Hi Tudor,
On 03/22/2013 05:19 PM, Tudor Popescu wrote:
Hello everyone,
I created a qdec.table.dat for my design (a simple group analysis, with gender as categorical factor and age as continuous factor), and created a gender.levels file containing the values that are found in qdec.table.dat, spellt in exactly the same way (all files attached). However, when loading the file in QDEC, I get the error:
/ERROR: Subject C01 has an invalid level 'controls' in the group column INFO: If 'group' is a discrete factor, then create a file named 'group.levels' containing the valid factor names one per line. Error loading the data table. /
The strings under the fsid column (subject ID) correspond to those that I used after the –subjid xxx, in my previous call to recon-all (e.g. "C01, C02, M01, M02.."). The qdec.table.dat was created in Excel and saved as a tab-delimited txt file. Replacing the tabs with spaces/commas does not change the error message. What could be the problem?
It looks like you made the group.levels file on a windows system. There are hidden characters that cause qdec to fail. Try creating it from scratch on a linux system or run dos2unix on the file.
I also have a few more questions, if I may:
- how much more accurate are the results of a QDEC analysis if an
average subject is computed from the subjects of the sample, with make_average_subject, as opposed to using FS's default average subject, fsaverage?
Accurate in what way? Changing the average subject in qdec just changes the gray scale curvature pattern. This will be a little more accurate if you use your subjects, but probably not much. It will change the stats a very small amount. Note that QDEC will not use a different registration even if you spec a new average subject.
- the qdec.table.dat in the tutorial
http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/QdecGroupAnalysiscontains more than just "demographical information", for instance it contains the volumes of the left and right hippocampi; are these types of informations normally brought in from previous analyses (that may have been done in a package other than FS), or does one first have to obtain any such relevant predictors in FreeSurfer, before starting the QDEC analysis?
They do not have to be. They are only there for convenience in case you have a hypothesis about them.
- the QDEC tutorial seems to be for whole-brain analyses; the ROI
analysis tutorial http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/AnatomicalROI mentions nothing of QDEC, however; can ROI analyses not also be done in QDEC?
No, not yet.
- in the Group Analysis (non-QDEC) tutorial
http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/GroupAnalysis, the design matrix (Xg.dat) is in the form MembershipF| MembershipM|AgeF|AgeM, while the values in the contrast are said to correspond to the 4 parameters (intercept for F,M; slope for F,M). Either way, how can the contrast vector be interpreted to give the correlation between thickness and age?
Each column in the Xg.dat is matched with a linear parameter (intercept or slope). If you have [0 0 1 0], it means to ignore the parameters matched with 0s and just test the parameter matched with 1, in this case, the AgeF (and so slope) parameter.
- in the same tutorial, why is the value 0 being given to the first
two elements in the matrix, when it should only be given to nuisance variables?
Effectively yes. They might only be "nuisance" for that particular contrast. For other contrasts they might be variables of interest. doug
Many thanks in advance for any help! Tudor
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer