> In this situation, percent signal change is not calculated before deconvolution,
> and group analysis proceeds simply with the t statistics (converted to z) attached
> to the respective beta. In that case, would you still consider the z score to be a
> scaled percent signal change?
In that situation, I would still try to convert the betas into percent signal change. It would be not appropriate to run group analysis directly on the beta values due to the baseline variation across subjects. If I were only allowed to choose between unscaled betas and z-score for group analysis, I would pick z-score. Yes, I still consider the z-score scaled percent signal change because the t value should be the same with or without scaling on the original signal.
Gang