NLOPTIONS Grid Search

[Cuneyt]

SAS provides a number of techniques (NEWRAP, QUANEW, etc.) under nonlinear optimization methods. One of these techniques is NONE where, according to documentation, SAS conducts a grid search. I have never used grid search before. Do I provide SAS with initial values, upper bounds, lower bounds? What is the proper way to set up a grid search? Any guidance will be much appreciated.

[SteveDenham]

Let's break this down into parts. The first is simple enough - an NLOPTIONS statement with the TECH=NONE option. The next part is determining which effects will go into the grid.  For that, we need to have the code you are currently using, to identify which variables are going to be included in the grid search. This will likely differ based on the statistical PROC you want to use.

In most cases, this is going to involve the PARMS statement. I would recommend heading to the documentation for the PROC you want to use in your analysis and read through the PARMS statement for that PROC.  I have used this several times with PROC GLIMMIX, so I find that my go-to.

SteveDenham

[Cuneyt]

Hi Steve,

Thanks for your response. I was looking into PROC NLMIXED. I have read further about the PARMS options. Now, I have a good idea about how proceed. Thank you very much for your guidance.

Cuneyt

[Rick_SAS]

This article discusses the PARMS statement and gives examples of how to use it in PROC NLMIXED:

"Use a grid search to find initial parameter values for regression models in SAS"