Dear FSL experts,
I'm running a TBSS analysis on a group of subjects and I'm interested whether a behavioral continuous variable is related to FA changes taking into account several covariates (both binary and continuous).
I just want to make sure I'm doing everything correctly.
1) When I create my design.mat I orthogonalise (demean) just my covariates but not the variable of interest?
2) My contract of interest would be 1 0 0 (where one corresponds to my behavioral variable and the two 0 to covariates)?
3) What is the appropriate level for cluster-based thresholding (-c) when subsequently running randomise? 2? 3? I have tried TFCE-thresholding and I'm not getting any significant results. Since this study is exploratory I want to make sure I'm not missing any possible trends in the data.
Many thanks in advance for your comments.
Maria.
|