"Hit constraint"

This checklist is for when the nonlinear regression was not able to finish, but rather reported "Hit constraint" at the top of the results. View other checklists if your curve fit ended with a different completion code: Completed, Bad initial values, Interrupted, Not converged, Ambiguous, Don't fit, Too few points, Perfect Fit or Impossible weights.

Constraints

Prism reports "Hit constraint" when the best-fit value of one or more parameters is right at the limit of a constraint.

The Constrain tab lets you tell Prism that a parameter can have values only within a certain range of values. Constraints are most helpful when they guide the iterations (prevent impossible values), but the best-fit value ends up within the allowed zone.

In some cases, the curve fitting process hits a constraint but isn't able to 'bounce back'. In that case, the best-fit value is right at the constraint border. For example, if you mistakenly set a constraint that a rate constant K must be less than or equal to zero, Prism's nonlinear regression might converge with the best-fit value equalling zero and report that you hit the constraint.

Note that hitting a constraint is not the same as constraining a parameter to a constant value. The best-fit values of all the parameters will be the same, but the standard errors, confidence intervals, covariance matrix, and dependencies will be different. Prism still treats the parameter that hit the constraint to be a parameter when it calculates these values (but doesn't when a parameter is constrained to a constant value).

When a fit hits a constraint, the results are unlikely to provide useful information. If you had a solid reason to constrain a parameter within a range of values, it ought to end up in that range, and not right at one of the limits. This usually happens when you set the constraint incorrectly.

Checklist

Did you enter the constraint correctly? Did you mix up "<" and ">".

Did you enter the model correctly?

Are you asking Prism to do the impossible -- too many parameters for your data?

If your model has two phases or components, consider switching to a one phase or component model.