Dear FreeSurfer experts,

 

I am trying to use the spatiotemporal mass-univariate LME approach for my data but am struggling with choosing the right contrast.

I have one group with 14 subjects, two timepoints for each subject and age. My goal is to test whether the subjects experienced any kind of change in cortical thickness from pre- to post-measurement, controlling for age.

I followed the tutorial up to the point where I figured out that only one random effect is needed as most vertices did not survive multiple comparisons correction. If I am correct the next step is to test, whether a quadratic or a linear model is needed. Therefore, I tested the contrast CM1.C = [0 1 0 0] on my design matrix X with subjects, time in years, (time in years)², age:

1.0000         0         0   20.0000

1.0000    1.4800    2.1904   20.0000

1.0000         0         0   19.0000

1.0000    1.4800    2.1904   19.0000

…

 

Multiple comparisons correction revealed an empty vector (as in the tutorial), so that the quadratic term of time was removed and a linear model with one random effect is tested with the new design matrix

1.0000         0   20.0000

1.0000    1.4800   20.0000

1.0000         0   19.0000

1.0000    1.4800   19.0000

…

 

Next, I applied the contrast CM2.C=[0 1 0] to test for change over time.

My question is if this is actually the right way to achieve my goal or whether I am doing something completely wrong.

And if it is correct, I would go on by correcting for multiple comparisons, right? How would I visualize the results & additionally get information about significant clusters? Unfortunately, it seems like I am not understanding the last part of the tutorial.  

Any help would be much appreciated!

 

Best,

Franziska

 

_____________________________

Franziska K. Haase, M.A.

Sportmedizinisches Institut

Department Sport und Gesundheit

Fakultät für Naturwissenschaften

Universität Paderborn

 

Warburger Str. 100

33098 Paderborn

Golfakademie 20b