You should make sure to have a balanced design (ie, approx the same pats and cons at each site). Then include site as a regressor, ie, if you had two groups and two sites, you would have four regressors. You could then test for the effect of group, the effect of site, and the interaction. If there is an interaction, then it is hard to interpret the differences between groups.

On 05/08/2017 01:24 PM, Laura Ferrero Montes wrote:

Dear FS team, In case I have data from different centers and I would like to do a group analysis between Patients and Controls, how different centers effects could be regressed out? Should I have to analyze if there is an interaction between center and group? I have seen examples of continuous variables and groups but not in the case of categorical ones. Thank you, Laura _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

For the interaction it would be

0.5 -0.5 -0.5 0.5

On 5/9/17 8:32 AM, Laura Ferrero Montes wrote:

Thank you for the information. I just want to be sure if I have understood it properly. If I had two groups and two sites, my design matrix would be like:

GroupDescriptorFile 1 Class PatientsGroup1 Class PatientsGroup2 Class ControlsGroup1 Class ControlsGroup2

If I want to test for the effect of group, contrast matrix would be: 0.5 -0.5 0.5 -0.5

If I want to test for the effect of site, contrast matrix would be: 0.5 0.5 -0.5 -0.5

Finally, in case of interaction test, how contrast matrix would be?

Douglas N Greve greve@nmr.mgh.harvard.edu escribió:

...

You should make sure to have a balanced design (ie, approx the same pats and cons at each site). Then include site as a regressor, ie, if you had two groups and two sites, you would have four regressors. You could then test for the effect of group, the effect of site, and the interaction. If there is an interaction, then it is hard to interpret the differences between groups.

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Do you have unix permissions to the file? Can you send the file?

On 5/15/17 6:25 PM, Laura Ferrero Montes wrote:

Dear FS team, I was trying to do an analylisis using mri_glmfit command. The analysis consists of two groups of patients and I wanted to know if there was an interaction between age and group. I wrote the following lines:

mri_glmfit --table table1.txt --fsgd Descriptor.txt --C C1.txt --glmfir firstanalysis.glmdir

• table1 was obtained afther command aparc2stats

• C1 is the contrast matrix: 0 0 1 -1 0 0

• Descriptor.txt if the design matrix:

  GroupDescriptorFile 1 Title PatientsControls Class Patient Class Control Variables Age IQ Input NYU_0050985 Patient 13.09 90 Input NYU_0051099 Control 13.72 87 Input NYU_0051021 Patient 23.66 112 Input NYU_0051067 Control 23.65 116 Input NYU_0050996 Patient 15.81 100 Input NYU_0051104 Control 15.27 104 . . .

I got this error:
gdfRead(): reading Descriptor.txt
ERROR: fio_FileHasCarriageReturn(): cannot open Descriptor.txt


I created this matrix using text editor in Linux. Would yo know

what is the problem? Thank you

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Do you mean that you are doing an ROI analysis on the volume of the ROI? If so, then you still need to correct for multiple ROIs, probably with bonferroni or FDR.

On 5/31/17 8:10 AM, Laura Ferrero Montes wrote:

Dear FS team, I was doing a statistical analysis using metrics that are provided in stats files like volume of different regions, so I am not doing a voxel based morphometry. It is necessary to correct for multiple comparisons or as I am only comparing the measures of the regions between to grups, it is not ?

Thank you in advance _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer