Bug: A negative standard error in nonlinear regression results

When Prism fits a curve with nonlinear regression, it reports the best-fit value of the parameters along with the corresponding standard error and confidence intervals. 

Standard errors (SE) are, by definition, always reported as positive numbers. But in one rare case, Prism will report a negative SE. This happens when you ask Prism to report P1^P2 where P1 and P2 are parameters and P1 < 1 and P2 > 0. This can happen when fitting the built-in Allosteric Sigmoidal equation in the enzyme kinetics family of equations, or any equation you enter where you ask to report P1^P2.

The bug has no impact on understanding the fit. The true SE is simply the absolute value of the reported one. The confidence interval, computed from the standard errors is correct. 

To be fixed in 6.06 and 6.0g.