Why results of nonlinear regression in Prism 7 can be different than prior versions

Nonlinear regression is an iterative procedure that requires numerous calculations of the derivative, and second derivative, of the change in Y with respect to each parameter. Prism 6 and earlier did this using numerical derivatives. For almost all the built-in equations, Prism uses analytical derivatives. This has two advantages:

• Fitting is much faster, which is noticeable when you have lots of data or a fitting lots of curves at once. 
• The results are more accurate. Usually the difference is trivial. But the differences can be more substantial for equations which are highly nonlinear with respect to the parameters (highly skewed), or when the parameters are poorly defined by data (high dependency).

Prism 7 uses the analytical derivatives for almost all the built-in equations, and for any user-defined equation that Prism recognizes as equivalent to a built-in equation. 

The results page does not indicate whether the derivatives were computed analytically or numerically. 

You can get Prism to compute built-in equations using numerical derivatives with this trick: Clone the equation, and then modify it in trivial ways (multiply everything by 1.0, add 0.0).