As with any nonlinear optimization routine, there is no guarantee that a nonlinear model can converge for a given data set. See Troubleshooting in the Details section in the PROC NLIN documentation which describes problems that can occur during the fitting process.

The following suggestions might be helpful when a model fails to converge:

• Examine the iteration history to verify that the algorithm is making progress toward a solution. If so, consider increasing the maximum allowed number of iterations in the MAXITER= option.
• Specify a different optimization method in the METHOD= option.
• Specify different starting values in the PARAMETERS (or PARMS) statement.
• Rather than specifying the derivatives, allow the NLIN procedure to compute the first and second derivatives automatically by removing any DER statements.
• In some cases, it might be necessary to reconsider the form of the nonlinear model.