Back to PLS Helpsingle subject single condition analysis
Posted on 12/01/08 17:37:12
Number of posts: 6
Hello,
I am a new PLS user and I am wondering if it is possible to run a ER-fMRI PLS analysis in which there is only one subject (500 scans) and one condition in the design matrix (anytime a stimulus is presented). I have tried running this analysis (I selected single subject and mean-centering PLS analysis) but I am not getting any results (grey brain).
Thank you for any assistance you can offer me,
Paul Metzak
I am a new PLS user and I am wondering if it is possible to run a ER-fMRI PLS analysis in which there is only one subject (500 scans) and one condition in the design matrix (anytime a stimulus is presented). I have tried running this analysis (I selected single subject and mean-centering PLS analysis) but I am not getting any results (grey brain).
Thank you for any assistance you can offer me,
Paul Metzak
Replies:
Untitled Post
I'm Online
Posted on 12/01/08 17:43:51
Number of posts: 291
You can use only one subject, but you will need to have at least 3 conditions (or 1 condition but have more than 3 groups) to run mean-centering PLS analysis.
Untitled Post
Posted on 12/01/08 17:44:58
Number of posts: 394
Hi Paul and welcome to PLS. The analysis you want to do is tricky since there is only one condition. Typically you would be interested in a change in signal between two or more conditions.
I think (and Jimmy please correct me if I am wrong) that if you do mean centring on one condition, that will subtract the mean from the mean and give you a value of zero at every voxel, hence a gray brain on the output.
If you simply want to look at the mean response for a given event, the easiest thing to do is run a "non-rotated" analysis with a single contrast assigning the task '1'. That should give you the mean event-related response. I don't think the permutation test will be very meaningful in this case since there are no conditions to scramble, but the bootstrap should tell you which voxels are responding most robustly.
I think (and Jimmy please correct me if I am wrong) that if you do mean centring on one condition, that will subtract the mean from the mean and give you a value of zero at every voxel, hence a gray brain on the output.
If you simply want to look at the mean response for a given event, the easiest thing to do is run a "non-rotated" analysis with a single contrast assigning the task '1'. That should give you the mean event-related response. I don't think the permutation test will be very meaningful in this case since there are no conditions to scramble, but the bootstrap should tell you which voxels are responding most robustly.
Untitled Post
Posted on 12/01/08 21:47:21
Number of posts: 6
Thank you for your help with my analysis, I really appreciate it.
I hate to trouble you further but when I was running the analysis you suggested I encountered an error message that I was unsure how to resolve. If you have any ideas on how I might resolve this, I would love to hear them.
Thanks again,
Paul Metzak
??? Error using ==> normalize at 43
A must be an ATOM.
Error in ==> fmri_taskpls_norotate at 39
crossblock = normalize(dev_design)'*dev_data;
Error in ==> fmri_pls_analysis at 576
[brainlv,s,designlv,b_scores,d_scores,lvintercorrs,design, ...
Error in ==> fmri_analysis_ui>ExecutePLS at 1847
fmri_pls_analysis(PLSoptions.profiles, PLSoptions.ContrastFile, ...
Error in ==> fmri_analysis_ui at 93
ExecutePLS;
??? Error using ==> waitfor
Error while evaluating uicontrol Callback
I hate to trouble you further but when I was running the analysis you suggested I encountered an error message that I was unsure how to resolve. If you have any ideas on how I might resolve this, I would love to hear them.
Thanks again,
Paul Metzak
??? Error using ==> normalize at 43
A must be an ATOM.
Error in ==> fmri_taskpls_norotate at 39
crossblock = normalize(dev_design)'*dev_data;
Error in ==> fmri_pls_analysis at 576
[brainlv,s,designlv,b_scores,d_scores,lvintercorrs,design, ...
Error in ==> fmri_analysis_ui>ExecutePLS at 1847
fmri_pls_analysis(PLSoptions.profiles, PLSoptions.ContrastFile, ...
Error in ==> fmri_analysis_ui at 93
ExecutePLS;
??? Error using ==> waitfor
Error while evaluating uicontrol Callback
Untitled Post
I'm Online
Posted on 12/02/08 07:19:12
Number of posts: 291
Never seem such an error message. Also tried in vain for a search on MATLAB web site.
Please send me:
Please send me:
- MATLAB version
- Datamat file (*.datamat.mat)
- Contrast file
Untitled Post
I'm Online
Posted on 12/02/08 18:11:02
Number of posts: 291
I have tried the session/datamat/contrast file you emailed me, and didn't receive any error message. (I tried both R2007a and R2008a). You can find the result file under:
ftp://ftp.rotman-baycrest.on.ca/pub/jimmy/forPaul
By the way, I also found the reason that I could not open your .mat file early this afternoon. Since we used up Linux MATLAB license, I tried it on Windows with MATLAB R2006a, that's why I cannot read your file. But it's better to use FTP server to transmit file anyway. I saw many times the file got broken because of using E-mail transmitting.
One thing strange is that when we save data under R2008a (later than your R2007b), by default, it saves in V7 format, so we can open data using any earlier version (later than MATLAB 7.0). But in your system, the files are saved into V7.3 format by default. Therefore, I am thinking that it could be the problem in your MATLAB. Are you using student version?
ftp://ftp.rotman-baycrest.on.ca/pub/jimmy/forPaul
By the way, I also found the reason that I could not open your .mat file early this afternoon. Since we used up Linux MATLAB license, I tried it on Windows with MATLAB R2006a, that's why I cannot read your file. But it's better to use FTP server to transmit file anyway. I saw many times the file got broken because of using E-mail transmitting.
One thing strange is that when we save data under R2008a (later than your R2007b), by default, it saves in V7 format, so we can open data using any earlier version (later than MATLAB 7.0). But in your system, the files are saved into V7.3 format by default. Therefore, I am thinking that it could be the problem in your MATLAB. Are you using student version?
multi-subject single condition analysis
Posted on 12/13/08 21:49:48
Number of posts: 6
Hi,
I just had a follow up question regarding the contrast to use for a similar analysis with non-rotated task PLS. If I use 24 subjects (in one group) with one condition (each), what is the correct contrast if I am interested in looking at stimuli vs. baseline? For the single subject analysis I used a single 1 as my contrast. Do I do the same here or do I need twenty-four 1's?
Thank you,
Paul
I just had a follow up question regarding the contrast to use for a similar analysis with non-rotated task PLS. If I use 24 subjects (in one group) with one condition (each), what is the correct contrast if I am interested in looking at stimuli vs. baseline? For the single subject analysis I used a single 1 as my contrast. Do I do the same here or do I need twenty-four 1's?
Thank you,
Paul
Untitled Post
Posted on 12/15/08 08:17:50
Number of posts: 394
quote:
Hi,
I just had a follow up question regarding the contrast to use for a similar analysis with non-rotated task PLS. If I use 24 subjects (in one group) with one condition (each), what is the correct contrast if I am interested in looking at stimuli vs. baseline? For the single subject analysis I used a single 1 as my contrast. Do I do the same here or do I need twenty-four 1's?
Thank you,
Paul
Hi Paul - for the full group analysis you use a single '1' as the contrast since the data will be averaged before computing the cross product of brain and contrast.
Hi,
I just had a follow up question regarding the contrast to use for a similar analysis with non-rotated task PLS. If I use 24 subjects (in one group) with one condition (each), what is the correct contrast if I am interested in looking at stimuli vs. baseline? For the single subject analysis I used a single 1 as my contrast. Do I do the same here or do I need twenty-four 1's?
Thank you,
Paul
Login to reply to this topic.
