> I would set go for (1) 3-way interaction in the end (b/w-ss group x TENT x covariate) or
> go for (2) b/w-ss group x covariate interaction for each TENT basis? I presume to do
> the option (2), I would have to set the glf for each TENT
Go with 1), and interpret the results as follows: interaction group x TENT x covariate shows the interaction of group x covariate across the *whole* hemodynamic response. Similarly, group x TENT reveals the group effect in terms of the overall HDR.
Gang