Dear FSL Experts,
I have some questions regarding the interpretation of contrasts and results in a TBSS study.
To compare the correlation of an external variable with the FA values between two clinical groups, I split the EV in 2 variables and set up the contrasts as described in Figure 8.10 in Smith, S. M., & Kindlmann, G. (2009). Cross-subject Comparison of local Diffusion MRI Parameters. In H. Johansen-Berg, & T. E. J. Behrens (Eds.), Diffusion MRI (1st ed., pp. 147-174). Amsterdam: Elsevier Acad. Press.
Design-matrix:
1.000000e+00 0.000000e+00 -2.110000e+00 0.000000e+00
1.000000e+00 0.000000e+00 -2.110000e+00 0.000000e+00
1.000000e+00 0.000000e+00 -2.110000e+00 0.000000e+00
1.000000e+00 0.000000e+00 2.890000e+00 0.000000e+00
1.000000e+00 0.000000e+00 -2.110000e+00 0.000000e+00
1.000000e+00 0.000000e+00 4.890000e+00 0.000000e+00
1.000000e+00 0.000000e+00 -1.110000e-01 0.000000e+00
1.000000e+00 0.000000e+00 -2.110000e+00 0.000000e+00
1.000000e+00 0.000000e+00 2.890000e+00 0.000000e+00
0.000000e+00 1.000000e+00 0.000000e+00 1.570000e+00
0.000000e+00 1.000000e+00 0.000000e+00 5.570000e+00
0.000000e+00 1.000000e+00 0.000000e+00 -1.430000e+00
0.000000e+00 1.000000e+00 0.000000e+00 -3.430000e+00
0.000000e+00 1.000000e+00 0.000000e+00 -3.430000e+00
0.000000e+00 1.000000e+00 0.000000e+00 -4.290000e-01
0.000000e+00 1.000000e+00 0.000000e+00 1.570000e+00
Contrasts:
0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00
0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00
0.000000e+00 0.000000e+00 1.000000e+00 -1.000000e+00
0.000000e+00 0.000000e+00 -1.000000e+00 1.000000e+00
It results in a significant cluster for contrast 4 (i.e. the comparison between the two parts of the EV, but no significant cluster for a correlation itself (i.e. contrasts 1 or 2). It is possible to interpret the significant cluster of the comparison contrast as a difference of the correlations, even if the correlation for each group is not significant?
Additional, i would like to add age as a covariate (as an additional EV with the contrast value 0). Do I have to split this variable as well (because there are two groups of subjects), even if i don't wont to test differences regarding this EV?
Thanks a lot for your kind help!
Antonius
|