Otimization function

[pedroy341]

Hello. I am trying to approximate a 36 points series to a quaterly data of 9 quarters. The objetive here is that the optimized series of 36 points fits the goal for those 9 quarters with minimal error while  being a smooth and monotonous function. I was trying to use the following code:

proc iml;

   /* Step 1: Define Bayes values from macro variables */

   bayes = {

       &bayes1,

       &bayes2,

       &bayes3,

       &bayes4,

       &bayes5,

       &bayes6,

       &bayes7,

       &bayes8,

       &bayes9

   };

   /* Step 2: Define initial guess (e.g., from macro variables &pd1–&pd36) */

    PD_ini = {

        &PD1, &PD2, &PD3, &PD4, &PD5, &PD6, &PD7, &PD8, &PD9, &PD10, &PD11, &PD12, &PD13, &PD14, &PD15, &PD16, &PD17, &PD18, &PD19, &PD20, &PD21, &PD22, &PD23, &PD24, &PD25, &PD26, &PD27, &PD28, &PD29, &PD30, &PD31, &PD32, &PD33, &PD34, &PD35, &PD36

    };

PD_init=t(PD_ini);

   /* Step 3: Define the objective function */

   start objetive(pd) global(bayes);

       S = j(1, 36, .);

       CumPD = j(1, 36, .);

       S[1] = 1 - pd[1];

       CumPD[1] = 1 - S[1];

       do t = 2 to 36;

           S[t] = S[t-1] * (1 - pd[t]);

           CumPD[t] = 1 - S[t];

       end;

       obj = 0;

       do i = 0 to 8;

           t = i * 3 + 1;  /* Corresponds to months 0, 3, ..., 24 */

           condPD = (CumPD[t+12] - CumPD[t]) / S[t];

           err = (condPD - bayes[i+1]) * 100;

           obj = obj + err##2;

       end;

       return (obj);

   finish;

   /* Step 4: Call optimizer */

   optn = {0};  /* 0 = minimize */

   call nlpqn(rc, result, "objetive", initPD, optn);

   /* Step 5: Output results */

   print rc result;

quit;

But I keep getting this error:
74 ! /* 0 = minimize */
75 call nlpqn(rc, result, "objetive", initPD, optn);
ERROR: (execution) Matrix has not been set to a value.

I have also tried to use the initial matrixes as constants but it didnt work either.
Ex:  bayes = {0.0002639011, 0.0003372667, 0.0004005217, 0.0004619689, 0.0005238603, 0.0005852403, 0.000645032, 0.0007037884, 0.0007612079 and  PD_ini = j(1,36,0.01);

Note: My base SAS software is 9.4 and the IML is For SAS/IML 15.2.

Can someone give some pointers or some help. I would apreciate it 😃

[Rick_SAS · Accepted solution]

The vector for the initial guess is called PD_init in the program, but you are passing initPD (which is undefined) as an argument to CALL NLPQN.

That information (which matrix was not set to a value) is part of the log, so please copy/paste the entire error message in future posts.

[pedroy341]

Hello! Thank you a lot for your answer. This was my first post and I didn't submit it in a acceptable format. Luckly I have sorted it out.
I will make sure to be more carefully next time.

[Rick_SAS]

No worries. Glad your problem is solved!