Dear FSL’ers,

I want to analyse my DWI data with TBSS with a model, which uses a group variable (3 levels, g1,g2,g3) and a behavioral continuous parameter (BehavP) as predictors and two control variables (sex, age). I am not sure about the exact design to answer the question, whether there are group differences and whether they are related to the behavioral parameter, controlling for sex and age.

Would the following model and the (interpretation of) contrasts be correct:

EV1: intercept

EV2: g1

EV3: g2

EV4: g3

EV5: g1xBehavP (BehavP demeaned across groups controlling for Sex and Age)

EV6: g2xBehavP (BehavP demeaned across groups)

EV7: g3xBehavP (BehavP demeaned across groups)

EV8: Sex (demeaned across groups)

EV9: Age (demeaned across groups)

Contrasts:

C1: Main effect of group

1 0 0

0 1 0

0 0 1

C2: g1 > g2 (“Group1 has higher FA than Group2 controlling for Sex and Age”)

1 -1 0

C3: g1 > g3 (“Group1 has higher FA than Group3 controlling for Sex and Age”)

1 0 -1

C4: g2 > g1 (“Group2 has higher FA than Group1 controlling for Sex and Age”)

-1 1 0

C5: g2 > g3 (“Group2 has higher FA than Group3 controlling for Sex and Age”)

0 1 -1

C6: g3 > g1 (“Group3 has higher FA than Group1 controlling for Sex and Age”)

-1 0 1

C7: g3 > g2 (“Group3 has higher FA than Group2 controlling for Sex and Age”)

0 -1 1

C8: Main interaction effect

0 0 0 1 0 0

0 0 0 0 1 0

0 0 0 0 0 1

C9: g1xBehavP > g2xBehavP (“Group1 with higher BehavP values than Group2 have increased FA in this voxel controlling for Sex and Age”)

0 0 0 1 -1 0

C10: g1xBehavP > g3xBehavP (“Group1 with higher BehavP values than Group3 have increased FA in this voxel controlling for Sex and Age”)

0 0 0 1 0 -1

C11: g2xBehavP > g1xBehavP (“Group2 with higher BehavP values than Group1 have increased FA in this voxel controlling for Sex and Age”)

0 0 0 -1 1 0

C12: g2xBehavP > g3xBehavP (“Group2 with higher BehavP values than Group3 have increased FA in this voxel controlling for Sex and Age”)

0 0 0 0 1 -1

C13: g3xBehavP > g1xBehavP (“Group3 with higher BehavP values than Group1 have increased FA in this voxel controlling for Sex and Age”)

0 0 0 -1 0 1

C14: g3xBehavP > g2xBehavP (“Group3 with higher BehavP values than Group2 have increased FA in this voxel controlling for Sex and Age”)

0 0 0 0 -1 1

Thank you in advance for your help!

Bests,

Philipp