Chapter Contents
Chapter Contents
The NLP Procedure

Functional Summary

The following table outlines the options in the NLP statement classified by function.

Description Option
Input Data Set Specifications 
data setDATA=
initial values and constraintsINEST=
quadratic objective functionINQUAD=
program statementsMODEL=
skip missing value observationsNOMISS
Output Data Set Specifications 
variables and derivativesOUT=
result parameter valuesOUTEST=
program statementsOUTMODEL=
combining various OUT... statementsOUTALL
CRP Jacobian to OUTEST= data setOUTCRPJAC
derivatives in the OUT= data setOUTDER
grid in the OUTEST= data setOUTGRID
Hessian to OUTEST= data setOUTHESSIAN
iterative output to the OUTEST= data setOUTITER
Jacobian in the OUTEST= data setOUTJAC
NLC Jacobian in the OUTEST= data setOUTNLCJAC
time in the OUTEST= data setOUTTIME
Optimization Specifications 
minimization methodTECHNIQUE=
update techniqueUPDATE=
version of optimization techniqueVERSION=
line-search methodLINESEARCH=
line-search precisionLSPRECISION=
type of Hessian scalingHESCAL=
start for approximated HessianINHESSIAN=
iteration number for update restartRESTART=
Initial Value Specifications 
best grid point numberBEST=
infeasible points in grid searchINFEASIBLE=
pseudorandom initial valuesRANDOM=
constant initial valuesINITIAL=
Derivatives Specifications 
finite-difference derivativesFD[=]
finite-difference derivativesFDHESSIAN[=]
compute finite-difference intervalFDINT=
use only diagonal of HessianDIAHES
test gradient specificationGRADCHECK[=]
Constraint Specifications 
range for active constraintsLCEPSILON=
LM tolerance for deactivatingLCDEACT=
tolerance for dependent constraintsLCSINGULAR=
Termination Criteria Specifications 
maximum number of function callsMAXFUNC=
maximum number of iterationsMAXITER=
minimum number of iterationsMINITER=
upper limit seconds of CPU timeMAXTIME=
absolute function convergence criterionABSCONV=
absolute function convergence criterionABSFCONV =
absolute gradient convergence criterionABSGCONV=
absolute parameter convergence criterionABSXCONV=
relative function convergence criterionFCONV =
relative function convergence criterionFCONV 2=
relative gradient convergence criterionGCONV=
relative gradient convergence criterionGCONV2=
relative parameter convergence criterionXCONV=
used in FCONV , GCONV criterionFSIZE=
used in XCONV criterionXSIZE=
Covariance Matrix Specifications 
kind of covariance matrixCOVARIANCE=
\sigma^2 factor of COV matrixSIGSQ=
determines factor of COV matrixVARDEF=
absolute singularity for inertiaASINGULAR=
relative M singularity for inertiaMSINGULAR=
relative V singularity for inertiaVSINGULAR=
threshold for Moore-Penrose inverseG4
tolerance for singular COV matrixCOVSING=
profile confidence limitsCLPARM=
Printed Output Specifications 
print (almost) all printed outputPALL
suppresses all printed outputNOPRINT
reduces some default outputPSHORT
reduces most default outputPSUMMARY
initial valuesPINIT
optimization historyPHISTORY
Jacobian matrixPJACOBI
crossproduct Jacobian matrixPCRPJAC
Hessian matrixPHESSIAN
Jacobian of nonlinear constraintsPNLCJAC
values of grid pointsPGRID
values of functions in LSQ, MIN, MAXPFUNCTION
approximate standard errorsPSTDERR
covariance matrixPCOV
eigenvalues for covariance matrixPEIGVAL
prints code evaluation problemsPERROR
model program, variablesLIST
compiled model programLISTCODE
Step Length Specifications 
damped steps in line-searchDAMPSTEP[=]
maximum trust-region radiusMAXSTEP=
initial trust-region radiusINSTEP=
Miscellaneous Options 
number accurate digits in objective functionFDIGITS=
number accurate digits in nonlinear constraintsCDIGITS=
general singularity criterionSINGULAR=
do not compute inertia of matricesNOEIGNUM
check optimality in neighborhoodOPTCHECK[=]

Chapter Contents
Chapter Contents

Copyright © 1999 by SAS Institute Inc., Cary, NC, USA. All rights reserved.