> Sounds like I should generate linear and quadratic betas for each subject and
> then conduct group analyses for each beta weight separately.
That should be fine. However, you don't have to do that. You can run such trend analysis at both individual and group level with the same weighting scheme.
Gang