The NLP Procedure

Functional Summary

The following table outlines the options in PROC NLP classified by function. An alphabetical list of options is provided in the Dictionary of Options .

Table 7.1: Functional Summary

Description

Statement

Option

Input Data Set Options:

Input data set

PROC NLP

DATA=

Initial values and constraints

PROC NLP

INEST=

Quadratic objective function

PROC NLP

INQUAD=

Program statements

PROC NLP

MODEL=

Skip missing value observations

PROC NLP

NOMISS

Output Data Set Options:

Variables and derivatives

PROC NLP

OUT=

Result parameter values

PROC NLP

OUTEST=

Program statements

PROC NLP

OUTMODEL=

Combine various OUT…statements

PROC NLP

OUTALL

CRP Jacobian in the OUTEST= data set

PROC NLP

OUTCRPJAC

Derivatives in the OUT= data set

PROC NLP

OUTDER=

Grid in the OUTEST= data set

PROC NLP

OUTGRID

Hessian in the OUTEST= data set

PROC NLP

OUTHESSIAN

Iterative output in the OUTEST= data set

PROC NLP

OUTITER

Jacobian in the OUTEST= data set

PROC NLP

OUTJAC

NLC Jacobian in the OUTEST= data set

PROC NLP

OUTNLCJAC

Time in the OUTEST= data set

PROC NLP

OUTTIME

Optimization Options:

Minimization method

PROC NLP

TECH=

Update technique

PROC NLP

UPDATE=

Version of optimization technique

PROC NLP

VERSION=

Line-search method

PROC NLP

LINESEARCH=

Line-search precision

PROC NLP

LSPRECISION=

Type of Hessian scaling

PROC NLP

HESCAL=

Start for approximated Hessian

PROC NLP

INHESSIAN=

Iteration number for update restart

PROC NLP

RESTART=

Initial Value Options:

Produce best grid points

PROC NLP

BEST=

Infeasible points in grid search

PROC NLP

INFEASIBLE

Pseudorandom initial values

PROC NLP

RANDOM=

Constant initial values

PROC NLP

INITIAL=

Derivative Options:

Finite-difference derivatives

PROC NLP

FD=

Finite-difference derivatives

PROC NLP

FDHESSIAN=

Compute finite-difference interval

PROC NLP

FDINT=

Use only diagonal of Hessian

PROC NLP

DIAHES

Test gradient specification

PROC NLP

GRADCHECK=

Constraint Options:

Range for active constraints

PROC NLP

LCEPSILON=

LM tolerance for deactivating

PROC NLP

LCDEACT=

Tolerance for dependent constraints

PROC NLP

LCSINGULAR=

Sum all observations for continuous functions

NLINCON

/ SUMOBS

Evaluate each observation for continuous functions

NLINCON

/ EVERYOBS

Termination Criteria Options:

Maximum number of function calls

PROC NLP

MAXFUNC=

Maximum number of iterations

PROC NLP

MAXITER=

Minimum number of iterations

PROC NLP

MINITER=

Upper limit on real time

PROC NLP

MAXTIME=

Absolute function convergence criterion

PROC NLP

ABSCONV=

Absolute function convergence criterion

PROC NLP

ABSFCONV=

Absolute gradient convergence criterion

PROC NLP

ABSGCONV=

Absolute parameter convergence criterion

PROC NLP

ABSXCONV=

Relative function convergence criterion

PROC NLP

FCONV=

Relative function convergence criterion

PROC NLP

FCONV2=

Relative gradient convergence criterion

PROC NLP

GCONV=

Relative gradient convergence criterion

PROC NLP

GCONV2=

Relative parameter convergence criterion

PROC NLP

XCONV=

Used in FCONV, GCONV criterion

PROC NLP

FSIZE=

Used in XCONV criterion

PROC NLP

XSIZE=

Covariance Matrix Options:

Type of covariance matrix

PROC NLP

COV=

$\sigma ^2$ factor of COV matrix

PROC NLP

SIGSQ=

Determine factor of COV matrix

PROC NLP

VARDEF=

Absolute singularity for inertia

PROC NLP

ASINGULAR=

Relative M singularity for inertia

PROC NLP

MSINGULAR=

Relative V singularity for inertia

PROC NLP

VSINGULAR=

Threshold for Moore-Penrose inverse

PROC NLP

G4=

Tolerance for singular COV matrix

PROC NLP

COVSING=

Profile confidence limits

PROC NLP

CLPARM=

Printed Output Options:

Display (almost) all printed output

PROC NLP

PALL

Suppress all printed output

PROC NLP

NOPRINT

Reduce some default output

PROC NLP

PSHORT

Reduce most default output

PROC NLP

PSUMMARY

Display initial values and gradients

PROC NLP

PINIT

Display optimization history

PROC NLP

PHISTORY

Display Jacobian matrix

PROC NLP

PJACOBI

Display crossproduct Jacobian matrix

PROC NLP

PCRPJAC

Display Hessian matrix

PROC NLP

PHESSIAN

Display Jacobian of nonlinear constraints

PROC NLP

PNLCJAC

Display values of grid points

PROC NLP

PGRID

Display values of functions in LSQ, MIN, MAX

PROC NLP

PFUNCTION

Display approximate standard errors

PROC NLP

PSTDERR

Display covariance matrix

PROC NLP

PCOV

Display eigenvalues for covariance matrix

PROC NLP

PEIGVAL

Print code evaluation problems

PROC NLP

PERROR

Print measures of real time

PROC NLP

PTIME

Display model program, variables

PROC NLP

LIST

Display compiled model program

PROC NLP

LISTCODE

Step Length Options:

Damped steps in line search

PROC NLP

DAMPSTEP=

Maximum trust region radius

PROC NLP

MAXSTEP=

Initial trust region radius

PROC NLP

INSTEP=

Profile Point and Confidence Interval Options:

Factor relating discrepancy function to $\chi ^2$ quantile

PROFILE

FFACTOR=

Scale for y values written to OUTEST= data set

PROFILE

FORCHI=

Upper bound for confidence limit search

PROFILE

FEASRATIO=

Write all confidence limit parameter estimates to OUTEST= data set

PROFILE

OUTTABLE

Miscellaneous Options:

Number of accurate digits in objective function

PROC NLP

FDIGITS=

Number of accurate digits in nonlinear constraints

PROC NLP

CDIGITS=

General singularity criterion

PROC NLP

SINGULAR=

Do not compute inertia of matrices

PROC NLP

NOEIGNUM

Check optimality in neighborhood

PROC NLP

OPTCHECK=