I’ve now attempted to use concatenation in melodic instead of tensor and my output is completely different. Many of the components are clearly noise. So I’m left unsure of what to stick with. According to the descriptions of both, tensor is what makes more sense. My participants all did the exact same task during the trial and I’m not looking at resting state.
--
Mayte Parada Ph.D.
____________________________
Laboratory for the Biopsychosocial Study of Sexuality
McGill University
1205 Docteur Penfield 
Montreal QC
Canada H3A 1B1
[log in to unmask]
(514) 398-5323

On Jan 9, 2015, at 1:08 PM, David V. Smith <[log in to unmask]> wrote:

Yes, I believe so. And yes, the t*.txt is indicating the temporal response of the component for all the inputs/subjects.

The dual_regression script is one line of code, so it should be pretty to easy to run. 

e.g.   dual_regression groupICA.gica/groupmelodic.ica/melodic_IC 1 design.mat design.con 500 grot `cat groupICA.gica/.filelist`

Work through the tutorials and guides, and then give it a try.

Cheers,
David



On Jan 9, 2015, at 1:00 PM, Mayte Parada, Ms. <[log in to unmask]> wrote:

Hi David, yes I chose tensor-based ICA this time…that explains the output then? I’ll try running it with concatenation instead and work with that output. Am I correct in assuming that those numbers in the t*.txt files represent activation for the entire network of brain regions across time?

Thanks for the tip on comparing men and women, I’m still really new to FSL so I’m not experienced enough to run it using code. Is there a way to do this with the GUI? 

Thanks for all your help,

--
Mayte Parada Ph.D.
____________________________
Laboratory for the Biopsychosocial Study of Sexuality
McGill University
1205 Docteur Penfield 
Montreal QC
Canada H3A 1B1
[log in to unmask]
(514) 398-5323

On Jan 9, 2015, at 11:58 AM, David V. Smith <[log in to unmask]> wrote:

Hi,

Looks like you might be running the tensor-based ICA? Did all of your subjects do the same task with the same timing? I suspect the extra columns you're seeing have something to do with the subjects (see https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=fsl;c54a748f.1004 for more details).

If you use concat instead of tica, you'll only have one column in each t*.txt file. 

In either case, you could consider using the dual_regression script. If you're interesting the activation of each network (rather than the connectivity with each network), you could regress the resulting dr_stage1*.txt files onto your design matrices.

Cheers,
David



On Jan 9, 2015, at 10:39 AM, Mayte Parada, Ms. <[log in to unmask]> wrote:

Hi David,
Thanks for the links, I have looked through those for melodic but I’m still not clear on what the 6 or 7 columns of numbers represent in .txt files. I want to correlate the change in activity across time with temperature data I’ve collected. If I click on the time course plot for each component (normalized response across time) I see several columns of numbers. I was expecting one column of numbers just like what I see in the power spectrum time courses so I’m not clear on what those six columns represent. 
--
Mayte Parada Ph.D.
____________________________
Laboratory for the Biopsychosocial Study of Sexuality
McGill University
1205 Docteur Penfield 
Montreal QC
Canada H3A 1B1
[log in to unmask]
(514) 398-5323

On Jan 9, 2015, at 9:33 AM, David V. Smith <[log in to unmask]> wrote:

Hi May,

You can compare connectivity using dual regression:

Here are some other guides and tutorials:

Cheers,
David



On Jan 9, 2015, at 7:56 AM, Mayte Parada, Ms. <[log in to unmask]> wrote:

Hi experts, I've run ICA on two groups (men and women) and want to compare the differences in the networks activated between those two groups but I'm not clear if that can be done.
Furthermore, I'd like to understand the output from ICA. Mainly, when you click on the t.txt files in the ICA reports folder there are several columns of data in each text file that corresponds to each component. I'm not clear on what those columns are. Is there a section in the tutorials that explains this?

May