Starting points for a nonlinear model.

[wutao9999]

I am wondering if SAS can allow the following way to set starting points for a nonlinear model:

1)  ignore all nonlinear terms in the model, and first solve the problem using linear programming solver.

2) using the solution of the linear programming solver as the starting point for the nonlinear programming solver.

I have no ideas for the above 2), does anyone know the answer?  Thank you very much.

[RobPratt · Accepted solution]

There is nothing automated to do this.  But in PROC OPTMODEL, you can call the LP solver by using SOLVE WITH LP and then later call the NLP solver by using SOLVE WITH NLP.  By default, the second solve will use the solution from the first solve as a starting point.  You can disable this behavior by using the NOINITVAR option in the PROC OPTMODEL statement:

SAS/OR(R) 13.1 User's Guide: Mathematical Programming.

You might find it helpful to use the PROBLEM and USE PROBLEM statements to control which parts of the model are passed to the solver:

SAS/OR(R) 13.1 User's Guide: Mathematical Programming