Hi Andrew,

This may or may not be right: it depends on whether the inputs are subtracted images (timepoint2 - timepoint1) or the actual cortical thickness maps without any subtraction. Depending on which case you'd need to add one intercept per subject, or you may need to subtract the IQ measurements and make other changes.

Also, note that the values of EV5 and EV6 are in a range very different than the others. This makes the numerical operations unstable. Consider dividing these EVs by some constant, e.g., 1e6, such that values become better behaved.

The above have to do with the design, and would apply to whatever quantities you might have had. Now, regarding thickness and IQ, it turns out that IQ is more related to area, not thickness, and volume is also more related to area, not thickness. So, with this design, even in the event that there is an actual effect on thickness, it may end up being absorbed by the volume variable, leaving you with negative results...

Hope this helps!

All the best,

Anderson

On Wed, 3 Oct 2018 at 15:18, Andrew Gordon <[log in to unmask]> wrote:
Dear FSL experts

I would like to use PALM on some cortical thickness measures from freesurfer and was hoping that someone could look over my design matrix/command and let me know if it is appropriate?

I have two groups of participants (EV1, EV2) who have been scanned at 2 time-points (T1, T2). I am looking to use PALM to compute the main effect of change in IQ (EVs 7+8 below) on cortical thickness (.mgh) while controlling for age (EVs 3+4), and total cortical volume (EVs 5+6). I would also like to determine whether this relationship differs between the two groups as well as determining if thickness differs while controlling for everything else. I followed the basic outline of the two Groups with continuous covariate interaction in the GLM wiki and just wanted to make sure I have set up the following correctly:

Design Matrix:
EV1     EV2     EV3     EV4        EV5        EV6              EV7              EV8
1       0       28.2    0          981438.3     0              39               0
1       0       38.5    0          1001527      0              40.05556 0
1       0       35.6    0          1015722      0              48.80645 0
1       0       41.7    0          1112665      0              33.625   0
1       0       39.8    0          1087298      0              20.51351 0
1       0       44      0          1148600      0              16.625   0
0       1       0       34.7       0                    1073384 0               33.42424
0       1       0       44         0                    1163399 0               39.20513
0       1       0       31.18   0                       1019120 0               48.34615
0       1       0       29.54   0                       956024  0               35.57692
0       1       0       28.19   0                       1049779 0               73.89655
0       1       0       30.49   0                       941737.5        0               36

Contrasts:                                                     

grp1>grp2                       1     -1        0       0       0       0       0       0
grp2>grp1                      -1       1       0       0       0       0       0       0
Pos. slope grp1         0       0       0       0       0       0       1       0
Neg. slope grp2         0       0       0       0       0       0     -1        0
Pos. slope grp1         0       0       0       0       0       0       0       1
Neg. slope grp2         0       0       0       0       0       0       0     -1
grp1>grp2 slope 0       0       0       0       0       0       1     -1
grp2>grp1 slope 0       0       0       0       0       0     -1        1


Call to PALM:
palm –i  lh.thickness.fsaverage.mgh  -d matrix.csv  -t contrasts.csv  -o LH_thickness  -fdr –n 10000 -corrcon -nouncorrected

Is this correct? Being that there are two groups should I also use the –vg option to define two variance groups, and does this also require a –eb input?

I would also like to run the analysis with lh and rh combined. I see that the command palm_hemimerge can be used for this, is it compatible with .mgh files?

Any help would be appreciated thanks!

########################################################################

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1