In IML: Error message is "unresolved module NLPSOLVE"

blund — Mon, 14 Oct 2024 17:02:48 GMT

The following code appeared in a 2024 WUSS presentation (below).

When running SAS OnDemand for Academics I get this IML error

ERROR: Invocation of unresolved module NLPSOLVE

Can someone provide a replacement IML function?

PROC IML;

start LL(param) global(x); /* normal log likelihood */

mu = param[1]; sigma = param[2];

LL = sum( logpdf("Normal", x, mu, sigma) );

return( -LL ); /* minimize -LL */

finish;

use Sashelp.Iris; /* read sample data */

read all var {SepalWidth} into x;

p = {35 5.5}; /* guess (mu, sigma), sigma>0 */

call nlpsolve(rc, soln, "LL", p) L={. 0}; /* sigma>0 */

print soln;

Re: In IML: Error message is "unresolved module NLPSOLVE"

Rick_SAS — Mon, 14 Oct 2024 17:15:16 GMT

Sure. ODA is running SAS 9.4, so use the NLPNRA procedure, as shown in

Maximum likelihood estimation in SAS/IML - The DO Loop

PROC IML;
start LL(param) global(x); /* normal log likelihood */
   mu = param[1]; sigma = param[2];
   LL = sum( logpdf("Normal", x, mu, sigma) );
   return( -LL ); /* minimize -LL */
finish;

use Sashelp.Iris; /* read sample data */
read all var {SepalWidth} into x;

p = {35 5.5}; /* guess (mu, sigma), sigma>0 */
/*     mu-sigma constraint matrix */
con = { .   0,  /* lower bounds: -infty < mu; 0 < sigma */
        .   .}; /* upper bounds:  mu < infty; sigma < infty */
call nlpnra(rc, soln, "LL", p) BLC=con;
print soln;

To understand where the SUM(LOGPDF(...)) function call comes from, see

Two simple ways to construct a log-likelihood function in SAS - The DO Loop

topic In IML: Error message is "unresolved module NLPSOLVE" in SAS/IML Software and Matrix Computations

In IML: Error message is "unresolved module NLPSOLVE"

Re: In IML: Error message is "unresolved module NLPSOLVE"