'PatientsA', 'PatientsB', 'PatientsC', 'Duration_PatientsA', 'Duration_PatientsB', and 'Duration_PatientsC' between-subject effects, and entering a contrast [0 0 0 1 0 0;0 0 0 0 1 0;0 0 0 0 0 1] (or by clicking on the contrast dropdown menu and selecting the one named "F-test Any effect across Duration_PatientsA & Duration_PatientsB & Duration_PatientsC")
Let me know if this is what you had in mind.
Alfonso
ps. also just for future reference, please post CONN related questions to the CONN-specific NITRC forum at https://www.nitrc.org/forum/forum.php?forum_id=1144 instead of to the general NITRC community forum (this one). To access the former you can also simply go to the NITRC CONN site at http://www.nitrc.org/projects/conn and click on 'support' (or from the CONN gui itself simply click on 'Help' and 'Support')
Originally posted by Bahri Mohamed Ali:
Yes, in CONN you would do that 'semi-manually'. If you have four groups (e.g. named 'PatientsA', 'PatientsB', 'PatientsC', and 'Controls') and a disease duration covariate (e.g. named 'Duration') with values for your patients and 0's for your controls, you could create the interaction terms manually doing the following (in Setup.Covariates.SecondLevel):
a) create a new variable named 'Duration_PatientsA' and enter in the values field the string "Duration.*PatientsA" (without the quotes). The new 'Duration_PatientsA' variable will contain the 'Duration' values for the PatientsA subjects and 0's for everyone else
b) repeat the step above for the PatientsB and PatientsC groups
If you want to then test the Duration by Group interaction with functional connectivity (i.e. if the association between connectivity and disease duration is different across your three patient groups) you can do so (in Second-level results tab) by selecting the 'PatientsA', 'PatientsB', 'PatientsC', 'Duration_PatientsA', 'Duration_PatientsB', and 'Duration_PatientsC' between-subject effects, and entering a contrast [0 0 0 1 0 0;0 0 0 0 1 0;0 0 0 0 0 1] (or by clicking on the contrast dropdown menu and selecting the one named "F-test Any effect across Duration_PatientsA & Duration_PatientsB & Duration_PatientsC")
Let me know if this is what you had in mind.
Alfonso
ps. also just for future reference, please post CONN related questions to the CONN-specific NITRC forum at https://www.nitrc.org/forum/forum.php?forum_id=1144 instead of to the general NITRC community forum (this one). To access the former you can also simply go to the NITRC CONN site at http://www.nitrc.org/projects/conn and click on 'support' (or from the CONN gui itself simply click on 'Help' and 'Support')
Originally posted by Bahri Mohamed Ali:
Hello,
I have 4 groups (3 of patients and 1 control group). I would like to add covariates. When I add disease duration, I add 0 for controls, of course. In spm if I enter all values for the 4 four groups in the same column spm assigns to each subject the mean of my values. In order to avoid that spm offers the possibility to split the covariates in groups (interaction and centering). Is there a possibility to do it with CONN?
Thanks in advance for your help!
I have 4 groups (3 of patients and 1 control group). I would like to add covariates. When I add disease duration, I add 0 for controls, of course. In spm if I enter all values for the 4 four groups in the same column spm assigns to each subject the mean of my values. In order to avoid that spm offers the possibility to split the covariates in groups (interaction and centering). Is there a possibility to do it with CONN?
Thanks in advance for your help!
Apr 24, 2015 11:04 PM | Alfonso Nieto-Castanon - Boston University
RE: covariates design CONN
Sorry, too little caffeine, this is a
correction on my previous post:
The contrast [0 0 0 1 0 0;0 0 0 0 1 0;0 0 0 0 0 1] (which would appear in the GUI listed as "F-test Any effect across Duration_PatientsA or Duration_PatientsB or Duration_PatientsC") is testing the main effect of duration without collapsing across groups (i.e. is functional connectivity associated with disease duration in any of the patient groups?), NOT the interaction of group by duration (i.e. is the functional connectivity association with disease duration the same or different across the three patient groups?) as it was originally stated in my previous post. That would of course be tested using the contrast [0 0 0 1 -1 0;0 0 0 0 1 -1] (which would appear listed as "F-test Any difference among Duration_PatientsA & Duration_PatientsB & Duration_PatientsC").
Hope this helps
Alfonso
Originally posted by Alfonso Nieto-Castanon:
The contrast [0 0 0 1 0 0;0 0 0 0 1 0;0 0 0 0 0 1] (which would appear in the GUI listed as "F-test Any effect across Duration_PatientsA or Duration_PatientsB or Duration_PatientsC") is testing the main effect of duration without collapsing across groups (i.e. is functional connectivity associated with disease duration in any of the patient groups?), NOT the interaction of group by duration (i.e. is the functional connectivity association with disease duration the same or different across the three patient groups?) as it was originally stated in my previous post. That would of course be tested using the contrast [0 0 0 1 -1 0;0 0 0 0 1 -1] (which would appear listed as "F-test Any difference among Duration_PatientsA & Duration_PatientsB & Duration_PatientsC").
Hope this helps
Alfonso
Originally posted by Alfonso Nieto-Castanon:
Hi,
Yes, in CONN you would do that 'semi-manually'. If you have four groups (e.g. named 'PatientsA', 'PatientsB', 'PatientsC', and 'Controls') and a disease duration covariate (e.g. named 'Duration') with values for your patients and 0's for your controls, you could create the interaction terms manually doing the following (in Setup.Covariates.SecondLevel):
a) create a new variable named 'Duration_PatientsA' and enter in the values field the string "Duration.*PatientsA" (without the quotes). The new 'Duration_PatientsA' variable will contain the 'Duration' values for the PatientsA subjects and 0's for everyone else
b) repeat the step above for the PatientsB and PatientsC groups
If you want to then test the Duration by Group interaction with functional connectivity (i.e. if the association between connectivity and disease duration is different across your three patient groups) you can do so (in Second-level results tab) by selecting the 'PatientsA', 'PatientsB', 'PatientsC', 'Duration_PatientsA', 'Duration_PatientsB', and 'Duration_PatientsC' between-subject effects, and entering a contrast [0 0 0 1 0 0;0 0 0 0 1 0;0 0 0 0 0 1] (or by clicking on the contrast dropdown menu and selecting the one named "F-test Any effect across Duration_PatientsA & Duration_PatientsB & Duration_PatientsC")
Let me know if this is what you had in mind.
Alfonso
ps. also just for future reference, please post CONN related questions to the CONN-specific NITRC forum at https://www.nitrc.org/forum/forum.php?forum_id=1144 instead of to the general NITRC community forum (this one). To access the former you can also simply go to the NITRC CONN site at http://www.nitrc.org/projects/conn and click on 'support' (or from the CONN gui itself simply click on 'Help' and 'Support')
Originally posted by Bahri Mohamed Ali:
Yes, in CONN you would do that 'semi-manually'. If you have four groups (e.g. named 'PatientsA', 'PatientsB', 'PatientsC', and 'Controls') and a disease duration covariate (e.g. named 'Duration') with values for your patients and 0's for your controls, you could create the interaction terms manually doing the following (in Setup.Covariates.SecondLevel):
a) create a new variable named 'Duration_PatientsA' and enter in the values field the string "Duration.*PatientsA" (without the quotes). The new 'Duration_PatientsA' variable will contain the 'Duration' values for the PatientsA subjects and 0's for everyone else
b) repeat the step above for the PatientsB and PatientsC groups
If you want to then test the Duration by Group interaction with functional connectivity (i.e. if the association between connectivity and disease duration is different across your three patient groups) you can do so (in Second-level results tab) by selecting the 'PatientsA', 'PatientsB', 'PatientsC', 'Duration_PatientsA', 'Duration_PatientsB', and 'Duration_PatientsC' between-subject effects, and entering a contrast [0 0 0 1 0 0;0 0 0 0 1 0;0 0 0 0 0 1] (or by clicking on the contrast dropdown menu and selecting the one named "F-test Any effect across Duration_PatientsA & Duration_PatientsB & Duration_PatientsC")
Let me know if this is what you had in mind.
Alfonso
ps. also just for future reference, please post CONN related questions to the CONN-specific NITRC forum at https://www.nitrc.org/forum/forum.php?forum_id=1144 instead of to the general NITRC community forum (this one). To access the former you can also simply go to the NITRC CONN site at http://www.nitrc.org/projects/conn and click on 'support' (or from the CONN gui itself simply click on 'Help' and 'Support')
Originally posted by Bahri Mohamed Ali:
Hello,
I have 4 groups (3 of patients and 1 control group). I would like to add covariates. When I add disease duration, I add 0 for controls, of course. In spm if I enter all values for the 4 four groups in the same column spm assigns to each subject the mean of my values. In order to avoid that spm offers the possibility to split the covariates in groups (interaction and centering). Is there a possibility to do it with CONN?
Thanks in advance for your help!
I have 4 groups (3 of patients and 1 control group). I would like to add covariates. When I add disease duration, I add 0 for controls, of course. In spm if I enter all values for the 4 four groups in the same column spm assigns to each subject the mean of my values. In order to avoid that spm offers the possibility to split the covariates in groups (interaction and centering). Is there a possibility to do it with CONN?
Thanks in advance for your help!
Jun 8, 2021 12:06 PM | Reza Momenan - CNIRC, NIAAA, NIH
RE: covariates design CONN
I am sorry if this question has redundancies. But, reading through
the posts every time I think I got it, I have worries about my
approach. So,as a novice user I wanted to make sure that I
understand this correctly. I have two groups of patients and a
control group, male and female and a number of covariates such as
age, education, score1, score2, score3. I have loaded covariate
files such as patient1 (0 & 1), pateinet2 (0 &1), control
(0&1), SexMale(0 &1), SexFemale (0&1), and the rest of
the covariates, resulting in a 9 elements vector.
My question is in order to see the effect of any of the covariates on the differences and interactions between any two groups and sexes, do I have to subpart each of the covariates to 12 (3groupsx2sex)? If that is the case how do I set up the contrast to see the difference in connectivity cased by score2 between male and female subjects of patient1 vs patient2? is this right?
[1 -1 0 1 -1 0 1 0]
Is there any simpler way without have to create multiple files for the same covariates?
Be happy to get on a call or whatever to make sure my question makes sense.
Thank you for your help.
My question is in order to see the effect of any of the covariates on the differences and interactions between any two groups and sexes, do I have to subpart each of the covariates to 12 (3groupsx2sex)? If that is the case how do I set up the contrast to see the difference in connectivity cased by score2 between male and female subjects of patient1 vs patient2? is this right?
[1 -1 0 1 -1 0 1 0]
Is there any simpler way without have to create multiple files for the same covariates?
Be happy to get on a call or whatever to make sure my question makes sense.
Thank you for your help.