help
help > RE: Questions about setup & within subject design
Mar 7, 2012 08:03 PM | Alfonso Nieto-Castanon - Boston University
RE: Questions about setup & within subject design
Hi Darren,
Just a few minor notes:
5) CONDITIONS: define two conditions instead ('pre' and 'post'), and associate each condition with a single session. For example, for the 'pre' condition you would select the first session and fill the onset/duration fields, and for the 'post' condition you would select the second session and fill the onset/duration fields (note that in both cases the other session onset/duration fields should be left empty; this basically tells conn that for the 'pre' condition it should use all scans in the first session while for the 'post' condition it should use all scans in the second session)
7) COVARIATES (second-level): this is fine, since you have only one group of subjects you would define just one covariate (named for example 'All subjects') that contains a list of 1's (one per subject)
10) FIRST LEVEL ANALYSIS TAB: for resting state data it is perfectly fine to leave the 'weight' field to 'none', other weighting functions are most helpful when you have block designs and want to minimize border effects between consecutive blocks (hrf weighting is the defaut value we use and it weights down the influence of the first few scans in each block, while hanning is a more conservative way of weighting down the influence of border scans within each block); of course you can still use hrf weighting for resting state data for example if you feel that you should weight down the influence of the first few scans within each session (we typically do that just to minimize any potential 'ramping' effects).
To compare session 1 and session 2 within subjects and then across subjects you would simply select in the second-level results tab both conditions (pre and post) and enter a [-1,1] 'between-conditions' contrast when defining your second level analyses (you would also select the single 'All subjects' effect in the 'between-subjects effects' list). Just for reference, 'conditions' and 'sources' (or 'measures' for voxel-to-voxel analyses) are always treated as within-subject effects while 2nd-level covariates are always treated as between-subject effects. Hope this clarifies a bit.
Best
Alfonso
Originally posted by Darren Gitelman:
Just a few minor notes:
5) CONDITIONS: define two conditions instead ('pre' and 'post'), and associate each condition with a single session. For example, for the 'pre' condition you would select the first session and fill the onset/duration fields, and for the 'post' condition you would select the second session and fill the onset/duration fields (note that in both cases the other session onset/duration fields should be left empty; this basically tells conn that for the 'pre' condition it should use all scans in the first session while for the 'post' condition it should use all scans in the second session)
7) COVARIATES (second-level): this is fine, since you have only one group of subjects you would define just one covariate (named for example 'All subjects') that contains a list of 1's (one per subject)
10) FIRST LEVEL ANALYSIS TAB: for resting state data it is perfectly fine to leave the 'weight' field to 'none', other weighting functions are most helpful when you have block designs and want to minimize border effects between consecutive blocks (hrf weighting is the defaut value we use and it weights down the influence of the first few scans in each block, while hanning is a more conservative way of weighting down the influence of border scans within each block); of course you can still use hrf weighting for resting state data for example if you feel that you should weight down the influence of the first few scans within each session (we typically do that just to minimize any potential 'ramping' effects).
To compare session 1 and session 2 within subjects and then across subjects you would simply select in the second-level results tab both conditions (pre and post) and enter a [-1,1] 'between-conditions' contrast when defining your second level analyses (you would also select the single 'All subjects' effect in the 'between-subjects effects' list). Just for reference, 'conditions' and 'sources' (or 'measures' for voxel-to-voxel analyses) are always treated as within-subject effects while 2nd-level covariates are always treated as between-subject effects. Hope this clarifies a bit.
Best
Alfonso
Originally posted by Darren Gitelman:
Hi.
I am trying to use the conn toolbox to analyze within subject resting state data. There is a single group and they have a resting state scan pre / post a treatment.
Please tell me if I am setting up the analysis correctly.
1) BASIC: number of sessions 2 (for each subject). Other entries self-explanatory
2) FUNCTIONAL: Associated resting scans with each session. (There are 196 scans, TR=2.5, duration = 490 seconds).
3) STRUCTURAL: no questions.
4) ROIs: I used all the default ROI's and selected the subject's segmented gray matter, white matter and CSF images.
Should I threshold these images? What value? Perhaps this will make each mask more exclusive?
The dimensions are set at default of 16 for white matter and csf and regress out covariates is checked.
For gray matter, dimensions is 1 and nothing is checked.
5) CONDITIONS: 1 condition called rest, and associated that with each session. Is that correct?
6) COVARIATES: First level: 1 covariate called movement. I associated it with each session and selected the relevant rp*.txt file.
7) COVARIATES: Second level: Not sure what to enter here. It only seems to accept a single number such as 1 or -1.
8) OPTIONS: I clicked all planned analyses.
9) PREPROCESSING TAB: I selected white matter, CSF and movement as confounds.
10) FIRST LEVEL ANALYSIS TAB: If I select a region such as PCC as source would my weight be "none". When would I use something like HRF or Hanning?
I'm not sure how to compare session 1 and 2 within each other within subjects and then across subjects?
Thanks
Darren
I am trying to use the conn toolbox to analyze within subject resting state data. There is a single group and they have a resting state scan pre / post a treatment.
Please tell me if I am setting up the analysis correctly.
1) BASIC: number of sessions 2 (for each subject). Other entries self-explanatory
2) FUNCTIONAL: Associated resting scans with each session. (There are 196 scans, TR=2.5, duration = 490 seconds).
3) STRUCTURAL: no questions.
4) ROIs: I used all the default ROI's and selected the subject's segmented gray matter, white matter and CSF images.
Should I threshold these images? What value? Perhaps this will make each mask more exclusive?
The dimensions are set at default of 16 for white matter and csf and regress out covariates is checked.
For gray matter, dimensions is 1 and nothing is checked.
5) CONDITIONS: 1 condition called rest, and associated that with each session. Is that correct?
6) COVARIATES: First level: 1 covariate called movement. I associated it with each session and selected the relevant rp*.txt file.
7) COVARIATES: Second level: Not sure what to enter here. It only seems to accept a single number such as 1 or -1.
8) OPTIONS: I clicked all planned analyses.
9) PREPROCESSING TAB: I selected white matter, CSF and movement as confounds.
10) FIRST LEVEL ANALYSIS TAB: If I select a region such as PCC as source would my weight be "none". When would I use something like HRF or Hanning?
I'm not sure how to compare session 1 and 2 within each other within subjects and then across subjects?
Thanks
Darren
Threaded View
| Title | Author | Date |
|---|---|---|
| Darren Gitelman | Mar 7, 2012 | |
| Tamara Sussman | Feb 23, 2018 | |
| Aisling O'Neill | Feb 14, 2019 | |
| Alfonso Nieto-Castanon | Feb 14, 2019 | |
| Aisling O'Neill | Feb 14, 2019 | |
| Alfonso Nieto-Castanon | Mar 7, 2012 | |
| Darren Gitelman | Mar 7, 2012 | |
| Alfonso Nieto-Castanon | Mar 8, 2012 | |