Bug: Confidence interval is wrong when the Remove Baseline analysis computes the percent difference.

When the Remove Baseline analysis is used to compute the percent difference i.e. 100 * (Y - X)/X, it computes the SE of that result (and thus the CI) improperly. The correct expression for the SEM of the (Y - X)/X is:

SEM((Y - X)/X) = (Y/X) * [SQRT(SEM(X)^2 / X^2] + [SEM(Y)^2 / Y^2)]

The bug is in Prism 6, and fixed in 7.00 and 7.0a.