Hello Freesurfer folks,
I hope this question is simple to answer: Is there any way to do a surface-based group analysis, a-la qdec, without the limitation of 2 continuous variables?
My goal is to analyze a thickness relationship with continuous variable X, which I have so far analyzed controlling for variable Y. However, I ALSO want to control for variable Z. As it is in qdec, it forces me to select only 2 of either X Y and Z.
I have tried converting one of the variables to a discrete factor, but the results do not look the same at all.
Any advice you can give will be very appreciated; and, as we are currently working under a deadline of sorts, the sooner someone might get back to me the better... as always, I really appreciate the help.
Victor Laluz : Imaging Coordinator : UCSF Memory and Aging Center 415-476-1680 : 415-476-1816 (fax)
you have a couple options:
1) assuming you are using 64b linux, you can download a new version of qdec which supports selecting 'nuisance' variables in the design. you can get it here: http://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc/linux-centos4_x86... copy qdec.bin to your $FREESURFER_HOME/bin/ dir. it will use your existing mri_glmfit to do the analysis, so its ok to overwrite the existing qdec.bin. let me know if you need it for another distribution.
or
2) dont use qdec, and create fsgd files instead (aka the command line option). some examples are found here:
https://surfer.nmr.mgh.harvard.edu/fswiki/FsgdExamples
n.
On Wed, 2010-03-17 at 11:34 -0700, Laluz, Victor wrote:
Hello Freesurfer folks,
I hope this question is simple to answer: Is there any way to do a surface-based group analysis, a-la qdec, without the limitation of 2 continuous variables?
My goal is to analyze a thickness relationship with continuous variable X, which I have so far analyzed controlling for variable Y. However, I ALSO want to control for variable Z. As it is in qdec, it forces me to select only 2 of either X Y and Z.
I have tried converting one of the variables to a discrete factor, but the results do not look the same at all.
Any advice you can give will be very appreciated; and, as we are currently working under a deadline of sorts, the sooner someone might get back to me the better... as always, I really appreciate the help.
Victor Laluz : Imaging Coordinator : UCSF Memory and Aging Center 415-476-1680 : 415-476-1816 (fax)
Nick,
Thank you so much. I'd love to try option 1, but I need centos 32bit if possible. This is great news.... thank you so much for the quick reply!
Victor ________________________________________ From: Nick Schmansky [nicks@nmr.mgh.harvard.edu] Sent: Wednesday, March 17, 2010 1:18 PM To: Laluz, Victor Cc: Bruce Fischl; freesurfer@nmr.mgh.harvard.edu Subject: Re: Replicating qdec analysis with more than 2 continuous variables
you have a couple options:
1) assuming you are using 64b linux, you can download a new version of qdec which supports selecting 'nuisance' variables in the design. you can get it here: http://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc/linux-centos4_x86... copy qdec.bin to your $FREESURFER_HOME/bin/ dir. it will use your existing mri_glmfit to do the analysis, so its ok to overwrite the existing qdec.bin. let me know if you need it for another distribution.
or
2) dont use qdec, and create fsgd files instead (aka the command line option). some examples are found here:
https://surfer.nmr.mgh.harvard.edu/fswiki/FsgdExamples
n.
On Wed, 2010-03-17 at 11:34 -0700, Laluz, Victor wrote:
Hello Freesurfer folks,
I hope this question is simple to answer: Is there any way to do a surface-based group analysis, a-la qdec, without the limitation of 2 continuous variables?
My goal is to analyze a thickness relationship with continuous variable X, which I have so far analyzed controlling for variable Y. However, I ALSO want to control for variable Z. As it is in qdec, it forces me to select only 2 of either X Y and Z.
I have tried converting one of the variables to a discrete factor, but the results do not look the same at all.
Any advice you can give will be very appreciated; and, as we are currently working under a deadline of sorts, the sooner someone might get back to me the better... as always, I really appreciate the help.
Victor Laluz : Imaging Coordinator : UCSF Memory and Aging Center 415-476-1680 : 415-476-1816 (fax)
Hi Nick,
could you please also provide us with a INTEL MAC qdec.bin binary?
Many thanks in advance,
Martin
On Wednesday 17 March 2010 21:18:19 Nick Schmansky wrote:
you have a couple options:
- assuming you are using 64b linux, you can download a new version of
qdec which supports selecting 'nuisance' variables in the design. you can get it here: http://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/misc/linux-centos4_x8 6_64/ copy qdec.bin to your $FREESURFER_HOME/bin/ dir. it will use your existing mri_glmfit to do the analysis, so its ok to overwrite the existing qdec.bin. let me know if you need it for another distribution.
Yuo can't do it in QDEC, but you can do it by calling mri_glmfit directly and creating an FSGD file. Do a search for FSGD on our wiki for more info.
doug
Laluz, Victor wrote:
Hello Freesurfer folks,
I hope this question is simple to answer: Is there any way to do a surface-based group analysis, a-la qdec, without the limitation of 2 continuous variables?
My goal is to analyze a thickness relationship with continuous variable X, which I have so far analyzed controlling for variable Y. However, I ALSO want to control for variable Z. As it is in qdec, it forces me to select only 2 of either X Y and Z.
I have tried converting one of the variables to a discrete factor, but the results do not look the same at all.
Any advice you can give will be very appreciated; and, as we are currently working under a deadline of sorts, the sooner someone might get back to me the better... as always, I really appreciate the help.
Victor Laluz : Imaging Coordinator : UCSF Memory and Aging Center 415-476-1680 : 415-476-1816 (fax)
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Douglas N Greve -
Laluz, Victor -
Martin Kavec -
Nick Schmansky