> 1) Is there a reason why Chen's suggestion would be more appropriate than Glen's? I found Glen’s
> more intuitive, but maybe it’s because it doesn’t take into account the centering?
3dTcorr1D is usually used to calculate the correlation of brain time series and a one-dimensional time series. More importantly, it does not provide any statistic testing for the correlation. Centering is only relevant if you're interested in the intercept.
> 2) Also, could you explain the output's colors? If I understand well, the correlation between the score
> and the maps would be represented by the marginal effect of the score (and its associated t-stat). But
> I am not sure of the link between marginal effect and correlation. Marginal effect would be how the score
> is a good predictor of the seed-based correlation score?
>
> So blue would be a bad predictor ==> no correlation
> Red would be a good predictor ==> strong correlation?
I cannot see your color bar, but if blue and red correspend to small and large values, respectively, then yes your understanding is correct. More specifically, the marginal effect indicates the amount of change in z-score when the explanatory variable increases by one unit.
Gang